Ferris Intelligence Before Overlord - MOD Intelligence Before Overlord 121 Expeditionary Force (...
Transcript of Ferris Intelligence Before Overlord - MOD Intelligence Before Overlord 121 Expeditionary Force (...
Ferris Intelligence Before Overlord
119
研究会記録
Intelligence Before Overlord: Knowledge and Assumption in Allied Planning for the
Invasion of Normandy, June 1943-June 1944.
John R.Ferris
Allied intelligence for OVERLORD usually is seen as a success story. That is the
truth-- just not all of it. From June 1943 to June 1944, intelligence revealed German
expectations, understanding and intentions. It showed the enemy was behaving as
OVERLORD assumed, how the allied deception plan, FORTITUDE, was working, and
the weakness of enemy wireless intelligence and aerial reconnaissance. The control of
MI 5, the British security service, over German agents revealed enemy expectations,
and shaped them, by enabling deception. Through Allied codebreaking, ULTRA and
MAGIC showed that German generals and Hitler exaggerated allied forces in Britain by
200% and the initial strength of a seaborne assault by 400%; remained fixed on the Pas
de Calais; but during May paid increasing attention to Normandy and strengthened its
garrison. 1 Intelligence was excellent on great issues yet mediocre about enemy strength
in Normandy on 6 June 1944. Scholars have noted these facts but not explained them.
Studies of intelligence also ignore the layer of assessment which guided planning. In
1994, Alexander Cochrane described intelligence and deception as the “missing
dimension” of studies in OVERLORD. 2 In 2013, studies of planning and intelligence in
An earlier version of this chapter was published as John Ferris, “Intelligence, A Snapshot from 6 June 1944”, in John Buckley (ed), The Normandy Campaign 1944: Sixty Years On, ( London, Routledge, 2006), pp. 185-201. Material in CAB, PREM and WO files may be found in The National Archives, London; the T and RG 165 and 331 series are held at The National Archives and Record Administration, College Park; and the RG 24 series at The National Archives of Canada, Ottawa. The Bernard Montgomery papers are held at The Imperial War Museum, London. 1 MI 14 Weekly Summaries, 15.5.44, 5.6.44, “The OKW and Allied Intentions-Apr 1944”, passim, WO 208/4312; JIC (44) 221 (0) Final, 29.5.44, RG 331/3/131; Hinsley, British Intelligence, Three, II, pp. 49-65, 790.
2 Alexander S. Cochran, “ULTRA, FORTITUDE and D-Day Planning: The Missing Dimension”, in Theodore Wilson (ed), D-Day, 1944, ( Lawrence, Ks, 1994), pp. 63-79, assesses the literature well. The best study of intelligence and OVERLORD, an excellent one, is F.H. Hinsley, E.E. Thomas, C.A.G. Simkins and C.F.G. Ransom, British Intelligence in the Second World War, Its Influence on Strategy and Operations, Volume Three, Part II ( Cambridge University Press, London, 1988). The best account of planning before Normandy, Carlo d’Este, Decision in
120
OVERLORD remain mutually exclusive. These problems cannot be solved in 16,000
words, but this paper will examine allied assessments before the invasion of the enemy’s
combat strength and order of battle in France, especially Normandy; and consider why
intelligence success and failure marched hand in hand for OVERLORD, and how far
either mattered. The intelligence record illuminates with unique force, the planning
which led to triumph in Europe, and challenges all conventional narratives of the matter.
The Struggle in Intelligence
OVERLORD involved a struggle between two sides in military rationality,
intelligence and deception. The allies won in every sphere. Military rationality led both
sides to calculate the ideal place for an allied attack would be the Pas de Calais, with its
long and open beaches on the northernmost coast of France, with Normandy second and
anywhere else a distant third; and that an attacker would have great logistical problems.
Thus, the Germans assumed any allied attack must aim immediately to seize a major
port. They spread their bets everywhere, but above all at the Pas de Calais. The allies
put all theirs on Normandy, where they faced a fragment of enemy strength and evaded
most of it; and carried a port in their hands, through the artificial MULBERRY
installations. Again, OVERLORD was the first priority for allied intelligence services,
aided by specialist organs focusing on issues like beach topography. They had good
human sources, excellent imagery and ULTRA and fused them. German sources failed
in coordination and quality. Their spies were controlled by MI 5, their imagery was
spotty, and their wireless interception and codebreaking were mediocre, missing truth
and eating lies. German commanders paid almost as much attention to deception as
allied ones, leaking false news about their strength in defences and divisions, through
many means. This effort had no impact. Conversely, FORTITUDE reinforced German
preconceptions, which the allies predicted by reason, detected by intelligence, and
exploited to victory.
In planning for OVERLORD, military rationality always came first, intelligence
second. Operational decisions were made by middle level planners at Chief of Staff
Supreme Allied Commander Designate ( COSSAC), Supreme Headquarters Allied
Normandy, 50th Anniversary Edition, ( London, 1994), offers good, but brief, accounts of intelligence.
Ferris Intelligence Before Overlord
121
Expeditionary Force ( SHAEF) and 21 Army Group, by the Supreme Allied
Commander and the chief of 21 Army Group, Dwight Eisenhower and Bernard
Montgomery, and sometimes by the British ( and to a lesser degree, American) Chiefs of
Staff (COS). The key decisions were made purely on the basis of military rationality,
with intelligence irrelevant, by COSSAC in July 1943, to avoid the Pas de Calais and
strike Normandy, tant pis; and by Eisenhower and Montgomery in January 1944, to
double the frontage and strength of the initial attack. Intelligence was secondary even in
the formulation of the initial plan for OVERLORD, though it became central when that
plan was refined and applied between April to June 1944. Most decisions emerged
through the routine interaction of mid level military bureaucrats, by their arguments,
agreements and differences.
This was especially true of intelligence. The British Joint Intelligence
Committee ( JIC ), MI 14, the German branch at the Military Intelligence Division, and
mid level figures in COSSAC, SHAEF and 21 Army Group, dominated estimates.
Specialists handled collection and deception. These officers were able, experienced, and
British. Montgomery and his chief of General Staff Intelligence ( GSI ), Edgar “Bill”
Williams, had served harmoniously for two years. Montgomery’s chief of staff, Freddie
de Guingand, thought Williams “the most able and clear-headed” GSI “I have ever met”.
From January to May, General John Whiteley, an Operations officer trusted by
Eisenhower but not Montgomery, and without experience as an intelligence chief, ran
SHAEF G-2. He was replaced just before D-Day by Ken Strong, Eisenhower’s
intelligence chief in 1943 and first choice for SHAEF G-2.3 MI 14 and JIC reports went
only to top commanders and staff officers; so too, some assessments by SHAEF G-2 and
Williams, though all staff officers and commands to division level received their weekly
estimates. During the planning for OVERLORD, MI 14 and the JIC dominated
analyses of strategic issues and enemy organisation, while Whiteley and Strong
harmonised all sources for their commander, and his forces. After D-Day, SHAEF G-2
dominated intelligence in the west, seizing power from MI 14 and Williams, but until
then it was a fifth wheel. 21 Army Group dominated operational planning, and Williams
the intelligence for it. 3 “Notes taken at a meeting of Army Commanders and their Chiefs of Staff, at HQ 21 Army Group, 7 Jan. 1944”, WO 205/16; Francis de Guingand, Operation Victory , ( Hodder & Stoughton, London, 1947) p 106; Nigel Hamilton, Montgomery, Master of the Battlefield, ( McGraw Hill, New York, 1983), p. 144.
122
Tensions there were within intelligence, because of opinion, personality and
tangled lines of command. In October 1943, referring mostly to Ultra, COSSAC
complained it was not receiving “all the essential intelligence available in the War
Office”. Its access to such material steadily rose. By January 1944, SHAEF and 21 Army
Group, and later First United States Army Group (FUSAG), received as much ULTRA
as commanders in Italy, and more JIC material. 4 Again, in February, the senior
operations officer at SHAEF thought Williams was addressing “a problem of a much
larger nature than any with which he had been previously faced”. 5 Yet they also pulled
together. Williams told Whitely, his senior in hierarchy but junior in experience, “I do
not want to be preaching a contrary doctrine to yours, for I feel there is real value in an
agreed text. If we are to be wrong, let’s all be wrong together. At least then our
Commanders will not have had muddled counsel. You will remember the loss of
confidence in the Middle East caused by the internicene but public disputes between the
“I” people which helped nobody, least of all the disputants”. 6 These officers avoided
disputes, but agreed texts were hard to find, because authorities confronted a situation
up to twelve months away. They made and argued over predictions as much as
assessments; all strove for accuracy, each understood its limits. COSSAC’s July 1943
plan for OVERLORD noted,
As it is impossible to forecast with any accuracy the number and location of
German formations in reserve in 1944, while, on the other hand, the forces
available to us have been laid down, an attempt has been made in this paper to
determine the wisest employment of our own forces and then to determine the
maximum number of German formations which they can reasonably overcome.
Apart from the air situation, which is an over-riding factor, the practicability of
this plan will depend principally on the number, effectiveness and availability of
German divisions present in France and the Low Countries in relation to our
own capabilities. 7
4 Minutes of meeting on “organisation of Intelligence for C.O.S.S.A.C.”, 3.10.43, WO 171/19; Whitefoord to Chief of Staff, 19.1.44, SHAEF/19BX/INT (SP), RG 331/1/58.
5 West to Whiteley, 17.2.44, RG 331/12/9. 6 RG 331/12/9, Williams to Whiteley, 2.2.44 7 PREM 3/342/2/8, C.O.S. (43) 416 (O), “Operation ‘Overlord’, Report and Appreciation”, 30.7.43, p. 4.
Ferris Intelligence Before Overlord
123
Commanders and planners thought less about Germany’s current strength in France,
than how it would meet amphibious attack, of its power relative to allied forces in recent
battles, its uncommitted forces across Europe and their speed of deployment. When
making these decisions, allied planners used whatever intelligence they had. It served
them well. Yet between the time plans were made and executed, much might change;
and many key points did not lie simply in the sphere of intelligence—like how good the
enemy would be, and how it would prepare its defences and counter attack. The value of
intelligence was defined by its ability to support planning, which in turn was based on
predictions of power and military rationality.
This process produced characteristics in preparations for Normandy.
Intelligence and planning never were combined effectively on paper, only in the minds of
commanders. The planning applied on 6 June 1944 rested on old estimates, not current
ones. Most planning was done by April, reflecting estimates of that era; Williams’
weekly assessment of 4 June reached generals only as their units were at sea. Even
more, these estimates and plans recognised a wide range of possibilities. Assessment
had to be fluid, because enemy strength and intentions were changing, and rapidly. The
allies originally planned to attack on 1 May. Had they done so, Normandy would have
had barely 50% its garrison of 6 June, which in turn might have received the extra few
divisions needed to force the postponement of OVERLORD or to defeat the operation,
had the attack waited until 1 July. Enemy intentions and capabilities were not distinct
matters, but fused—what Germany would do, was what it could do. If the enemy
guessed right, it could defeat OVERLORD, or deter it. Meanwhile, a dynamic tension
emerged between worst-case and better-case assessments. Worst-case logic has a bad
reputation, because in open-ended circumstances it causes timidity and costs
opportunity. Yet in OVERLORD, uncertainty on basic issues and the need to control
risks were unusually high. Worst cases were hard to avoid, and took an odd role in
planning. Caution was the better part of error, doubly so because planners and
commanders recognised they were applying worst-case logic, tried to minimise its
impact, and sought to understand how Germans ( not allies) would act. The tension
between worst case logic and attempts to control it, in one wave after another, marked
all planning for OVERLORD.
At the highest of levels, the War Office, JIC and COS took care to avoid
presenting worst case estimates to the British Prime Minister, Winston Churchill,
124
presumably so not to provoke an effort to cancel the operation. They emphasized that
German air and land forces exceeded the conditions which COSSAC originally had
defined for success, but also were less formidable than their numbers indicated, while
allied strength in the assault had risen more. 8 Otherwise, higher layers of authority
outlined worst cases, primarily for the sake of form; which subordinates received, but
regarded as beyond their concerns; with the key cogs in the system remaining free to
assess and act as they chose. These worst cases had little effect at the time, but remain
embedded in evidence, ready to trap unwary scholars. They shape conventional ideas
about key aspects of planning before OVERLORD. Thus, Montgomery’s celebrated
briefing of generals at St. Paul’s School on 15 May 1944, treated as the touchstone for
expectations before 6 June 1944, made points he knew were misleading. He rated
German strength more than his own intelligence agency did. He credited them with 22
assault divisions in France, ten Panzer and 12 Field Infantry. He forecast that between
D + 6 and D + 8, 24 divisions would surround the bridgehead, including 10 Field
Infantry divisions, and also 10 Panzer divisions launching a “full blooded counter attack”.
9 These figures matched the very worst case assessments of MI 14, SHAEF and the JIC.
They were twice the level of the COSSAC conditions, and also of the figures reckoned by
Williams, which governed planning in 21st Army Group. Presumably Montgomery
delivered this overestimate because these were the figures defined from above: to use
them caused no problems, to use anything else would raise questions. They aided
showmanship, and spurred preparations, making generals sweat more over training,
only to find operations easier than expected. They gave him even greater credit for
victory, when it occurred. None the less, this assessment does not reflect Montgomery’s
expectations of German numbers, nor his planning, though his prediction at St. Paul’s of
what German forces would seek to do between D + 1 and D + 8 was prescient and honest.
That numerical overestimate, and the inadvertent but widespread misinterpretation of
the meaning of the map that 21 Army Group displayed about the expansion of the
bridgehead between D Day and D + 90, which was intended for logistical purposes but
understood to predict a steady advance, meant that this briefing misled its audience
about key issues, and later commentators as well.
8 Hinsley, British Intelligence, Three, II, p. 713. 9 Nigel Hamilton, Master of the Battlefield, Monty’s War Years, 1942-1944, ( New York, McGraw-Hill), pp 582-9.
Ferris Intelligence Before Overlord
125
The Allied Assessment of German Forces
COSSAC assumed that the quality of enemy units would remain constant and
the enemy would use them as well as it could. COSSAC’s July 1943 assessment of the
air aspect of OVERLORD was a classic statement of worst-case planning:
The following estimates and appreciations are based on factual intelligence and
consider the scale of enemy air effort possible under ideal conditions, not taking
into account any effect that the present Allied air offensive may have on the
G.A.F.’s fighting value in the future...Furthermore, the rates of reinforcement
are those of which the enemy is physically capable, considering the scale of
efficiency of his organization as a whole, and do not take into account the effects
of any counter-action by Allied forces. The scale of effort envisaged assumes also
that the enemy would concentrate the maximum strength possible even at the
expense of adequate defence elsewhere. 10
In this instance, worst-case logic was accurate: during June 1944 the Luftwaffe did
throw most of its remaining aircraft across Europe into the battle of Normandy. Had
that logic been applied to land forces, planners would have had to abandon
OVERLORD; here, the worst case centred on the nature and the number of “full-
strength first quality divisions” Germany could use against OVERLORD. In July 1943,
COSSAC concluded that OVERLORD had a “reasonable chance of success” only if
German forces and defences in Normandy, its reserves in the west and all aircraft
deployable there, did not rise above the level of 30 July 1943, including 12 “full-strength
first quality divisions” in France; if no more than three such formations joined the local
forces of three infantry divisions on D-Day, five by D + 2 and nine by D+ 8; and if no
more than 15 good formations moved in from other theatres by D + 60. 11 These
calculations approximate the level of German forces actually in Normandy between D-
Day and D +8. Their effect in combat confirms the accuracy of COSSAC’s predictions.
Soon, however, the quality of German forces sagged, and were seen to be
sagging. In July 1943, COSSAC calculated Germany had 1740 first-line aircraft in the 10 PREM 3/342/8, C.O.S. (43) 416 (O), “Operation ‘Overlord’, Report and Appreciation”, 30.7.43,
pp. 1, 69 11 PREM 3/342/8, C.O.S. (43) 416 (O), “Operation ‘Overlord’, Report and Appreciation”, 30.7.43,
pp. 25, passim
126
west, but its key concern was the “steady rise” in fighter strength, which must be
“checked and reduced” before invasion. Air intelligence estimated that between 1
January and 1 August 1943, German first-line strength in fighters had grown by 245
aircraft, to 1340 ( or 2260, counting second-line reserves), and that in the west from 305
to 600 machines. 12 In February 1944, planners expected to confront 1650 first-line
German aircraft on D + 1, perhaps soon joined by 950 more but, like COSSAC, they
noted the Luftwaffe’s lack of “appreciable depth” and inability to sustain losses. Soon,
SHAEF thought German fighter pilots “vastly inferior in quality” to American ones. By
May, the British COS agreed that the Luftwaffe’s strength exceeded July 1943 levels, by
5250 to 4870 aircraft, including 2700 fighters versus 2175 ( first-line and all reserves),
but this was irrelevant: its decline in quality, production and reserves changed the
meaning of the numbers. 13 These figures still distorted German air strength— the
Luftwaffe did throw most of its fighters into OVERLORD, but they numbered barely
1300—yet that problem was minor. All planners abandoned the worst case about the
Luftwaffe save Trafford Leigh-Mallory, commander of the Allied Expeditionary Air
Force; this division provoked needless debate over any part of OVERLORD related to air
support, crippled the use of airborne forces and made decisions imperfect, but not
seriously so. Worst case logic ruled the equation at air and sea, but led to little worse
than wasted effort. Given the circumstances confronted by allied air and naval forces—
with forces to spare, and superiority over the enemy in vital areas, where risks could not
be run—worst case logic was the best means to define the level of insurance. For a quick
and single premium, the cost was cheap.
The story was different regarding assessments of the German army. As
COSSAC wrote, the issue was not so much “the precise number” of enemy divisions as
their “effectiveness”: if they “are below strength or of low morale, or if German ability to
12 CCS 309, 15.8.43, PREM 3/333/15. 13 Fourth Draft, 30.1.44, “Neptune, Initial Joint Plan”, WO 205/15, 1.2.44, Neptune, Initial Joint
Plan, WO 171/126; SHAEF Weekly Intelligence Review, No 6, Part II, RG 165/79/2566; PREM 3/342/10, COS to Churchill, 23.5.44, “Opposition to Overlord”; F.H. Hinsley, E.E. Thomas, C.A.G. Simkins and C.F.G. Ransom, British Intelligence in the Second World War, Its Influence on Strategy and Operations, Volume Three, Part II ( Cambridge University Press, London, 1988), pp. 35, 103-125. The discrepancies between these figures occur because allied estimates of what German air strength would be on 31 July 1943 changed, while two different issues were being assessed, the “initial establishment” of first-line units, and total Luftwaffe strength, including second-line reserves.
Ferris Intelligence Before Overlord
127
move them is reduced, we can face and defeat a proportionately larger number”. 14 When
assessing the balance between the allied and German build-ups at Normandy, a major
figure in the Operations section at SHAEF, General H.R. Bull, noted, we are tempted to
look on the enemy forces as fully-equipped, fully-trained, battle experienced 1940
divisions, instead of regarding them as 1944 divisions, diluted by foreign elements, and
in several cases far from being up to strength or effectively trained... I feel that the
whole difficulty in estimating the enemy’s course of action lies in appreciating what real
strength we are up against as opposed to theoretical strength. This, I think, must be
evaluated in the monthly estimate of the rate of reinforcement, to avoid giving us the
one-sided impression a purely logistical calculation is apt to do. 15
Great problems emerged in understanding the enemy’s real strength. Several good
sources, ULTRA above all but also agents in France and imagery, kept an extraordinary
grip on the enemy’s order of battle. From August 1943 to June 1944, the number of
German formations in France constantly shifted, as did Allied estimates of them Until
April, when SHAEF thought its information “unsatisfactory”, they included several
German formations which were not in that theatre. 16 Estimates were filled with
speculation about the value and name of formations. Given the nature of planning and
the situation, and German security and deception, these problems were hard to avoid
and errors often cancelled each other out. At any time, allied intelligence was good on
the identifications, numbers and locations of formations in, entering or leaving the
theatre, and almost perfect on these matters in the weeks before D-Day. It gave
commanders accurate information they trusted, reinforcing their faith that enemy
strength still was within the COSSAC conditions. 17
Yet, simply to identify the quantity of divisions was not enough; their quality
was of equal importance. Virtually every division in France was rebuilding, some
starting in early 1944 from just 10%-25% their establishment of equipment and trained
soldiers; to determine how good they would be on the day, one needed to know not just
their current strength, but how far and fast each would improve. Two types of
formations proved especially hard to assess. “Training” divisions, infantry and armoured,
14 PREM 3/342/8, “Overlord”, p. 10. 15 331/29A/119, Bull, AC of S, G-3, to AC of S, G-2, SHAEF/17100/12/Ops, 24.2.44 16 SHAEF to AGWAR, S-50558, 21.4.44, RG 331/1/114. 17 Material on these matters is contained in RG 165/79/2566, RG 331/29A/119, RG 331/12/9, RG
331/1/114.RG 331/1/59, WO 208/4312 and WO 171/102.
128
had an uncertain value: though usually that title described their function, to provide
drafts to other formations, rarely they did serve in combat, and over time their rating
could rise. Meanwhile, the allies ranked German infantry divisions either as Field
Infantry, capable of independent assault in mobile operations, or Lower Establishment
( LE), defensive formations ranging from poor to decent in quality, about half having just
two instead of three regiments, including some “Ost” battalions ( with German officers
and ex-Soviet soldiers). The allies recognized another, and worse, category, “Static”
divisions, but believed none were in the west. In any case, they had mixed success in
determining the real strength of most German infantry, and also of the best enemy
forces.
ULTRA illuminated the manpower and training of Panzer, Panzer Grenadier
and Paratroop formations, but not their equipment. ULTRA and the Combined Service
Document and Interrogation Centre ( CSDIC ), which collected and assessed material
from these two sources, offered good material about LE divisions. 18 The allies’ grasp of
these formations was sound, no easy task, yet not perfect. Conversely, intelligence on
Field Infantry was lacking, and estimates of their value were arbitrary and erroneous.
These problems were fundamental to assessment, producing worst case assessments
which drove far into the heart of planning, though irrelevant on the battlefield,
fortunately. The JIC’s last full estimate of German strength in France identified 14
Field Infantry divisions. It overrated by 33% the manpower of four of them, and that of
eight others and most LE formations by 10%. The JIC slightly underrated the quality of
the best infantry in France, 3 Paratroop Division, which also had 33% more soldiers
than expected—18,000 men, twice the strength of the average infantry division in
France. With that exception, every other so-called Field Infantry formation had just
50%-66% the manpower of average ones of 1939. 19 Only one of them, 5 Paratroop
Division, had the strength even of what in 1943 the British termed “ 2nd Quality
Infantry Divisions” ( 12,000 soldiers), while four were much weaker than “ Defensive
18 RG 165/179/660, C.S.D.I.C. ( U.K.) S.I.R.s No 110, 115, 117, 119, 236, 256, 277, 340; RG
165/179/659, C.S.D.I.C. ( U.K.) No 293 19 Compare RG 331/3/131, JIC (44) 215 (0) 25.4.44, to the figures in David Westwood, “The
German Army in France 6 June 1944”, paper presented to the conference on Normandy, Wolverhampton University, July 2004.
Ferris Intelligence Before Overlord
129
Infantry Divisions” ( 10,000 men). 20 Manpower was only one indicator of combat value,
but a central one, and the allies knew German infantry placed increasing reliance on
organic firepower. All told, allied intelligence distorted the combat quality of German
infantry—in reality, only one Field Infantry division stood in France, Third Paratroop.
The rest were just slightly better than usual LE formations, while many of the latter
were far worse than even that title would indicate, Static divisions. Fewer offensive
divisions fought at Normandy than expected, in part because the Germans had fewer of
them in France than the allies estimated. This worst case logic distorted assessments
about the effect of FORTITUDE at the time, and afterward, because observers could see
no other reason for Germany to retain forces on the Pas de Calais after D-Day, and
exaggerated their value, especially of the 18 LE divisions left on the Pas de Calais after
I July.
A little intelligence and knowledge of standard enemy practices provided a fair,
if overstated, account of the number of flak and anti-aircraft guns which could be
deployed at Normandy, calculated around 1300. 21 This approach failed in another
sphere. In its greatest technical failure before D-Day, allied intelligence did not know
the number of tanks in Panzer Divisions, or in France, and knew it did not. This failure
stemmed from limits in ULTRA, which did worse in tank counting than ever since 1941,
because changes in the way German formations reported their tank strengths stymied
the codebreakers at Bletchley Park. The latter solved messages, but could not
understand their meaning. Bletchley began to overcome these problems on D-Day itself,
but all prior calculations rested on guesses. 22 These took many forms. Allied planners
knew the number, location and manpower of armoured formations in France, but errors
over tank strength crippled views of their combat value. The Germans had fewer tanks
than allied intelligence feared, but more than allied planners assumed. On 6 April 1944,
ULTRA showed that the German commander in chief in the west, Gerd von Rundstedt,
aimed to make his three Panzer training divisions combat worthy as quickly as possible.
Williams, thinking this “a desperate improvisation”, never took them seriously. He did
not formally assess their value but at most, he seemingly thought each of these divisions
might provide a weak but combat worthy battlegroup, of regimental or brigade strength.
20 Appendix ‘E’, COSSAC (43) 4, RG 331/3/124 21 JIC (44), 215 (0), 25.5.44, RG 331/3/128. 22 Ralph Bennett, Ultra in the West.
130
Before von Rundstedt’s message, the JIC rated these formations and all non-divisional
armour as together matching two divisions in strength, in which calculation training
units were a minor factor. In early April, conversely, MI 14 expected each Panzer
training division to be fully operational by D-Day. By 30 April, the JIC rated them as
equal to two operational divisions; by 25 May it thought only “elements” of each “are at
present capable of employment in an offensive role”. In fact, they did not fight until the
last days of Normandy, with mediocre quality. 23 Yet throughout this period, these
formations always were counted at face value on the books, crediting the enemy with
three ( or 43% ) more Panzer/Panzer Grenadier Divisions in France than the seven
which were there on 6 June. Hard intelligence on these formations was scanty— one
training division had no Panther battalion and was short of motor transport; 2 SS
Panzer and 17 SS Panzer Grenadier lacked motor transport while 12 SS Panzer did not;
“slight evidence” suggested 21 Panzer had two tank regiments ( in fact, it had a weak
regiment). 24 The absence of intelligence drove the allies to predict German tank
strength by their understanding of enemy establishments. This approach caused
overestimates, as the allies understood these establishments no more than the Germans
followed them.
MI 14, thinking it knew from ULTRA the formal establishment of Panthers in
Panzer Divisions but ignorant of their real strength, assumed all but training divisions
had a full strength battalion with 81 Panthers, unless it had proof to the contrary,
reversing the analytical process used to determine the number of Panzer Divisions in
France.25 MI 14 and SHAEF’s grasp of the strength of Tigers and Panthers in non-
divisional units was accurate enough ( estimated as 135 Tigers and 283/320 Panthers,
against in reality 102 and 250 on 6 June, or the armoured elements of three Panzer
divisions). They were, however, wrong on important elements of German organisation.
They did not know 21st Panzer Division lacked its Panther battalion, while those of two
armoured formations and a training one were at 50%--66% the strength assumed by the
allies; and thought four assault gun battalions were additional non-divisional units,
rather than being part of formations. Predictions of strength derived from
establishments could take many forms. If one added the figures for non-divisional 23 Hinsley, British Intelligence, Three, II, pp. 72-77,814. 24 JIC (44) 215 (0) 25.5.44, RG 331/3/131. 25 “Allotment of Panthers to Armoured Divisions in the West”, MI14 /Apprec/8/44, 18.5.44, WO
208/4312; JIC (44) 215 (0) 25.5.44, RG 331/3/131.
Ferris Intelligence Before Overlord
131
armour and assault guns, to seven Panzer divisions on what was taken to be full
establishment, plus one Panther regiment above establishment, and three training
divisions weak in Panthers, one had 1830 armoured fighting vehicles ( AFVs)—or 1560,
if the training divisions were entirely removed from that figure.
In reality, even the low ball projections exaggerated the problem. On 6 June
1944, in France the Germans had 1891 AFVs , including 179 captured ones, 39 Mark
IIIs ( for command purposes), 758 Mark IVs, 102 Tigers, 655 Panthers and 158 assault
guns. Of these, 26% had limited combat value—captured tanks, Mark IIIs, and the 253
Mark IVs and 40 Panthers in training divisions—leaving 1420 battleworthy AFVs. In all
fairness, German strength was hard to determine. Allied estimates of it varied wildly,
and many were grossly wrong. Just before D-Day, Third US Army thought the Germans
had 1750 to 2600 tanks. FUSAG, relying on estimates of enemy strength from February,
assumed each of the seven Panzer and Panzer Grenadier Division it thought were in the
theatre would have what it took to be a full complement of 160 tanks, but ignored non-
divisional forces and training divisions, for a total of 1120. MI 14 , the JIC and SHAEF
thought the Germans had 800-850 Panthers alone, though they could not offer “a
reliable estimate”, and “may possibly” have a total of 3000 tanks—159% above the true
strength. 26 The figure of 1120 came from knowledge of average German complements;
so too that of 3000, added to every piece of certain and uncertain information, including
aircraft sightings of empty flat cars, commonly used to carry Panthers on trains, moving
eastward toward the Rhine, indicating they had done so in an unobserved inward trip.
At the strategic level, imagery on train movements created more mysteries then it
solved.
Allied estimates of German tank strength were wrong, but in many ways, not
one. Calculations of enemy armour were all over the map. On 6 June, FUSAG and 21
Army Group overestimated this strength and SHAEF underestimated that to be found
in Normandy. They led units to underrate the armour before them at H hour, but to
overrate that which could arrive there by H + 24. The problem was uncertainty, rather
26 After Action Report Third U.S. Army, I August 1944-1 May 1945, ( Scholarly Resources), Reel
One, Chapter 2, p 10; “Operation Neptune”, FUSAG, 20.5.44, First United States Army Group, Report of Operations, 23 October 1943-1 August 1944, Reel Two, Book Two ( Scholarly Resources). “Allotment of Panthers to Armoured Divisions in the West”, MI14 /Apprec/8/44, 18.5.44, WO 208/4312; JIC (44), 215 (0), 25.5.44, RG 331/3/128.
132
than fixed error. It had little impact at the strategic level, since the issue was academic
and the mistakes cancelled each other out: but it did cause problems at the tactical level.
The Allies Assess the German Reaction to Invasion
The issue was not just the enemy's total and real strength, but how much of it
could intervene in the battle, and when. This matter was central to planning for
OVERLORD. Allied planners were split on this point, with almost everyone grossly
overestimating the problem; fortunately, Williams came close to truth, and 21 Army
Group acted on his assessments.
The allies were uncertain about the enemy’s true combat strength in France and
its likely one at the front on the day of battle, and those which would follow. In October
1943, COSSAC thought that if the attack were launched immediately with good security,
given the number of “full strength, first quality” German divisions in France, assessed
around 10, their weak establishments and the effects of wastage and interference with
their movements, on D Day the allies would confront local forces and two Panzer
Divisions ( with the real strength of one); and by D + 4, three Panzer Divisions and four
infantry divisions ( with the real strength of 1.5 and three, respectively). 27 This
assessment may have been accurate, but this situation was changing; the questions
being, how much and how fast? In February 1944, SHAEF and 21st Army Group agreed
that “when the flag falls”, the chief German commanders, Erwin Rommel and von
Rundstedt, would have 20 divisions able to attack in mobile operations ( roughly 12
armoured or mechanised, and 8 Field Infantry). 28 The quantity of divisions assessed at
this quality rose toward 20, and passed it by late May, in a confusing flurry of estimates,
often contradictory in details. MI 14 and the British COS expected the Germans to have
17-21 divisions able to conduct mobile warfare, ( equaling 12-16 “full strength and first
quality” formations in value). The JIC predicted 16 to 27 good ones, though between 20
and 25 May, without explanation, its estimate of the number of Field Infantry divisions
fell from 16 to 14. On 15 April, SHAEF G-2 thought the enemy had eighteen offensive
divisions in hand ( 8 Panzer/Panzer Grenadier and 10 Field Infantry) and 28 ( 10
27 “Operation ‘Overlord”, Calculation of Enemy Rates of Reinforcement-Conditions of –1 October
1943”, WO 171/19. 28 Whiteley to West, 3.3.44, RG 331/12/9.
Ferris Intelligence Before Overlord
133
armoured and 17 infantry) on 2 June; both cases included the three training Panzer
Divisions. Beyond this, allied planners rightly agreed that non-divisional units held a
divisions’ worth of Tiger tanks and two of Panthers. 29 By the absolute worst case, in
France the Germans might have the equivalent of 14 armoured divisions, with 3000
tanks, and 17 Field Infantry divisions, on D Day—far above COSSAC’s conditions and
enough to abort the invasion. Against these calculations, planners knew these numbers
meant less than they seemed—SHAEF G-2 emphasised “the old question of the goods in
the shop window and the empty shelves behind, without which no appreciation of enemy
capabilities would be complete” 30--and doubted all of these forces would strike the
beachhead. They did not expect their worst case, neither could they ignore it, nor was
their better case pleasant, especially because of the way they assumed the enemy
machine would function.
In Tunisia, Sicily and Italy, German commanders responded aggressively and
immediately to seaborne invasion. At Salerno and Anzio, the Germans succeeded in
sealing allied forces on the beachhead. One had to expect similar efforts in Normandy,
which the enemy publicly predicted would occur. In a press interview, which Williams
distributed to his readers, von Rundstedt warned that his coastal defenceshave depth
and cannot be outflanked. In their rear there exists a system of field fortifications and
strongpoints. Mining, swamping and flooding obstacles and tank walls are so manifold
that they could not even have been dreamt of by those who build the Maginot Line.
Water obstacles along the beach for enemy landing craft and a wide mined belt on the
beach would make a landing difficult even in its initial phase before the enemy had even
been able to gain a footing. Against expected large-scale landings by airborne forces in
the rear, measures have been taken long ago, but they must be kept confidential. In my
sector there is no possibility of side-stepping or retreat. The coast line and the deeply
echeloned fortifications can be held to the last. Thus it is possible that landed enemy
forces will be split up, because they must fight for every single fortification. Numerous
vast reserves, above all Panzer and mechanised divisions, are kept in readiness in
29 Chiefs to Staff to Churchill, 23.5.44, PREM 3/342/10; JIC (44), 215 (0), 25.5.44, RG 331/3/128;
SHAEF Weekly Intelligence Summaries, Nos 6, 7, 9 and 10, 29.4.44, 6.5.44, 20.5.44, 27.5.44, RG 165/79/2566; SHAEF to AGWAR, S-50365, 15.4.44, S-53053, 2.6.44, RG 331/1/114; 21.5.44, “Statement as to the PZ, PZ Gren and Field Divisions in the west in terms of full strength first class divisions:”, WO 208/4312.
30 SHAEF Weekly Intelligence Summary, No 6, 29.4.44, RG 165/79/2566
134
groups in such a way that within a few hours they could make a powerful counter-attack.
The Germans do not indulge in the tired Maginot spirit. We conduct our fighting by
making use of strong fortifications and of other means which, for quite obvious reasons, I
shall not divulge. 31
Some of these statements reflected von Rundstedt’s campaign of deception-- after the
war, he claimed to have believed that the Atlantic Wall was an “unmitigated fake”
which any attack quickly would penetrate—but the comments on counter-attack did
reveal how he expected to fight, until he was frustrated by other commanders, German
and allied. 32 These comments reached the highest levels of allied command, and
reinforced their greatest area of uncertainty. This knowledge first divided allied
planning, and then drove it to adopt measures which overcame the problem, and the
enemy. It overcame a hard case by treating it as a worst case.
The allies had little intelligence on how German command would meet an
attack. A fair amount of material from ULTRA, captured documents and prisoners
showed that divisions in France responded to emergencies with unpleasant speed. Units
sometimes were on the move 30 minutes after receiving orders, regularly within 90; as
Williams emphasised, “troop trains are kept standing by in important divisional areas”
while “time-tables are prepared”; probably, several entrained divisions could move at
any one time from Russia to France in 7-10 days. 33 This evidence showed the best the
machine could do when it worked well, but not whether it would do so. The allies neither
understood nor pretended to understand German command. They knew personal
politics might affect decisions—Whitefoord doubted that Rommel and von Rundstedt
“would work harmoniously, as their outlooks are so different”, Montgomery that
“quarrels might arise between the two of them”—but did not bank on it. 34 Allied
analysts overestimated the rationality and autonomy of generals in France and
underrated their rivalries. They knew Hitler hated to abandon ground but
underestimated his interference. The allies assumed all enemy armour would be
31 “GSI 21 Army Group, Weekly Review, No. 2”, 20.2.44, WO 171/102. 32 Interrogation of General Gerd von Rundstedt, 4-14, 1.2.46, WO 205/1020 33 Sherrington to Fass, 230/4/42, 3.8.43, memorandum by FUSAG G-2, 8.12.43, “German
Movement of Reinforcements in the West”, passim, RG 331/12/9; 21 Army Group Weekly Intelligence Report No 9, 17.1.44, WO 171/102
34 COSSAC “Intelligence Review, 17.12.43, RG 331/12/13; Brief by Montgomery, “Brief summary of Operation OVERLORD as affecting the Army”, 7.4.44, BH Montgomery Papers, No 74, Imperial War Museum.
Ferris Intelligence Before Overlord
135
commanded by Armies, throwing all their forces immediately at the beachhead, and by
Army Groups or OKW, acting after thorough and rapid thought. They read German
reactions from Rommel, his characteristics and his proclamations of smashing invasion
on the beaches. These assumptions were reasonable, unavoidable, but wrong, and
shaped by cultural and military ethnocentrism, and mirror imaging. MI 14 described
German command as a rational bureaucracy, rather like that of the allies. Hitler was
Supreme Commander, but his power was neither defined nor discussed. The chief
soldier at OKW, General Keitel was “in effect, a Minister of Defence”. Its General Staff
handled “the general conduct of the war, and the main outlines of strategy, and is
responsible for operational planning”. Its role was “not exactly known”, butthe procedure
is however believed to be as follows. Once an operation has been decided upon, the
General Staff of the Armed Forces prepares plans in a broad sense. It then selects a
commander to carry out the plan. He is chosen with special regard to his experiences
and qualifications, and may belong to any of the services. He is, except probably in the
case of a large-scale campaign, allowed to chose his own force and his own subordinate
commanders. He defines his wishes to the General Staff of the Armed Forces, which
pass on his requirements to the Service ministries, leaving them to work out the
detailed plans in conjunction with the commander and his staff. 35The Germans would
have done better at Normandy to act that way, but they did not. The allies expected
German command at least to match their own; they were not.
Allied planners tried not to exaggerate this matter, but still they did. In late
1943, COSSAC’s Operations Branch noted “the Intelligence Branch have determined to
show the worst possible picture in order that they may be on safe ground”. The
GERMAN forces move with mechanical precision, without hesitation, without error of
destination, and without interference to the right locations where they arrive, deploy
and attack at the right time on the right plan. Even admitting that the nearer reserves
will know of and have practiced their counter attack role, does that means that it can be
carried out as if every allied move had been foreseen months before, or that the later
reserves will be equally conversant with the terrain and the situation? If so, we may as
35 “The Supreme Command of the German Armed Forces ( Oberkommando der Wehrmacht-
O.K.W.)”, 12.5.44, MI14/g/Apprec/4/44, WO 208/4312.
136
well call off ‘OVERLORD” and try an operation which we can keep secret from the
enemy more successfully. 36
In January 1944, General Morgan called for an estimate of the German response based
on “what I may call a rather more military, rather than mathematical, angle. Up till
now, we have always made a comparison of two sums in simple addition, ours and theirs.
I think we should now interest ourselves in forming some estimate of the way in which
the enemy will handle the resources that have been so carefully deduced by
mathematical means as being available to him, at various times in relation to D day”. 37
Planners agreed about how the Germans would react. Complete surprise was
unlikely for OVERLORD—impossible for an attack much after first light. Assault at
night or dawn might provide “a limited degree of tactical surprise”, but even so, several
hours earlier the enemy would know attack on north-west France, and probably
Normandy, was imminent, by finding convoys at sea before dusk. 38 In January,
Eisenhower noted the “extreme importance of the enemy not having even 48 hours
warning of the actual points of assault”. 39 Planners were more demanding and
optimistic. They expected the enemy to realise Normandy would be attacked four hours
before the assault ( H - 4) , perhaps more; to remain uncertain for some time whether
this was the only assault, or if its target was Le Havre or Cherbourg, the major ports on
either side of the Normandy peninsula; but quickly to regard it as a major threat. Then
trouble would start, the only questions being, how much and how soon? The answer was
measured in hours, and divisions.
Staff at COSSAC, forced to work with a small beachhead and rating enemy
quality high, always doubted the attackers could survive the planning condition of a
36 illegible minutes by G-3 Ops Div, 26.10.43, COSSAC/00/6/15/Ops, RG 331/29A/119 37 RG 331/1/58, Morgan to ACOS G-2, 4.1.44. 38 Appendix B to NJC 5 ( Final), 14.12.43, WO 205/12. This version abandoned the view of the
preceding draft that an attack at night or dawn might gain tactical surprise, perhaps even preventing enemy moves from occurring before first light, when allied troops “”’touch down’” and the enemy “can estimate the strength of our convoys”, Appendix “B” to NJC (4) Final, 14.12.43, WO 205/12. In fact, the latter situation did occur on D-Day, because of the decision to land in bad weather, combined with German incompetence. Though initial warning of the allied attack came just after midnight on 6 June, largely through radar reports of ships at sea and news of allied airborne landings, local German commanders did not begin to act for almost ten hours afterward, and took almost as long to execute these actions, while decisions by senior commanders about the release of reserves were even worse.
39 Memorandum by COS for Churchill, 5.2.44, PREM 3/345/4.
Ferris Intelligence Before Overlord
137
counter-attack by five divisions on D + 2.40 That reaction was universal, leading senior
authorities to favour an expanded beachhead, new commanders and revised
assessments. In February 1944, 21 Army Group planners expected local German forces
to react instantly at H – 4, but hoped those elsewhere would not. Williams wrote, “In my
view the decision to go the whole hog and risk everything on turning us out of the
NEPTUNE area cannot be taken earlier than evening D + 2. Until then whether
immediately available reserves are going to be enough for the task cannot be evident.
After that it must be neck or nothing, and he may then be prepared, if forced, to sell out
of the no longer vital areas of the West”. Most mechanised divisions should reach the
front by D + 4. Williams envisaged the “horrifying prospect” of a “full-blooded counter-
attack” on D + 6 by all mechanised forces in France plus whatever infantry was in the
area, which he then rated at 12 and three divisions. 41 At SHAEF, Bull thought the
enemy “will react almost automatically and according to pre-arranged plan with his
Army Reserves and very promptly with Army Group Reserves close behind, accepting
only the necessary delay required for preparing his estimate of the situation”. Whiteley
too accepted Williams views about what would occur at H-4 and D + 2. He thought local
forces would react immediately, while all mechanised divisions in the west would reach
Normandy by D + 7. 42 By April, Whiteley assumed “the enemy counter attack forces”,
based at Brest, Normandy, Le Havre and the Pas de Calais, “laying back from the
coastal area, will have perfected and rehearsed plans to deal with the various situations
likely to arise. These divisions, therefore, will move swiftly to the threatened point once
orders have been given committing them to the battle”. Between D + 1 and D + 3, local
forces would attack Normandy, where all other reserves would arrive by D + 7, and
decision would begin. 43 The final edition of “Operation Neptune” expected a “piecemeal
counter-attack initially... at points where the assault is making good progress”, rather
than “a co-ordinated counter-attack”. Divisions “arriving later will be employed in a
coordinated counter-attack according to preconceived plans. The crisis of ground battle is
40 COSSAC G-2, 12.1.44, COSSAC/3DX/INT, Timing of Enemy Counter-Attacks Against Neptune
Bridgehead”, passim, RG 331/12/9. 41 RG 331/12/9, Williams to Whiteley, 2.2.44; 21 A Gp/CO/INT/1101/4/15, “Delay of Enemy
Strategic Reserves”, 21 A Gp/00/74/39/G ( Plans), 21.2.44, “Neptune, Employment of Armoured Forces”, WO 205/19B.
42 Bull to D/AC of S, 5.2.44, G-2 Div, 5.2.44, “Examination of Enemy Rates of Reinforcement”, RG 29A/119; SHAEF/17100/1/Ops, by Betts, 3.5.44, WO 171/19.
43 G-2 Estimate of the Enemy Build-Up Against Operation “OVERLORD”, 5.4.44, RG 331/1/59
138
estimated between D plus 3 and D plus 6”. 44 By these estimates and Montgomery’s
initial planning, the Germans would have numerical parity against the 10 1/3 allied
infantry or airborne divisions and five armoured brigades ashore by D + 4, and the 17
and five by D + 8, and afterward. 45 So to overcome this problem, the allies boosted their
build-up between D + 2 ( 12 divisions) and D + 5 ( 15 divisions), in both cases joined by
five armoured brigades; but after D + 8, the old balance returned.
When combined, the concepts of enemy strength and behaviour produced an
alarming image of opposition to OVERLORD. Planners expected to engage whatever
forces were in Normandy, estimated initially at three or four infantry divisions, but
which reached six ( including four LE formations) and a parachute regiment in the days
before D-Day. Just before the invasion, good intelligence from many sources on these
reinforcements led planners to consider abandoning the attack on the Cotentin
peninsula and almost 30% of their assault frontage, and forced commanders radically to
reconfigure preparations there. 46 In January 1944, even after doubling the strength and
frontage of the landing, Montgomery noted “The theoretical rate of enemy reinforcement
compares unfavourably with our own build-up, and there is nothing to prevent him from
concentrating all his forces against the central assault sector, on which we are primarily
dependent for maintenance. If he succeeded in defeating us in this sector, it is unlikely
that our other landings could survive”. 47 Eisenhower feared that by “about the ‘10th
day’...pressure against OVERLORD might become dangerously strong”. 48 His chief of
staff, Walter Bedell Smith, thought “the buffer of German divisions confronting us
across the channel is just now approaching the absolute maximum we can handle”. 49
That buffer, alas, continued to swell. In February, planners estimated that by D + 6,
nine to 12 mechanised divisions would strike the beachhead; in April, American
44 20.5.44, HQ First US Army, “Operation Neptune”; p. 122, First United States Army, Report of
Operations, 20 October 1943-1 August 1944, SR, Reel One, Book Two. 45 “Build Up”, note from Montgomery to Churchill, 23.1.44, PREM 3/342/7. 46 Bradley to Montgomery 26.5.44, “Memorandum for; General Montgomery”, 381 (C), WO 205/5;
20th SCAEF meeting, 29.5.44, WO 205/12; memorandum by Bull for SHAEF COS, 26.5.44, “Implications of the Reported Enemy Reinforcement of the COTENTIN PENINSULA”, RG 331/1/76.
47 “Notes for the Commander-in-Chief for the Supreme Commander’s Meeting”, 21.1.44, WO 205/12, “Notes for the Commander-in-Chief’s Meeting with the Supreme Commander”, 21.1.44, WO 205/16.
48 Dill to COS, telegram FMD 93, 14.1.44, PREM 3/342/1. 49 D.K.R. Crosswell, Beetle, The Life of General Walter Bedell Smith, ( University Press of
Kentucky, Lexington, Ky., 2010), pp. 586.
Ferris Intelligence Before Overlord
139
intelligence in Washington predicted a strength of 7 Panzer and 10 field infantry and LE
divisions by that stage. Just before the invasion, estimates of this strength between D +
6 / D + 8 surged; the JIC predicted 16-17 divisions and SHAEF G-2, 15 ( along with
whatever remained of forces there on D-Day). 50 Counting the local garrison, SHAEF’s
estimates of enemy divisions around the beachhead by D + 8 grew from 13 to 22—above
COSSAC’s conditions, even when accounting for quality. Meanwhile, planners assumed
one infantry division per day would enter the battle from D + 8, either because they
advanced ( in the worst case, causing stalemate) or the allies did ( in the best case;
driving deep, with continuous pressure everywhere, “by alternate thrusts toward the
EAST and towards the SOUTH-WEST, we should be able to retain the initiative, reap
the benefit of interior lines, and keep the enemy from moving his reserves from one
flank to the other”). 51 This picture soon darkened. In May, SHAEF and American forces
accepted that by D + 25, the allies would be engaging or have destroyed 11-12 Panzer
and 17-26 infantry divisions, with some 20 more remaining on the Biscay, Belgian and
Mediterranean coasts. 52 Bedell Smith told another officer that “he had no misgivings
about our troops getting ashore, but gave me the alarming prediction that our chances of
holding the bridgehead, particularly after the Germans get their build-up, is only fifty-
fifty”. 53 This dark picture, however, was irrelevant, because it disagreed with the only
text which counted, at the time that mattered.
21 Army Group Solves the Problem.
As OVERLORD approached, Williams’ analyses differed from those of SHAEF,
MI 14 and the JIC. 54 They all noted the interaction between fronts, especially Russia
and France, but this issue was more central to his analyses, which were pointed,
50 JIC to SHAEF, 6.4.44, RG 331/12/9; passim. 51 21 Army Group to FUSAG, 1 US Army and 2 British Army, 21 A Gp/20651/19/G ( Plans),
13.5.44, “Appreciation of the Relative Build-Up Between Allied and Axis Forces-Operation ‘OVERLORD’”, WO 171/203.
52 After Action Report Third U.S. Army, I August 1944-1 May 1945, Reel One, Third United States Army Outline Plan, Operation Overlord, nd; SHAEF G-2 Estimate, No 2, 14.5.44, RG 331/12/9; SHAEF Intelligence Staff, Appreciation No. 3, Part II, German Conduct of the Campaign in the West”, SHAEF/CIS/102/INT, 22.4.44, WO 171/19.
53 Harry C. Butcher, My Three Years with Eisenhower, ( New York, Simon and Schuster, 1946), p. 538.
54 William’s GSI 21 Army Group, Weekly Reviews, may be found in WO 171/102.
140
optimistic, consistent and focused on what enemy offensive capabilities would ( not,
might ) be from D-Day to D + 60. On 13 February, he told his readers, “The needs of the
West await the decision of the East...German resources are stretched on all fronts and
there is little to spare”. On 2 April, “the enemy is courting further and further a
deepening disaster in the East to retain a good chance in the West, a strange gamble
militarily, made intelligible politically by the prospect of a compromise peace if the
Western decision bore fruit; in short, more and more Stalingrads in the hope of one
Dunkirk”. In February, Williams thought Germany could redeploy 15 good divisions to
France from other theatres by D + 60. That number soon fell to 13 and in April, after
devastating German losses in Ukraine, to six, with perhaps two more from Italy. These
losses, Williams thought, would force the Germans to send replacements to Russia, and
so slow the rebuilding of formations in France, where on D-Day Germany would have
only 12 to 16 offensive divisions, most likely closer to 12. On 9 April, these numbers fell
by five. A Field Infantry division, much non-divisional armour and two crack formations,
9 and 10 SS Panzer Divisions, went from France to the east, as did spare forces in other
theatres. Williams crowed, “The great decision has had to be taken... The Red Army has
forced a revision of the proposed strategy of 1944; the West has lost its priority and its
insulation ...in their doughty turn, the Russians make ‘Overlord’ more possible”. He also
held (correctly), that two Panzer Divisions previously thought to be in France were not
there, though the Germans might use their training Panzer Divisions in combat, with
uncertain effect. For Williams, as with every analyst, mid April was the most optimistic
period before D-Day, when German strength was inadequate against OVERLORD, and
its ability to grow was uncertain. From then until 6 June, like every analyst, Williams
estimate of German strength in France rose. Still, he held that on D-Day Germany
would have fewer offensive divisions than the 20 he and SHAEF had forecast in
February, probably only 12, perhaps less, and be able to transfer only half as many
formations to France after D-Day as had then been expected, or less—probably far less.
Every analyst had a worst case and a better one. Williams’s cases were more
optimistic than those at SHAEF and MI 14— SHAEF planners, heirs to COSSAC’s
tendencies toward worst case planning, thought his far too much so. 55 His worst case
was their best, because of personality and function. Williams liked clear analysis and
55 West to Whiteley, 17.2.44, passim , RG 331/12/9.
Ferris Intelligence Before Overlord
141
bold prediction; so did his chief consumer. Whiteley preferred bean counting and
narrative; under Strong, SHAEF assessments became like those of Williams, though
often their views differed. MI 14 and the JIC reached conclusions about out of theatre
transfers like Williams’, sometimes before he expressed them, but their views were more
academic-- mutable and sensitive to new or uncertain information. They, and Whiteley,
were responsible to authorities who could make only one decision—to unleash
OVERLORD or postpone it. Their job was to assess the range in which enemy strength
would fall and report if it might be too high; legitimately, they could change their minds
overnight without explanation, or define the worst case--just for the record, or to show
what might go wrong. Williams did the same when he stood in similar shoes, when
outlining German tank strength on the day for his subordinates in 21 Army Group. In
the month before D-Day, the JIC’s worst case assessment of the number of German
divisions able to reach France from other fronts by D + 60, switched from 15, to 13, to 18,
and then toward eight, as the German front in Italy collapsed; just before the assault it
warned that 3000 German tanks and 27 offensive divisions might be in France. 56 If
taken literally, such predictions would have forced decision makers to consider
postponing OVERLORD and constantly to change their plans; but they were not so
intended. In March, the COS asked whether JIC reports meant that enemy strength
was above the COSSAC conditions. The JIC’s Secretary replied that its estimates of out
of area transfers did not mean what they said—they were ”very conservative” and
“highly unlikely” to occur. SHAEF, “have not, as far as I know, ever in so many words
stated that they are prepared to meet” the reported rise in German strength in France,
but it had done so “by implication”--because it continued to plan for an invasion despite
knowing the JIC’s views! 57 The JIC explicitly presented its figures of 3000 tanks and 13
divisions transferred by D + 60 as being the very worst possible case. Williams could not
adopt such an attitude, because his job was to assess intelligence to aid the planning of
operations which would occur; nor did he think these views were right. He selected the
most optimistic of best and worst cases. His predictions were close enough to work.
On 6 June, the Normandy garrison was stronger than any analyst predicted
before 25 May —well past the level COSSAC thought acceptable, measured in divisions. 56 JIC (44) 66 (0), 1.3.44, RG 331/1/114; JIC ( 44) 127 ( 0 ) Final, 3.4.44. XXX; JIC (44) 127 (0)
( FINAL), 3.3.44, RG 331/3/126; JIC (44) 188 (0) Final, 13.4.44, JIC (44) 198 (0) (Final), 13.5.44, JIC (44) 210 (0) (Final), 20.5.44, JIC (44) 215 (0), 25.5.44, RG 331/3/128.
57 Hinsley, British Intelligence, Three, II, p. 71b
142
That indicator, however, was misleading, because German divisions had fewer units
than in 1943, diluted with “Ost” troops. Measured by a constant indicator, purely
German units, on 6 June that garrison included 30 infantry battalions ( three divisions
and one regiment on a 1943 scale), one parachute and one panzer grenadier regiment,
with the rest of 21 Panzer Division nearby. In numerical terms, this garrison was one
division above the COSSAC conditions for success on D-Day, but perhaps slightly below
them in qualitative terms. In any case, this rise in German strength did not produce
failure, but it did prevent success. German reinforcements were smaller and slower than
any allied analyst dared predict, though close to the level Williams hoped, and within
the COSSAC conditions. By D + 12 ( 18 June), the first wave of five Panzer/Panzer
Grenadier divisions, three infantry divisions ( including 3 Paratroop) and elements from
two more reached Normandy. Compared to SHAEF’s predictions of 3 June, this was
66% the expected level of reinforcements in armoured formations and 40% in infantry,
producing a total of 60% the mechanised divisions ( six) and 65% the infantry ( nine
divisions and three battlegroups ) at the front. Even more, two of these Panzer Divisions
were not yet in action. By D + 25 ( 1 July), the second wave, of another infantry division,
elements of four others and two Panzer divisions, reached the front. Compared to
SHAEF’s predictions of 3 June, this was 100% the expected gains in armoured
formations but only 20% of the infantry, producing at the front 80% the Panzer/Panzer
Grenadier Divisions ( eight) and 47% the infantry—much less, given their losses ( 10
divisions, and elements from another seven). (In both cases, battlegroups are treated as
33% of a division, an overstatement which perhaps balances 3 Paratroops’ superiority
over any other German division). On these issues, Williams was much more accurate
than any other analyst, and good by any standard. All allied analysts, however, were
wrong on two major matters, which together involved almost 33% of the German tanks
deployed to Normandy. On 6 June, 19% of German tanks in France ( including 40% of
its Panthers and Tigers) were in non-divisional units. Planners never forecast their
arrival, which transformed the capabilities of any formation they joined, and so the
fighting. Roughly half these tanks reached the front by D + 12, the rest by D + 25.
Meanwhile, no analyst predicted the main German transfer to France, the return from
Poland of 9 and 10 SS Panzer Divisions, with 15% the total tank strength in France on 6
June. Thus, the Germans transferred as many elite forces to France from outside it as
the allies ever feared, but only one other infantry division. These transfers were smaller
Ferris Intelligence Before Overlord
143
than allied analysts had predicted before 25 May, but larger than they had hoped just
before D-Day.
Compared to what did happen, Williams’ assessments of what would happen
were good, but not perfect. He slightly underrated enemy capabilities, which all other
analysts significantly exaggerated. Allied planning followed his estimates, gaining from
their strengths and losing from their flaws, which stemmed less from intelligence per se
than its relationship to planning. The allies could use strategic intelligence until the
moment landing craft were launched, but just in one sense, to postpone OVERLORD.
Tactical intelligence could be incorporated up to 48 hours before, but only in matters like
fire plans on the beaches. From mid May, however, operations and operational
intelligence were prisoners of prior planning. The naval commander forbade any change
to plans for D-Day involving seaborne landings or supply. Except on the Cotentin
peninsula, where in late May airborne attacks were changed to reflect enemy
redeployments, the planning applied on D-Day was made between February and April.
It also rested on William’s intelligence estimates from that period, the most optimistic
reading of evidence from the most optimistic time, which no one, including himself,
entirely accepted on 5 June. These plans assumed the enemy would have fewer
formations at Normandy between D + 8 and D + 20 than SHAEF expected and more
than 21 Army Group did; and that the Normandy garrison would be weaker than every
analyst thought it was.
Williams rightly predicted the value of German infantry and armoured divisions,
and their strength in formations around the beachhead by D + 8 and D + 20. When
these plans were formulated between February and April, he was more right on those
issues than anyone else; and so too when they began to be executed. Yet his estimates on
the day of execution were different from those during the days of planning. Moreover,
because he accurately assessed the numbers of combat worthy formations in France, his
estimates were upset more by unexpected matters, like the rise in the Normandy
garrison just before OVERLORD, or the arrival of 9 and 10 SS Panzer Divisions, or the
impact of non-divisional armour, than were those of other analysts, with fat free from
their worst cases to cover underestimates elsewhere. The allies paid for Williams’
accuracy in assessment by underestimating the enemy in important particulars, though
the gains were worth the costs. Again, by 5 June Williams had a remarkably good
144
picture of the enemy, which, alas, could not help units on 6 June, though it guided
actions in coming days. So too, like other planners at 21 Army Group, he was more
accurate about numbers of enemy formations than the situation they would produce
once battle was joined. His estimates and their planning attempted precisely to gauge
the nature of many variables and their interactions and combine them into one whole.
He and they were surprisingly accurate, but imperfect, and the battle took a different
form than they had expected. This situation should not be surprising.
OVERLORD and the Limits to Intelligence
Normandy was a hard operation. We tend to forget that fact because it was done
so well. The Allies, forced to launch a seaborne assault against heavily defended
positions, could not leave the beachhead until their strength had built up, by which time
the Wehrmacht had surrounded them in a powerful ring, at a time when the balance of
military power on land had swung away from the attacker, defensive systems became
elaborate and killing zones deeper and deadlier, especially along the narrow fronts of the
west. OVERLORD, the most complex operation of the Second World War, faced
interlocking problems that required precise and accurate solutions. Either inaccuracy or
imprecision could hamper operations or wreck them; yet the more precise planning was,
the less accurate it could be, because chance must strike like a tsunami. Planners for
OVERLORD controlled just one part of a dynamic and reciprocal system. They
determined what to do on the basis of guesses about what the enemy would do, and the
interaction between their wills and actions and chance.
21 Army Group understood these facts. It played to the odds and understood
them. What went right outweighs what went wrong, and avoided the worst of outcomes;
but still some mixed ones occurred, because of the way 21 Army Group and the allies
made decisions. Some aspects of OVERLORD were planned in a tight and centralised
way—in particular, anything to do with seaborne issues, like beach assault and
logistics—others not. OVERLORD, after all, was defined by a staff without power or
responsibility. Its land component was refined by another staff parachuted in five
months before the attack; and it was fought by armies of several nationalities, each with
its own command, doctrine, training and preparations on stream. This produced a
strikingly loose approach to intelligence, planning and training. Estimates of German
Ferris Intelligence Before Overlord
145
armoured strength were imprecise, contradictory, and left that way—every authority
was free to believe what it wished; so too, views of enemy strength between D Day and D
+ 25. 21 Army Group followed Williams, while American forces were more pessimistic.
In May 1944, FUSAG’s version of NEPTUNE used for intelligence an estimate of 10
February 1944, so old that many of the formations it listed no longer were in the west,
while others known to be there were not mentioned. The Third US Army gave its forces
an assessment based on SHAEF’s views of 17 April. 58 That every command had its own
estimates created some confusion on 6 June. Above all, it blinded generals and units to
tactical realities. In preparing for OVERLORD, the allies faced three circumstances,
each requiring distinct training: beach assault, attrition, and mobile war. Everyone
expected and trained for the best case, quick breakthrough leading to mobile war, and
ignored the prospect of having to fight through the bocage, rectangular fields enclosed by
hedgerows, which seemed pessimistic and old fashioned. This issue fell through that gap
in intelligence, planning and training. No allied commander matched Montgomery in
the ability to fight this kind of battle, but even his plans did not play to this strength.
Where British and Canadian units had a tribal understanding of how to fight such a
battle, American ones did not, and their commanders ignored these weaknesses. This
shaped a key phenomenon of the first six weeks of the battle of Normandy: the failure of
the American army.
Meanwhile, the allied ability to gain what usually is the easiest material to
find, tactical intelligence, was constrained. In the 50 days before 6 June, eight dedicated
squadrons conducted 735 photographic reconnaissance (PR) missions in north west
France, including seven overflights of the main beachhead up to 4,000 yards inland, and
detailed coverage of beach gradients and enemy forces. Yet the allies could not focus
these resources on the real target for fear of tipping their hand—as it was, the Luftwaffe
drew dangerously accurate predictions of allied intentions from analysis of allied
reconnaissance flights. The allies aimed to fly two imagery missions elsewhere for every
one flown over Normandy. When free to concentrate on the real target, this work was far
more effective—from 6 June 1944, 380 PR sorties were flown over Normandy each day.
58 First United States Army, Report of Operations, 20 October 1943-1 August 1944, SR, Reel One,
Book One, 20.5.44, HQ First US Army, “Operation Neptune”; After Action Report Third U.S. Army, I August 1944-1 May 1945, Reel One, Third United States Army Outline Plan, Operation Overlord, nd.
146
59 Had this effort been possible in the weeks before OVERLORD, perhaps it would have
overcome the problems in tactical intelligence. So too, in February 1944, the allies
ceased trying to take prisoners or conduct seaborne reconnaissance in Normandy,
though they understood this would cripple the collection of tactical information. The only
such mission run thereafter, aimed to examine the nature of outer beach defences, was
conducted at the Pas de Calais, for security reasons and with effect—though the
Germans detected the mission, this information reinforced OKW’s predisposition to fear
for that region. 60 The allies hoped to make up for the damage done to tactical
intelligence by interrogating prisoners taken in Italy who previously had been in
Normandy, with mixed success. 61 Allied forces in Italy did not capture prisoners from
any division presently in Normandy, though some based there in 1943 did illuminate old
procedures, while other captives gave excellent and recent accounts of defences and their
defenders in Brittany, Belgium and the Netherlands. 62 CSDIC provided useful
background to assessments of defences in Normandy, but nothing concrete; imagery
provided many facts, which guided planners and commanders, and the maps issued to
assault units. Still, much was missed.
These circumstances and constraints prevented good intelligence services from
discovering basic points. Errors that normally would be inexcusable, were unavoidable.
Few things should be easier in intelligence than to determine the strength and defences
in a tactical sector one plans to attack; but this proved beyond the grasp of services
which otherwise achieved marvels. Though prior planning and the initial assault rested
on a good picture of the topography and defences on the beaches, the allies missed
important details: and at H hour, a prior failure to locate individual guns could have
deadly costs. In particular, the allies were unsure of the deployment of two of the
enemy’s seven divisions in Normandy, its best formations there, standing precisely at
places the allies must attack on D-Day. 352 Infantry Division, a formation with full
59 “Operation ‘Neptune”, Joint Commanders-in-Chief Memorandum No. 7, Air Reconnaissance-
Requirements and Resources”, NJC/00/74/6, 15.4.44, WO 205/15. 60 Hinsley, et.al., British Intelligence, Three, II, p. 89; Helmut Heiber and David Glantz eds),
Hitler and his Generals, Military Conferences 1942-1945,The First Complete Stenographic Record of the Military Situation Conferences, from Stalingrad to Berlin, ( Now York, 2003), pp. 434.
61 3rd Joint Commanders in Chief meeting, 22.12.43, WO 205/12; SHAEF/45BX/1/INT, 9.2.44, SHAEF/17/225/Ops, 12.2.44, RG 331/12/9
62 RG 165/179/660, C.S.D.I.C. ( U.K.) S.I.R No 130, 132, 136; RG 165/179/659, C.S.D.I.C. ( U.K.), No 160, 213, 260.
Ferris Intelligence Before Overlord
147
strength and some combat veterans, raised by 150% the strength of the garrison on Gold
and Omaha beaches and made them formidable, while 21 Panzer placed its anti-tank
units and half its infantry between the Orne River and Caen, key British and Canadian
targets on D-Day, with its armour in easy reach behind. This position was well suited to
its task and its enemy-- the Germans deployed their forces as well in Normandy, as they
did badly in France. Allied intelligence looked hard and objectively for information on
these forces, and Williams suspected, but could not prove, they might be near their true
locations. His last report before D-Day, of 4 June, warned that 352 Division might be
where it really was, while noting 21 Panzer was unlocated, but thought to be entirely
outside the assault area, from which it likely would advance toward Caen as a whole
around H + 12. He gave worst case, but sound, estimates of the amount of German
armour his forces would meet on D-Day, 240 tanks and 40 assault guns ( only 10 % off
the real strength of the two armoured formations in the sector, 12 SS and 21 Panzer).
Second Army intelligence, conversely, warned its units to expect immediate local
counterattacks by some armour, while between H + 12--H + 24, they might encounter
540 tanks ( more than twice the real strength of those formations), and perhaps 160
more from 17 SS Panzer Grenadier Division ( which in fact had only 42 assault guns). 63
Second Army correctly warned some armour might be met before Caen, and much more
there, but it grossly distorted that strength--one of the greatest examples of worst case
logic before OVERLORD. Here, as in similar instances with the JIC or SHAEF, the aim
probably was to state the absolute worst case, so to make people think about the
problem in advance, to show the limits to bad news, and for the record, rather than to
state one's true belief. After all, if Second Army was right, all that could be done was
cancel OVERLORD--the allies could not possibly defeat a force so much larger than that
defined in the COSSAC conditions.
Williams’ last estimate did not reach divisional commanders before the attack,
but that of Second Army was known to all unit commanders. They ignored it, for reasons
which remain unclear. Units essentially seem to have treated that issue as academic--
above their pay scale. The mere process of hitting the beaches and forming up to
advance were difficult and time consuming. Units seem to have preferred to focus on
known problems, and largely ignored anything which would happen later, because such
63 Hinsley, et.al., British Intelligence, Three, II, pp 847-51.
148
matters obviously were unpredictable, and their superiors were uncertain about them.
Thus, British units assumed they would drive straight off the beaches and first meet
enemy forces around Caen, in an encounter battle, possibly with Germans attacking
British positions, rather than charging themselves into enemy units deployed in
defensive systems. They expected to meet substantial enemy armour after Caen, but
little before it. Nothing but a clear warning the enemy already controlled the approaches
to Caen could have affected their attitudes; which Williams would have done had he
been able to; but he could not. Even such a warning merely could have saved some lives,
tanks and, above all, confidence in high command; the real problem was that the
Germans held Caen, not the fact this surprised the allies. These failures in the
intelligence cost lives, but not the battle. Even had they known the truth, the allies
would have had to try to take Caen on 6-7 June, and largely in the same way—
otherwise they would have been condemned for passivity. In any case, they crushed the
only German counterattacks launched during the landing, and ensured ultimate victory.
So too, certain knowledge of the location of 352 Division could have done nothing to stem
the slaughter at Omaha beach. These intelligence failures were indicative, but not
influential
More significant was a muted version of an old problem. In the CRUSADER
operations during 1941, British intelligence was precise in classic ways—99% accurate
on the enemy’s order of battle and numbers of tanks-- and let its forces time their attack
with optimum effect. Yet they were grossly wrong on basic matters, like the quality of
enemy forces and weapons relative to their own. 64 Meanwhile, an extremely precise
plan, relying on intelligence collapsed, as did the Eighth Army. Poor training negated
good intelligence. British intelligence was better before CRUSADER than OVERLORD,
and integrated as much into planning; but by 1944, allied armies, plans and
commanders were better. Even so, before OVERLORD the allies misinterpreted the
relationship between terrain and tactics. Planners and units assumed that once ashore,
they quickly would enter the rolling country inland from Normandy, and fight mobile
battles. Even an able “tactical study of the terrain correlated with the latest Intelligence
estimates of enemy capabilities” of 7 May, which discussed fighting in circumstances
64 John Ferris, "The 'Usual Source': Signals Intelligence and Planning for the Crusader Offensive,
1941", In David Alverez (ed), Allied and Axis, Signals Intelligence during the Second World War, (London, Frank Cass, 1999), pp. 84-118.
Ferris Intelligence Before Overlord
149
like and roughly on the line of 28 June 1944, was over optimistic. It defined that
territory as “not an easy one for forces to advance through rapidly in the face of
determined resistance, but it will likewise be most difficult for the enemy to prevent a
slow and steady advance by infiltration”. It was “difficult to judge” whether the bocage
would better suit attacker or defender; tactics for the matter “should be given
considerable study by formations to be employed therein”. In perhaps the greatest
failure in allied planning before the invasion, this advice was not followed. 65 More
broadly, allied planners did not realise they would be fighting a battle of attrition in
terrain well suited to the defender, and did not train for it. They knew Panthers, Tigers
and 88 mm anti-tank guns were formidable, not that they would be attacking the largest
force to space densities in these weapons the Germans ever managed on the defence, in
terrain well suited to their strengths and allied weaknesses. Through these gaps in
tactical intelligence flowed tragedies.
Again, intelligence enabled effective planning for the beach assault, and the first
days of the fight. It could not do so for subsequent operations, because too much was
unpredictable. 21 Army Group ordered Canadian forces to plan for a breakout toward
Le Havre and the channel ports on D +20, on the assumption that the bridgehead had
reached the point it roughly did around 28 July—8 August, D + 62 to D + 74, during the
Cobra offensive. Canadian forces, on a line from Argentan to the coast at Houlcourt,
allegedly would confront a few infantry divisions, undefined in number and implicitly
low in quality: “the enemy’s principal anxiety will be to defeat or contain the forces
further SOUTH-WEST”, driving on the Loire river and Brittany. These assumptions
anticipated Hitler’s orders around 28-30 July, but not the disposition of German forces.
The commander of First Canadian Army, H.D.G. Crerar, retorted that these
assessments rested on a “very sketchy” basis. “While I am without authoritative
information on this matter, my personal judgment leads me to somewhat contradictory
conclusions”, which represented the disposition of German forces around 28 July, but
not their actions, or Hitler’s orders: the enemy would “treat the threat to the LOIRE and
BRITTANY ports as of secondary importance, and consequently… counter attack with
maximum forces, and as soon as possible, the left sector, or Eastern flank, of the Allied
bridgehead”. Crerar’s staff should assess “this alternative problem ( the defence by First
65 21 A Gp/20651/55/G ( Plans), 7.5.44, “Appreciation of Possible Development of Operations to
Secure a Lodgement Area, Operation Overlord”, WO 205/118
150
Canadian Army of the Eastern flank of the Allied bridgehead against heavy and
determined German counterattack)”. When Williams was asked to assess this debate, he
concluded that much is pretty obviously guesswork at this stage and my guess is that
the Canadians should be prepared to face 2-3 Pz plus 7-8 infantry divs: depending on the
success of 2 British Army. Therefore it seems essential to me that the Canadian Army
chooses ground vital at once to the retention of their sector and as a diving-board for
“AXEHEAD”. I am afraid that General Crerar will find these comments very unhelpful.
That is, First Canadian Army, with four infantry and two armoured divisions,
should plan an assault leading to breakthrough, while absorbing singlehandedly a full
blooded counter-attack by half the German Army at Normandy! That assessment
echoed the Canadian position at the Falaise Gap. Instead, 21 Army Group told Crerar to
plan an assault which also could handle one large counter-attack, but not more, though
“with the information now available, the possibility cannot be excluded”. 66 Not
surprisingly, these plans had little value when Canadian and Polish forces assaulted the
remnants of five panzer and ten infantry divisions on both the defense, and attack,
around Argentan and Falaise between 8--21 August. These commanders and planners
were competent and informed men, fairly accurate in their net assessment of
capabilities and intentions, but intelligence simply could not provide the surety needed
to produce a plan able to survive first contact with the enemy. Their discussions
illustrate that problem.
Intelligence and Deception
Intelligence also shaped a struggle in deception. It showed in real time what the
Germans expected and perceived, enabling adjustments to allied behaviour ( both
positive and negative) so to minimize the chance of inconvenient actions. That the allies
conducted deception before OVERLORD is famous; that Germans did so, is not. Von
Rundstedt, however, later claimed to have developed “ a huge deception programme”,
aimed to mislead the allies about his forces, including “intensive propaganda about the
invincibility of the Atlantic Wall”, the laying of dummy minefields, and systematic
66 21 Army Group 00/316/1/G Plans, 20.3.44, WO 171/126; Memorandum by Crerar for C in C
and 21 Army Group, CPS/1-6-4, 14.4.44, passim, NAC, RG 24/C 17/13607; Appreciation by Williams, “Axehead”, 30.4.44, and de Guingand to Mann, 24.4.44, WO 205/5C.
Ferris Intelligence Before Overlord
151
efforts to fool authorities across France into exaggerating the strength of German forces,
so spurring false rumours to London. He passed false maps of defences “to the Allies by
means of German agents in Paris and Switzerland”, and exaggerated statements of his
forces to the Japanese ambassador at Vichy, so “to lull both him and his government
into a sense of security as to the strength of German forces in the west”. 67 Von
Rundstedt’s claims are plausible. Germany often conducted such practices, and wireless
deception. It attempted to simulate the presence of Luftwaffe forces in France during
1941, as part of the cover for Operation Barbarossa, though British intelligence
penetrated this effort, so finding a clue to German intentions. 68 Von Rundstedt’s chief of
staff, General Gunther Blumentritt, confirmed his claims, while other generals and
soldiers described efforts to exaggerate the presence of panzer and infantry divisions in
France during 1943-44. 69 The effect of these efforts is hard to trace. British intelligence
never used ( and probably never received) the maps allegedly shipped through
Switzerland. Perhaps it did pick up some of the rumours circulating through France,
which might explain certain errors about German order of battle in early 1944, but if so,
the skill and sources of British intelligence overcame that threat, without detecting the
danger, at little cost to planning.
The material allegedly leaked to the Japanese embassy in Vichy raises greater
problems. That information was intended to deceive Germany’s ally, not its enemies: von
Rundstedt could not have known this material might reach his foes, because they read
Japanese diplomatic traffic. Nor does any allied intelligence acquired from Japanese
authorities in Vichy fit von Rundstedt’s descriptions. His statement, however, suggests
that Germans sought to deceive other Japanese authorities who inspected the defences
in France in 1943-44, like General Oshima Hiroshi, the Japanese ambassador, and
Admiral Kojima Hideo, the naval attaché, to Germany. Reports from these men rank
high within the intelligence available for planning in OVERLORD. When critically
assessed, they reflect either a mistaken overestimate of Nazi power or else, and more 67 RG 165/179/649, CSDIC UK, G.R.G.G. 326 (0), C in C West and the Allied Invasion Plan, 1943-
1944; Interrogation of General Gerd von Rundstedt, 4-14, 1.2.46, WO 205/1020. 68 Brieftagebuch Haifisch I u. II, H.Gr.D. 85566, T 311/34, NARA; Air Historical Branch, The
Second World War, 1939-1945, The Royal Air Force, Air Ministry Intelligence Part II: Chapter 6, paragraph 58, undated; Hut Three History, Section III, Development as a Research Organization ( Spring 1940-Autumn 1942), HW 3/104.
69 RG 165/179/659, CSDIC ( UK) , S.I.R. 520. “285 I.D, ( Action “Landfraf”); RG 165/179/649, CSDIC UK, G.R.G.G. 326 (0), op. cit; RG 165/179/663, CSDIC ( UK), S.I.R. 1466, “German Deception Measures”.
152
likely, German efforts at disinformation, tempered by Japanese professionalism in
observation and assessment. Oshima and Kojima tended systematically to overstate
German capabilities in the west. They distorted the power of the Atlantic Wall, while
their account of how the Germans planned to smash an allied attack reflected von
Rundstedt’s disinformation, tempered by accurate information from other sources. The
more they relied on what they were told, as against what they could see, the less
accurate became their comments. In October 1943, Oshima distorted German strength
in three major ways, claiming that it included fifteen “mechanized” divisions, 50% more
than was true, if his terminology meant formations capable of offensive action; with
several more on the way-- including two paratroop divisions from Italy, which never
arrived; while the two thirds of the thirty “coastal” divisions that had just two regiments,
were being “greatly reinforced up to the strength of three regiments”, which did not
happen with half of them, with the rest receiving merely “Ost” troops. In May 1944,
Kojima indicated that the forces in France were “all the pick of the German Army”,
implied that almost all LE divisions had three regiments, and overestimated by 100%
the number of Luftwaffe aircraft able to enter the battle. British intelligence found
treasure in these reports, while throwing out the trash, by comparing them to material
from all other sources, including Ultra, and by recognising that these officers were being
misled. As MI 14 noted,” although German senior officers were clearly out to impress
( Oshima) his report, while exaggerating factual information, is a reasoned and well
considered document”. 70 Inadvertently, however, the Japanese estimate of the high
quality of the German preparations for attacks, and of their response to invasion,
reinforced British preconceptions of these matters, and their worst case assumptions on
these matters.
If von Rundstedt’s claims and this analysis of it are correct, one confronts an
abundance of ironies. A great intelligence source, Magic, exposed the allies to a
deception campaign that could have reached them through no other means, but which
did so in a way uncontrolled by its masters. This deception was undone by the
professionalism of officers intended to receive it, and those who did so only through
intelligence, Japanese and British who, penetrating the bodyguard of lies in different
ways, gave truths to Tokyo, and even more to London. German deception failed
70 British Intelligence III/2, pp. 771-75, 787-92.
Ferris Intelligence Before Overlord
153
because it was countered through excellent intelligence by its allies and enemies, though
even had the effort succeeded in technical terms, probably little would have changed:
defenders never gain easily from deception, especially on so small a scale as this. The
greatest effect of these reports was counter-productive: to reinforce public statements by
German authorities about the power of their defences, and private allied expectations on
the matter, so driving the demand for effective counter measures, including
FORTITUDE.
Equally, British intelligence was essential to the success of FORTITUDE. That
success commonly is overstated, focusing on the eccentricities of intelligence officers and
agents, and also controversial, for general and particular reasons. Deception reinforces
preconceptions, which complicates judgment of its significance, compared to other
elements of its victim’s process of decision making. The influence of FORTITUDE on
OVERLORD also overlaps with that of Special Forces, airpower, and all factors which
limited the redeployment of enemy forces, kinetic or perceptual, internal or external to
German minds. All considered, however, FORTITUDE was essential to victory in
OVERLORD. 71
Preconception drove German assessment, aided by two engines. First, the
Germans failed in the collection of intelligence, the best means to avoid error,
uncertainty and the division of force; even worse, much, perhaps most, of their evidence
was shaped by the enemy. After the war, some staff officers also claimed to have
politicized assessment, by overstating all indicators of allied strength, so to drag
reinforcements to France. If so, inadvertently, these officers served as a document
delivery service for deception. They amplified the effect of the false order of battle
peddled by the deceivers, which led Germans to overrate allied strength in divisions by
almost 50%. Otherwise, the Germans never could have believed that the invasion of
71 Michael Howard, Strategic Deception, British Intelligence in the Second World War, Its
Influence on Strategy and Operations ( Cambridge, 1991), is essential, but must be supplemented by several essays in Michael Handel (ed) Strategic and Operational Deception in The Second World War, ( London, 1989). Other valuable accounts include Katherine Barbier, D-Day Deception: Operation Fortitude and the Normandy Invasion, ( Stackpole, Westport, CT, 2007) , John Ferris, “The Roots of Fortitude: The Evolution of British Deception in The Second World War”, in Thomas Mahnken (ed), The Paradox of Intelligence: Essays in Memory of Michael Handel ( London, 2003), Thaddeus Holt, The Deceivers: Allied Military Deception in The Second World War, ( Scribners, New York, 2004) and Nicholas Rankin, A Genius for Deception, How Cunning Helped the British Win Two World Wars, ( Oxford University Press, New York, 2008) .
154
Normandy was a feint, with another army poised to strike the true target, the Pas de
Calais. Second, the allies understood enemy expectations, and exploited them by
evading German strength, striking its weaknesses, dividing its forces and degrading
their deployment, by many means. Among them, security and deception had unique
power, tools to manipulate the assessments which caused actions, rather than stalling
movements as they were made. Germans commonly saw Normandy as an obvious sector
for attack, especially the ports which bracketed it, Cherbourg on the Cotentin peninsula
and Le Havre north of the Seine River. Before the attack, few things prevented
Germany from sending the three to five more divisions to Normandy which would have
cancelled or contained the invasion. On 6 June, even just another good infantry division
at Normandy might have crippled the allies at two places, instead simply of Omaha
Beach. Once the Germans insured the Pas de Calais heavily against risk, Normandy
reached the top of priorities for resources. In the week before D Day, the enemy almost
doubled the capability of that garrison, increasing its German infantry battalions by
over 50% and adding an armoured division; but still it stood below the average force to
space ratio on the Pas de Calais. British deception and security rank high among the
things which blocked that development: they were perhaps the only thing left. That the
Germans came so close to seeing the real danger and to blocking it, proves not the
failure of FORTITUDE, but its success. Almost half of the German forces in the west,
good and mediocre, were badly placed for the attack on Normandy, while the formations
there were just too weak to crush or contain the invasion. All that Germany needed to
defeat an allied attack on Normandy was superiority in intelligence. They were inferior.
FORTITUDE was one of several factors which were essential for the allies to
establish a viable beachhead by D + 2. It was less important after the invasion, but still
notable, on the border between necessary and secondary causes, by short circuiting
German defences. FORTITUDE added little to the collapse in German command above
the divisional level which bought almost 24 precious hours on 6 June. The delay in
decisions enabled the allies to penetrate the beach defences and shatter their defenders,
before the immediate reserve entered the fray, to slow the deployment of the first wave
of reinforcements to the theatre, and to crush counterattacks on 6 and 7 June. These
successes stemmed from the convoluted and fragile German command system in the
west, fuelled by the politics of commanders facing uncertainty and ignorance about
allied capabilities and intentions, which were shaped by deception and security; but also
Ferris Intelligence Before Overlord
155
because a false sense that no attack was imminent, stemming from incompetence in
their intelligence, and Eisenhower’s willingness to attack in poor weather, disrupted
decision making, as Rommel flew home for his wife’s birthday, for example. In an alloy
with German preconceptions, however, FORTITUDE slowed the decisions to deploy the
second wave of reinforcements ( including two of the seven Panzer divisions in France)
by two weeks, and of a third wave, of a dozen LE divisions, for almost two months, until
just before the foe collapsed. It also helped to keep six weak divisions ever from entering
the battle. Special Forces and airpower slowed these actions once they occurred, and
others. Without FORTITUDE, none of these achievements are likely to have happened,
let alone all of them, nor to the same degree. Without them, OVERLORD would have
been much less successful, though probably after D + 2, with the establishment of a long
and deep bridgehead, and certainly by D +10, when the allies seized the Cotentin
peninsula and surged on Cherbourg, ensuring control of a major port, Germany could
have contained the allies only had every division in the west immediately moved to
Normandy. Even then, allied victory perhaps was inevitable.
On the Day.
In OVERLORD, intelligence was excellent, and influential. This case shows the
best that can be expected from intelligence, and its limits, and the degree to which
decision makers always must be able to act in the face of uncertainty, error and
ignorance. The conventional assumption is that intelligence for OVERLORD was precise
and powerful, and almost everything right and certain, with some errors. In fact, much
was erroneous or uncertain, but most key things were correct and trusted by the right
people, who were excellent consumers. Montgomery had a weird ability to read an
enemy, and a battle, before the fact; Williams was an analyst of acumen: together, they
were a great intelligence team. Without them, planning for OVERLORD might have
failed. Again, allied intelligence was filled with errors, but since they were not
systematic, many of them cancelled each other out. Others shaped matters which did
not matter —so too often the successes, including the most famous or technically
complex of them. Success was most pronounced at the strategic level, where intelligence
bracketed the truth closely enough to show where it was, or was not. Since planners
and commanders cared about gaps in their information, intelligence was fundamental as
156
a source of knowledge, and psychological certainty. Even at this level key issues were
uncertain, like enemy armoured strength, and the speed and scale with which it could
reinforce Normandy. Allied assessments overstated the efficiency of German divisions in
France, and also of their command system, though none could safely have predicted that
the latter would perform so poorly at Normandy as it did. In this instance, worst case
logic spurred effective action. It led SHAEF and 21 Army Group to focus on jamming the
German machine through every means possible, including airstrikes, Special Forces,
mobilization of the French resistance, and deception, in particular, the demand that
FORTITUDE shape German actions long after OVERLORD opened. These means
achieved their end; they might have worked less well had they not been driven by such
fear and uncertainty. Ultimately, these errors had little cost, because they spurred
effective counteraction, their significance was small, and 21 Army Group made the
fewest mistakes and acted on its own views. The greatest area of failure lay at the
operational level; indeed, as on 6 June 1944 tactical and operational problems flowed
together, errors in minor forms of intelligence had major consequences. However, one
may doubt that these failures mattered much, or that successes with them would have
done so either. On 6 June 1944, the problem was the enemy, not intelligence about it.
Intelligence had done well enough, as much as could be expected. The rest was up to the
men.
(カナダ・カルガリー大学歴史学部教授)