Selling Platforms - Boston University...Selling Platforms Hemant K. Bhargava UC Davis with Olivier...

22
Selling Platforms Hemant K. Bhargava UC Davis with Olivier Rubel Platforms Symposium, Boston, July 2016 Hemant K. Bhargava UC Davis 07/14/2016 Selling Platforms 1 / 13

Transcript of Selling Platforms - Boston University...Selling Platforms Hemant K. Bhargava UC Davis with Olivier...

Page 1: Selling Platforms - Boston University...Selling Platforms Hemant K. Bhargava UC Davis with Olivier Rubel Platforms Symposium, Boston, July 2016 Hemant K. Bhargava UC Davis 07/14/2016

Selling Platforms

Hemant K. Bhargava

UC Davis

with Olivier Rubel

Platforms Symposium, Boston, July 2016

Hemant K. Bhargava UC Davis 07/14/2016 Selling Platforms 1 / 13

Page 2: Selling Platforms - Boston University...Selling Platforms Hemant K. Bhargava UC Davis with Olivier Rubel Platforms Symposium, Boston, July 2016 Hemant K. Bhargava UC Davis 07/14/2016

“SELLING” PLATFORMS

I exciting part of economy and society

I some grow virally, due to network effects, many must be “sold”

I single-sided network goods: e.g., Kyruus

I two-sided goods: e.g., OpenTable, American Well, CreditKarma

I “selling” is fraught with uncertainty, moral hazard ... managed via

risk-sharing compensation plans (commission rate)

I NE alter rewards, productivity and risk exposure of selling agent

I what is the net influence on plan design?

I how should network and platform firms manage sales agents?

Hemant K. Bhargava UC Davis 07/14/2016 Selling Platforms 2 / 13

Page 3: Selling Platforms - Boston University...Selling Platforms Hemant K. Bhargava UC Davis with Olivier Rubel Platforms Symposium, Boston, July 2016 Hemant K. Bhargava UC Davis 07/14/2016

“SELLING” PLATFORMS

I exciting part of economy and society

I some grow virally, due to network effects, many must be “sold”

I single-sided network goods: e.g., Kyruus

I two-sided goods: e.g., OpenTable, American Well, CreditKarma

I “selling” is fraught with uncertainty, moral hazard ... managed via

risk-sharing compensation plans (commission rate)

I NE alter rewards, productivity and risk exposure of selling agent

I what is the net influence on plan design?

I how should network and platform firms manage sales agents?

Hemant K. Bhargava UC Davis 07/14/2016 Selling Platforms 2 / 13

Page 4: Selling Platforms - Boston University...Selling Platforms Hemant K. Bhargava UC Davis with Olivier Rubel Platforms Symposium, Boston, July 2016 Hemant K. Bhargava UC Davis 07/14/2016

BACKGROUND AND RESEARCH QUESTIONS

I platforms need to be “sold” (too)

I salesforce management literature: principal-agent model - does not

recognize role of network effects

I our research: impact of NE on

I mix of guaranteed and performance-based incentives?

I risk and reward sharing between firm and agent?

I how should firm respond to externalities created by NE

Hemant K. Bhargava UC Davis 07/14/2016 Selling Platforms 3 / 13

Page 5: Selling Platforms - Boston University...Selling Platforms Hemant K. Bhargava UC Davis with Olivier Rubel Platforms Symposium, Boston, July 2016 Hemant K. Bhargava UC Davis 07/14/2016

BACKGROUND AND RESEARCH QUESTIONS

I platforms need to be “sold” (too)

I salesforce management literature: principal-agent model - does not

recognize role of network effects

I our research: impact of NE on

I mix of guaranteed and performance-based incentives?

I risk and reward sharing between firm and agent?

I how should firm respond to externalities created by NE

Hemant K. Bhargava UC Davis 07/14/2016 Selling Platforms 3 / 13

Page 6: Selling Platforms - Boston University...Selling Platforms Hemant K. Bhargava UC Davis with Olivier Rubel Platforms Symposium, Boston, July 2016 Hemant K. Bhargava UC Davis 07/14/2016

RESULTS AND INSIGHTS

I network effects exert externalities on sales agent: increase both mean

and variance of sales (⇒ compensation risk)

I spectrum of influence, depending on nature of network effects

I one-sided (direct) vs. two-sided (indirect) NE

I which side to meter for commission

I one vs. two agents

I firm’s ability to leverage network effects depends on balance between

# externalities vs. # instruments to manage them.

Hemant K. Bhargava UC Davis 07/14/2016 Selling Platforms 4 / 13

Page 7: Selling Platforms - Boston University...Selling Platforms Hemant K. Bhargava UC Davis with Olivier Rubel Platforms Symposium, Boston, July 2016 Hemant K. Bhargava UC Davis 07/14/2016

CONCEPTUAL FRAMEWORK AND BENCHMARK CASE

compensation design without network effects

I agent’s influence on sales: Q = V + βw + ε

V=base sales; β=agent’s productivity; ε ≈ N(0, σ2)

I risk-averse agent, earns ω(w) = α0+α1Q, picks effort level w∗

(max. U(ω(Q), w) = −e−ρ(ω(Q)−C(w)) ≥ R

)⇒ w∗ = βα1

I firm designs (α0, α1) to max. E[Π] = E[Q]− (α0+α1E[Q])

⇒ α∗1 =

β2

β2+ρσ2; Λ0 =

α1E[Q]

α0+α1E[Q]=

2β2

β2+ρσ2β4+V (β2+ρσ2)

β4+2R(β2+ρσ2)

Hemant K. Bhargava UC Davis 07/14/2016 Selling Platforms 5 / 13

Page 8: Selling Platforms - Boston University...Selling Platforms Hemant K. Bhargava UC Davis with Olivier Rubel Platforms Symposium, Boston, July 2016 Hemant K. Bhargava UC Davis 07/14/2016

CONCEPTUAL FRAMEWORK AND BENCHMARK CASE

compensation design without network effects

I agent’s influence on sales: Q = V + βw + ε

V=base sales; β=agent’s productivity; ε ≈ N(0, σ2)

I risk-averse agent, earns ω(w) = α0+α1Q, picks effort level w∗

(max. U(ω(Q), w) = −e−ρ(ω(Q)−C(w)) ≥ R

)⇒ w∗ = βα1

I firm designs (α0, α1) to max. E[Π] = E[Q]− (α0+α1E[Q])

⇒ α∗1 =

β2

β2+ρσ2; Λ0 =

α1E[Q]

α0+α1E[Q]=

2β2

β2+ρσ2β4+V (β2+ρσ2)

β4+2R(β2+ρσ2)

Hemant K. Bhargava UC Davis 07/14/2016 Selling Platforms 5 / 13

Page 9: Selling Platforms - Boston University...Selling Platforms Hemant K. Bhargava UC Davis with Olivier Rubel Platforms Symposium, Boston, July 2016 Hemant K. Bhargava UC Davis 07/14/2016

CONCEPTUAL FRAMEWORK AND BENCHMARK CASE

compensation design without network effects

I agent’s influence on sales: Q = V + βw + ε

V=base sales; β=agent’s productivity; ε ≈ N(0, σ2)

I risk-averse agent, earns ω(w) = α0+α1Q, picks effort level w∗

(max. U(ω(Q), w) = −e−ρ(ω(Q)−C(w)) ≥ R

)⇒ w∗ = βα1

I firm designs (α0, α1) to max. E[Π] = E[Q]− (α0+α1E[Q])

⇒ α∗1 =

β2

β2+ρσ2; Λ0 =

α1E[Q]

α0+α1E[Q]=

2β2

β2+ρσ2β4+V (β2+ρσ2)

β4+2R(β2+ρσ2)

Hemant K. Bhargava UC Davis 07/14/2016 Selling Platforms 5 / 13

Page 10: Selling Platforms - Boston University...Selling Platforms Hemant K. Bhargava UC Davis with Olivier Rubel Platforms Symposium, Boston, July 2016 Hemant K. Bhargava UC Davis 07/14/2016

SELLING ONE-SIDED NETWORK GOODS

direct network effects, intensity η

I with Q = V + βw + ηQe + ε, and rational expectations,

q =V + βw

1− η+

ε

1− η; η increases mean AND volatility

I η makes agent more productive, puts in more work, w∗ = β α11−η

and has more compensation risk, Var(ω(q)) = α21

σ2

(1−η)2

how to adjust commission rate and reward structure?

Hemant K. Bhargava UC Davis 07/14/2016 Selling Platforms 6 / 13

Page 11: Selling Platforms - Boston University...Selling Platforms Hemant K. Bhargava UC Davis with Olivier Rubel Platforms Symposium, Boston, July 2016 Hemant K. Bhargava UC Davis 07/14/2016

SELLING NETWORK GOODS: RESULTS

I η has no effect on commission rate, α∗1 = β2

β2+ρσ2

∵ costs (risk-disutility) and gains (compensation) both ≈ α21

(1−η)2

I firm takes more risk; more of agent’s compensation as fixed salary

0.3 0.4 0.5 0.6 0.7

0.0

0.4

0.8

η

total compensation

revenues

performance based incentives

total cash compensation

I yet gives agent a greater share of earnings (... net profit increases)

Hemant K. Bhargava UC Davis 07/14/2016 Selling Platforms 7 / 13

Page 12: Selling Platforms - Boston University...Selling Platforms Hemant K. Bhargava UC Davis with Olivier Rubel Platforms Symposium, Boston, July 2016 Hemant K. Bhargava UC Davis 07/14/2016

SELLING NETWORK GOODS: RESULTS

I η has no effect on commission rate, α∗1 = β2

β2+ρσ2

∵ costs (risk-disutility) and gains (compensation) both ≈ α21

(1−η)2

I firm takes more risk; more of agent’s compensation as fixed salary

0.3 0.4 0.5 0.6 0.7

0.0

0.4

0.8

η

total compensation

revenues

performance based incentives

total cash compensation

I yet gives agent a greater share of earnings (... net profit increases)

Hemant K. Bhargava UC Davis 07/14/2016 Selling Platforms 7 / 13

Page 13: Selling Platforms - Boston University...Selling Platforms Hemant K. Bhargava UC Davis with Olivier Rubel Platforms Symposium, Boston, July 2016 Hemant K. Bhargava UC Davis 07/14/2016

SELLING TWO-SIDED NETWORK GOODS (B,S)

cross-market network effects, intensity ηb, ηsI agent hired to recruit side S participants, paid based on S sales

Qb = Vb + ηbQs + εb

Qs = Vs + ηsQb + εs .

I similar to network goods, agent works more, w∗ = β α11−ηbηs

and has more compensation risk, = f (ηb, ηs)

how to adjust commission rate and reward structure?

Hemant K. Bhargava UC Davis 07/14/2016 Selling Platforms 8 / 13

Page 14: Selling Platforms - Boston University...Selling Platforms Hemant K. Bhargava UC Davis with Olivier Rubel Platforms Symposium, Boston, July 2016 Hemant K. Bhargava UC Davis 07/14/2016

SELLING TWO-SIDED NETWORK GOODS: RESULTS

α∗1 =

β2

β2+ρ(σ2s +σ2bη

2s

) ; Λ∗2 = 2

(Vs+Vbηs)(1−ηbηs)

β2+

2β2

β2+ρ(σ2s +σ2bη2s )

I ηb behaves like η ! (no impact on α∗1) but α∗

1 varies with ηs

to internalize externality (agent not rewarded for Qb which ηs affects)

I high ηb is good for firm (like η), but high ηs may not be!

∵ ηs affects Qb, not accounted for in agent’s compensation

too many externalities, too few ways to manage the effects

Hemant K. Bhargava UC Davis 07/14/2016 Selling Platforms 9 / 13

Page 15: Selling Platforms - Boston University...Selling Platforms Hemant K. Bhargava UC Davis with Olivier Rubel Platforms Symposium, Boston, July 2016 Hemant K. Bhargava UC Davis 07/14/2016

SELLING TWO-SIDED NETWORK GOODS: RESULTS

α∗1 =

β2

β2+ρ(σ2s +σ2bη

2s

) ; Λ∗2 = 2

(Vs+Vbηs)(1−ηbηs)

β2+

2β2

β2+ρ(σ2s +σ2bη2s )

I ηb behaves like η ! (no impact on α∗1) but α∗

1 varies with ηs

to internalize externality (agent not rewarded for Qb which ηs affects)

I high ηb is good for firm (like η), but high ηs may not be!

∵ ηs affects Qb, not accounted for in agent’s compensation

too many externalities, too few ways to manage the effects

Hemant K. Bhargava UC Davis 07/14/2016 Selling Platforms 9 / 13

Page 16: Selling Platforms - Boston University...Selling Platforms Hemant K. Bhargava UC Davis with Olivier Rubel Platforms Symposium, Boston, July 2016 Hemant K. Bhargava UC Davis 07/14/2016

SELLING TWO-SIDED NETWORK GOODS: RESULTS

α∗1 =

β2

β2+ρ(σ2s +σ2bη

2s

) ; Λ∗2 = 2

(Vs+Vbηs)(1−ηbηs)

β2+

2β2

β2+ρ(σ2s +σ2bη2s )

I ηb behaves like η ! (no impact on α∗1) but α∗

1 varies with ηs

to internalize externality (agent not rewarded for Qb which ηs affects)

I high ηb is good for firm (like η), but high ηs may not be!

∵ ηs affects Qb, not accounted for in agent’s compensation

too many externalities, too few ways to manage the effects

Hemant K. Bhargava UC Davis 07/14/2016 Selling Platforms 9 / 13

Page 17: Selling Platforms - Boston University...Selling Platforms Hemant K. Bhargava UC Davis with Olivier Rubel Platforms Symposium, Boston, July 2016 Hemant K. Bhargava UC Davis 07/14/2016

TWO-SIDED INCENTIVES FOR TWO-SIDED GOODS?

I hire agent to recruit side S , but pay him also for B sales!

ω(qs , qb) = α0 + α1qs + α2qb; w∗ = βα1 + α2ηb1− ηbηs

I higher ηb, ηs ⇒ higher commission rate; α∗1 = 1

(1−ηsηb)(1+ρσ2s )

I “pay to play” ... α∗2 = −α∗

1ηs

I firm is better off with stronger network effects (both ηb and ηs)

second metric ⇒ better tuning for multiple externalities

Hemant K. Bhargava UC Davis 07/14/2016 Selling Platforms 10 / 13

Page 18: Selling Platforms - Boston University...Selling Platforms Hemant K. Bhargava UC Davis with Olivier Rubel Platforms Symposium, Boston, July 2016 Hemant K. Bhargava UC Davis 07/14/2016

TWO-SIDED INCENTIVES FOR TWO-SIDED GOODS?

I hire agent to recruit side S , but pay him also for B sales!

ω(qs , qb) = α0 + α1qs + α2qb; w∗ = βα1 + α2ηb1− ηbηs

I higher ηb, ηs ⇒ higher commission rate; α∗1 = 1

(1−ηsηb)(1+ρσ2s )

I “pay to play” ... α∗2 = −α∗

1ηs

I firm is better off with stronger network effects (both ηb and ηs)

second metric ⇒ better tuning for multiple externalities

Hemant K. Bhargava UC Davis 07/14/2016 Selling Platforms 10 / 13

Page 19: Selling Platforms - Boston University...Selling Platforms Hemant K. Bhargava UC Davis with Olivier Rubel Platforms Symposium, Boston, July 2016 Hemant K. Bhargava UC Davis 07/14/2016

TWO-SIDED INCENTIVES FOR TWO-SIDED GOODS?

I hire agent to recruit side S , but pay him also for B sales!

ω(qs , qb) = α0 + α1qs + α2qb; w∗ = βα1 + α2ηb1− ηbηs

I higher ηb, ηs ⇒ higher commission rate; α∗1 = 1

(1−ηsηb)(1+ρσ2s )

I “pay to play” ... α∗2 = −α∗

1ηs

I firm is better off with stronger network effects (both ηb and ηs)

second metric ⇒ better tuning for multiple externalities

Hemant K. Bhargava UC Davis 07/14/2016 Selling Platforms 10 / 13

Page 20: Selling Platforms - Boston University...Selling Platforms Hemant K. Bhargava UC Davis with Olivier Rubel Platforms Symposium, Boston, July 2016 Hemant K. Bhargava UC Davis 07/14/2016

MUTLIPLE AGENTS FOR MULTIPLE TERRITORIES ?

I agents (i = 1, 2) exert indirect externality on each other, because

participation is fueled by overall network size

Qi = V + βiwi + η (Qe1 + Qe

2 ) + εi

Qsi = Vs + ηsQb + βiwi + εsi .

I one-sided network goods: η does impact optimal commission rate

(firm must use α∗1 to manage externalities across agents)

I two-sided goods: ηb now impacts α∗1

Hemant K. Bhargava UC Davis 07/14/2016 Selling Platforms 11 / 13

Page 21: Selling Platforms - Boston University...Selling Platforms Hemant K. Bhargava UC Davis with Olivier Rubel Platforms Symposium, Boston, July 2016 Hemant K. Bhargava UC Davis 07/14/2016

SUMMARY: IMPACT OF NETWORK EFFECTS ON DESIGN

One Agent Two Agents

Traditional Good β2

β2+ρσ2β2

β2+ρσ2

Network Good β2

β2+ρσ2β2(1−η)

β2(1−η)2+ρ(1−2(1−η)η)σ2

Platform Good β2

β2+ρ(σ2s+σ

2bη

2s )

β2(1−ηbηs)β2(1−ηbηs)2+ρ(σ2

s (1−ηbηs)2+η2s (σ2s η

2b+σ

2b))

optimal commission rate α∗1

Hemant K. Bhargava UC Davis 07/14/2016 Selling Platforms 12 / 13

Page 22: Selling Platforms - Boston University...Selling Platforms Hemant K. Bhargava UC Davis with Olivier Rubel Platforms Symposium, Boston, July 2016 Hemant K. Bhargava UC Davis 07/14/2016

CONCLUSION AND GENERAL INSIGHTS

I network effects create externalities on selling outcomes and risks

I compensation plan design must account for network effects,

in spectrum of ways depending on type of network good

I firm must deploy suitable number of incentives, and in suitable ways,

to manage multiple externalities

Hemant K. Bhargava UC Davis 07/14/2016 Selling Platforms 13 / 13