Big Data Francesco Civardi - THE INNOVATION GROUP · Data • Collecting • ... • Vesenda...

21
Analytics for Product Profitability, Customer Profitability, Organizational Unit Profitability and Risk Management Francesco Civardi, DaisyLabs 17th October 2013 BIG DATA ANALYTICS CONFERENCE 2013

Transcript of Big Data Francesco Civardi - THE INNOVATION GROUP · Data • Collecting • ... • Vesenda...

Page 1: Big Data Francesco Civardi - THE INNOVATION GROUP · Data • Collecting • ... • Vesenda eLegere, MSFT SQL Server, Analysis Services, PowerPivot and PowerViewer, RoamBI, KNIME,

Analytics for Product Profitability, Customer Profitability,Organizational Unit Profitability and Risk Management

Francesco Civardi, DaisyLabs

17th October 2013

BIG DATA ANALYTICS CONFERENCE 2013

Page 2: Big Data Francesco Civardi - THE INNOVATION GROUP · Data • Collecting • ... • Vesenda eLegere, MSFT SQL Server, Analysis Services, PowerPivot and PowerViewer, RoamBI, KNIME,

22

Basic Ideas

Analytics for product profitability, customer

profitability, organizational unit profitability,

risk management:

• Profitability and Risk go hand in hand, but are

often managed by different Company Units

• It is essential to integrate and correlate the two info

• Temporal and spatial correlation must be considered

• Bisociative thinking can give new insights

Page 3: Big Data Francesco Civardi - THE INNOVATION GROUP · Data • Collecting • ... • Vesenda eLegere, MSFT SQL Server, Analysis Services, PowerPivot and PowerViewer, RoamBI, KNIME,

33

Agenda

• DaisyLabs:

• Who? Where? What? How?

• The Kernel (core components of our solutions)

• The Engines, the WebConsole, the GeoConsole

• The Accelerators (working templates of «condensed experience»)

• NetRisk Monitor

• Marklus (Markov Clusters)

• Geneco (GenEconomics)

• Surcus (Customer Survival)

Page 4: Big Data Francesco Civardi - THE INNOVATION GROUP · Data • Collecting • ... • Vesenda eLegere, MSFT SQL Server, Analysis Services, PowerPivot and PowerViewer, RoamBI, KNIME,

44

DaisyLabs: who?

• DaisyLabs started ten years ago, as an IT Company

devoted to IT consulting through state-of-the-art

technology.

Page 5: Big Data Francesco Civardi - THE INNOVATION GROUP · Data • Collecting • ... • Vesenda eLegere, MSFT SQL Server, Analysis Services, PowerPivot and PowerViewer, RoamBI, KNIME,

55

DaisyLabs: where?

• We moved in 2012 to the Pavia Techno Park, in order to

strenghten our contacts with the research environment,

and to attract more easily the best graduates

Page 6: Big Data Francesco Civardi - THE INNOVATION GROUP · Data • Collecting • ... • Vesenda eLegere, MSFT SQL Server, Analysis Services, PowerPivot and PowerViewer, RoamBI, KNIME,

66

DaisyLabs: what?

Our mission is to «Deliver Information Assets», through

Data

• Collecting

• Organizing

• Integrating

• Visualizing

• Analyzing

• Presenting

Page 7: Big Data Francesco Civardi - THE INNOVATION GROUP · Data • Collecting • ... • Vesenda eLegere, MSFT SQL Server, Analysis Services, PowerPivot and PowerViewer, RoamBI, KNIME,

77

DaisyLabs: how?

• We work closely with our customers – better, partners.

• We build together mixed teams

• We inject them with our knowledge and our

methodology.

• We are fully commited on the results

Page 8: Big Data Francesco Civardi - THE INNOVATION GROUP · Data • Collecting • ... • Vesenda eLegere, MSFT SQL Server, Analysis Services, PowerPivot and PowerViewer, RoamBI, KNIME,

88

Kernel: the Engines

• Daisylabs integrate in their solutions the best analytical

engines available. We haven’t married a specific

technology, but we choose it according to the

ecosystem of the customer.

• Vesenda eLegere, MSFT SQL Server, Analysis

Services, PowerPivot and PowerViewer, RoamBI, KNIME, Panorama Software and R are examples of

technologies used.

8

Page 9: Big Data Francesco Civardi - THE INNOVATION GROUP · Data • Collecting • ... • Vesenda eLegere, MSFT SQL Server, Analysis Services, PowerPivot and PowerViewer, RoamBI, KNIME,

99

Kernel: the WebConsole

• The WebConsole collects information coming from

Company reporting systems or from analytical systems

(developed by DaisyLabs or by others) in a user-friendly

environment accessible by different people in a strictly

controlled way.

• Profitability and Risk Information can be shown side by side.

9

Page 10: Big Data Francesco Civardi - THE INNOVATION GROUP · Data • Collecting • ... • Vesenda eLegere, MSFT SQL Server, Analysis Services, PowerPivot and PowerViewer, RoamBI, KNIME,

1010

Kernel: the GeoConsole

• The GeoConsole is an innovative way proposed by DaisyLabs

to visually integrate “mapped” information assets.

• Information coming from internal or external sources are

organized in channels available for user analysis.

• Every channel is an independent information source; different

channels are displayed on a map for visual exploration and

discovery.

• Risky and profitable areas can be identified, and visually

“space-correlated”.

10

Page 11: Big Data Francesco Civardi - THE INNOVATION GROUP · Data • Collecting • ... • Vesenda eLegere, MSFT SQL Server, Analysis Services, PowerPivot and PowerViewer, RoamBI, KNIME,

1111

Visual Data Exploration11

Page 12: Big Data Francesco Civardi - THE INNOVATION GROUP · Data • Collecting • ... • Vesenda eLegere, MSFT SQL Server, Analysis Services, PowerPivot and PowerViewer, RoamBI, KNIME,

1212

Mapped reports12

Page 13: Big Data Francesco Civardi - THE INNOVATION GROUP · Data • Collecting • ... • Vesenda eLegere, MSFT SQL Server, Analysis Services, PowerPivot and PowerViewer, RoamBI, KNIME,

1313

Accelerators: RiskNet Monitor

• RiskNet monitors the risks taken by financial institutions,

through their commercial networks.

• The engine of the system computes indicators from

operations stored in the DW, calculate risk rates and

aggregate them along the Organizational Structure, properly

weighing them

• The webconsole raises alarms, when unexpected risk rates

show up, and allows users to go from risks back to indicators,

in order to understand root causes of potential problems.

13

Page 14: Big Data Francesco Civardi - THE INNOVATION GROUP · Data • Collecting • ... • Vesenda eLegere, MSFT SQL Server, Analysis Services, PowerPivot and PowerViewer, RoamBI, KNIME,

1414

Accelerators: Marklus (Markov Clusters)

• This accelerator enables the statistical segmentation (i.e. clustering) of business entities (customers, products…)

• Segmentation can be based on risk info, on profitability info, or both.

• Segmentation is not static but dynamic; e.g. the goal of marketing efforts (Campaigns etc.) is indeed moving customers from less to more profitable segments.

• Cluster analysis is therefore dynamically updated, and enriched with Markovian analysis, to analyze transitions between segments.

14

Page 15: Big Data Francesco Civardi - THE INNOVATION GROUP · Data • Collecting • ... • Vesenda eLegere, MSFT SQL Server, Analysis Services, PowerPivot and PowerViewer, RoamBI, KNIME,

1515

Bisociative thinking

«The pattern underlying the creative act is the

perceiving of a situation or idea, L, in two self-

consistent but habitually incompatible frames of

references, M1 and M2.»

Arthur Koestler, The act of creation

15

Page 16: Big Data Francesco Civardi - THE INNOVATION GROUP · Data • Collecting • ... • Vesenda eLegere, MSFT SQL Server, Analysis Services, PowerPivot and PowerViewer, RoamBI, KNIME,

1616

Accelerators: Geneco (GenEconomics)

• This accelerator has been inspired by our research work on genomics, where the problem is to find which genes are differently expressed in cancer tissues, with respect to normal tissues.

• It scans hundreds, or even thousands, of variables (genes, but they can be products, product attributes, customer attributes, nodes of a commercial network etc. ) and identify the best «predictors» of desired vs undesired status (loyalty vs churn, wealth vs bankrupcy, good vs bad quality, good vs risky branch, profitable vs unprofitable customers or producs etc.)

• Visualization techniques used in genomics, like heatmaps and vulcano plots, can be easily adapted to a business scenario.

16

Page 17: Big Data Francesco Civardi - THE INNOVATION GROUP · Data • Collecting • ... • Vesenda eLegere, MSFT SQL Server, Analysis Services, PowerPivot and PowerViewer, RoamBI, KNIME,

1717

Vulcano plots17

X: Log Odds, a measure of effect size

Y= AUC,a measure of predictive power

e.g. P25 is a predictor of «good»,P33 of «bad»

Page 18: Big Data Francesco Civardi - THE INNOVATION GROUP · Data • Collecting • ... • Vesenda eLegere, MSFT SQL Server, Analysis Services, PowerPivot and PowerViewer, RoamBI, KNIME,

1818

Heatmaps18

Visuallyanalyze correlations

Page 19: Big Data Francesco Civardi - THE INNOVATION GROUP · Data • Collecting • ... • Vesenda eLegere, MSFT SQL Server, Analysis Services, PowerPivot and PowerViewer, RoamBI, KNIME,

1919

Accelerators: Surcus (Customer Survival)

• This accelerator has been inspired by our research work on clinical prediction models, and particularly on survival analysis, which is now often applied to business problems (e.g. see G. Linoff & M. Berry «Data Mining Techniques»).

• A special regression technique (Cox regression) is applied, in order to identify the drivers of churn, or more generally, of customer «mortaliy», and survival curves (Kaplan-Meyer) are compared.

• This technique can be applied together with Cluster Analysis.

19

Page 20: Big Data Francesco Civardi - THE INNOVATION GROUP · Data • Collecting • ... • Vesenda eLegere, MSFT SQL Server, Analysis Services, PowerPivot and PowerViewer, RoamBI, KNIME,

2020

Kaplan-Meier20

Different Customer Groups (Clusters)have different«Survival Curves»,i.e. different risk of leaving

HR= Hazard risk, p=stat. significance

Page 21: Big Data Francesco Civardi - THE INNOVATION GROUP · Data • Collecting • ... • Vesenda eLegere, MSFT SQL Server, Analysis Services, PowerPivot and PowerViewer, RoamBI, KNIME,

2121

Questions & Answers

Thanks for your attention!

[email protected]

21