Customer data model for analytics

Which DATA INPUTS are required to answer the business question

Out of the plethora of business questions which frame the 'optimal contact strategy for a stores loyalty card' holders, What data inputs are required to answer the business question. In this case its

1. Loyalty card membership information
2. POS Data of loyalty card holders
3. Purchase channel

Ellaborating Loyalty card information further we have
- Loyalty card number
- Customer name
- Member since
- Profession
- Age / Date of birth
- Sex
- City
- State
- Zip

Ellaborating POINT OF SALE data furtger we have
- Store identifier
- Purchase date
- Total Purchase amount
- Product identifier
- Product quantity
- Product value

Framing the customer behavior related questions which help in optimizing the decision

Lets say that we have zeroed in on the decision to optimize. Taking the example of a retail context lets say that we want to "Optimize the decision to initiate an outbound communication to a loyal customer". What are the business questions which define our attempt to optimially 'touch' the customer.


- What are the behavorial characteristics of my loyal customers who shop in my store ?
- What is the mix of products that they purchase typically ?
- Is this behavior in line with normal purchase behavior for this market segment ?
- What are the underlying drivers of shopping behavior ? Is it price, brand, niche products, instore experience, customer service, optimal shelf placement ?

Zeroing in on right Customer Decision to optimize

Please click on above to see a much larger image of customer decisioning framework

Very often it is easy to get seduced by the statistical techniques and various algorithms and start off analytics in an area which is in the comfort zone of the practitioner. The classic "Hammer looking for a nail in the wall" problem. For example it is common for analysts who have done multiple churn scoring models to assume that scoring customers for churn is the best place to start deriving value in an analytical engagement. But taking a structured approach revealed in this case that trying to build a customer propensity model for cross selling complimentary products provided more value than the churn score model. How does one not fall into such traps ? Here are some key questions to ask before applying analytics to a business processes are

Which are the possible Customer facing ecisions which I can optimize for my operations ?

What is the current yield of these customer decisions which are taken ?

Which of these are strategic customer decisions ? Which of them are operational customer decisions ?
Which of the customer decisions , if optimized bring additional revenue to my organisation ?
Which of the customer decisions , if optimized reduce cost for my organisation ?

This is possibly best explained with an example. Lets take the example of a retail store and apply the above questions framework to retail industry

STEP-1.1 : What are the 13 core decisions taken in a retail industry ?

- Cross sell decisions
- Migrating Valuable customers to 'High touch channels'
- Promotion decisions ( Temporary price reductions, Gift offs, Coupon promos )
- Decision on where to open/close new Stores ( location decisions )
- Decisions on Pricing for newly launched products
- Which slow moving products to move off the shelf using campaigns
- Mix of product Categories to stock
- Value added retail Services to bundle
- Decisions regarding which customers to waive Loyalty fees for
- Decisions regarding which interventions to win back loyal customers are at risk
of churning
- Decisions regarding migrating low value customers to email and web channel instead of call center channel ( Channel migration decisions )
- Decisions regarding store layout
- Decisions required to heighten in store experience
STEP-1.2 : Which of these are strategic in nature ? ( window of 4-6 months )

- Decision on where to open/close new Stores ( location decisions )

- Decisions on Pricing for newly launched products

- Decisions regarding store layout

- Decisions required to heighten in store experience

STEP-1.3 : Which of these are OPERATIONAL in nature ? ( every day, every week decisions taken by front line folks )
- Cross sell decisions

- Migrating Valuable customers to 'High touch channels'

- Promotion decisions ( Temporary price reductions, Gift offs, Coupon promos )

- Which slow moving products to move off the shelf using campaigns

- Mix of product Categories to stock

- Decisions regarding which customers to waive Loyalty fees for

- Decisions regarding which interventions to win back loyal customers are at risk of churning

- Decisions regarding migrating low value customers to email and web channel instead of call center channel ( Channel migration decisions )

Customer behavior analytics - A framework

Customer behavior analytics is an area where there is a lot of new found interest . Given the fact that a lot of organisations have collected terrabytes of detailed raw customer transaction information using CRM,ERP,Online and home grown applications, its imperative leverage the insights bueried in tons of raw data. The starting point in understanding customer behavior information is to define the questions which frame the issue at hand. Some questions framing this issue at hand are


QUESTION-1 : Which are the top 5 revenue impacting customer behavior patterns ?

QUESTION-2 :What specific channel interventions can I put in place when I encounter these behavorial pattern ? How do I operationalize my customer insights ? Which are customer facing decisions which can be optimized by the insights ?

QUESTION-3 :Given that there are tons of areas to fish for insights where do I begin ? Do I look for insights in store sales ? online clickstream info ? call center complaints ? payment information ? campaign responses ?

QUESTION-4 : What are some best practices when one undertakes the journey to generate penetrating customer insights from raw data ? What are some of the landmines to avoid while generating customer insights ?

QUESTION-5 :What solution architecture can I use to implement a customer behavorial targetting system ? What are the components of this system ? How does one define the logical architecture for a customer behavorial targetting system ?Which tools do I use - SAS ? Kxen ? R ? MSFT Datamining services ? SPSS ? Oracle Data miner ?

QUESTION-6 : Is there a method to the madness ? Can we evolve a structured methodology to surface customer insights ?

I would encourage readers to submit their questions so that we have an exhaustive list of issues which we can address in a systematic fashion. We can then take ONE QUESTION AT A TIME and seek answers to the same
This entry was posted on July 2, 2008 at 6:25 pm