Customer Segmentation approach  

Customer Segmentation and Profiling

Customer Segmentation and profiling is the first step before we decide on how to retain, satisfy and maximize their profitability.

Customer Segmentation approach

This is just a high level statement on the approach, and some cautions along with. In customer segmentation, one has to be prudent to maintain a balance on the level of segment. One also has to know early on the invalid or irrelevant segments.

It’s easy to create a list of parameters on which you can segment a customer, but one needs to be extremely smart to ensure that one does not get lost into over-complexity. Before one gets into final round of customer segmentation and defining the specific actions you will take for those segments, one can have following approach:

Steps of Customer Segmentation Approach

STEP 1 - Define core (defining) vs. non-core (fine-tuning) parameters:

Question is on what are the top line parameters, which will make a difference to your strategy. For example if your strategy is to be a cost-leader, following could be among your defining segmentation parameters:

  • Income group (lower income groups will be higher potential customers),
  • Usage patterns (a higher usage is expected as that will give you good revenue on lower margins),
  • Risk profile (you cannot have a high risk with low margins)

STEP 2- Customer Segmentation at single core parameter level

One can first identify the customer segments at a single parameter level. This means that one should make multiple sets of customer segments:

  • One set of segments for their risk profile. (high risk, medium risk, low risk..)
  • One set of segments for product affinity. (Market basket analysis..)
  • One set of segment for Existing/potential value/profitability. (High value, medium value, low value..)
  • One set of segment for customer satisfaction. (Highly satisfied, Medium Satisfied, Less satisfied, least satisfied..)

The above can be done by running data mining routines (OR a less effective, but The above can be done by running data mining routines (OR a less effective, but “to start with”

STEP 3- Customer segmentation across multiple core parameters:

You can now make next level groupings on combination of multiple parameters, for example – risk profile along with income group and customer profitability. This will generate clusters of customer segments. A customer in this case will belong to a single cluster, for a given combination parameters.

STEP 4- Validate the Segment clusters, and churning out the invalid OR sparse combinations of core parameters

You can take out the combinations which are either invalid combination OR are sparse combinations (with very few customer belonging to it)

STEP 5- validate your actions, to go through the re-iterations

Once you get the details on how many customers are lying in a given combinations of parameters, you will do the following:

  • Check if the numbers of customers in each cluster, are aligned to your strategy Assumptions

    Let’s say that you have the strategy assumption of large population of high usage, low income group customers. However, after the segmentation you may find the your assumption is not correct.

  • Check the possible actions on these segment are aligned to your SWOT:

    Let’s say that you have a cluster of customer, which will demand a major change in the way your product is to be designed. If you feel that you that product design is your weakness, you may like to nullify this cluster and redefine your segmentation strategy.

STEP 6 - Repeat Step 1 to Step 5

NOTE- This is a highly iterative exercise, and once you have done certain iterations, you may like to merge step 2 and step 3.

NOTE- The above process can be applied very successfully on the existing customers as you have a good amount of data on those customers. For potential customers, one has to rely upon many other methods to get the data, which will be incomplete and partially-reliable in most cases. In that kind of scenario, one has to work on Data Correction, and Enrichment as well as do smart extrapolation. One can also rely upon 3rd party research companies OR industry associations to get the data.