Solutions by Function

Segmentation & Clustering

Segmentation is a way to have more targeted communication with the customers. Based on the adage 'not every customer is the same', techniques in statistical clustering and segmentation are employed to group 'identical customers' from a pool of customer population. Customers are segmented based on a variety of variables like geographic/demographic/behavioral, and the clusters thus received are further used as a populous to promote a sales strategy or a marketing campaign. Because each of these segments contains customers possessing a similar profile, these attributes form a good basis to reflect & design outcomes of various marketing channels.

Based on the business problem and the quality of data available, appropriate technique is employed to segment the customers using SAS, SPSS, R, or MS Excel -

Scoring Technique

Using know-how of the trade and statistical measures of count (weights/scores for each attributes), each of the customer is assigned a score from 1-10. Customers are then ranked and plotted across a variable to capture the inflection points for each variable i.e. the score where change in behavior is most significant. Further, the weights/scores are recalibrated to find a 'good fit' of customers who are able to follow Pareto's Principle i.e. 20% of the customers who are bringing 80% of the effect. This technique is most used in cases where variables are categorical.

Linear & Logistic Regression

Using target definition, a behavioral model is built on the many demographic and behavioral variables contained in the data. Intended targets are identified in the population and each customer is given a score on 1-10 that demonstrates the propensity of the event rate we are trying to measure. The model is then verified using out of sample & out of time data.Mostly used in the cases where variables are numeric and targets can be identified in the historical data.

K-Means Clustering

Using a hierarchical method to define the number of clusters, the K-means procedure is then employed to form the clusters in the customer population. Being one of the simplest unsupervised learning algorithms that solve the clustering problem, the method is mostly used in cases where target definition is not clearly available.

Our Experts are here to help you in your needs Contact Us