Customer Insight - Solution Approach...

There are a number of processes needed to deliver a really effective solution which starts with the linkage of the disparate data in the sales and account systems. As it is doubtful the unique reference numbers in both files could be used for linkage, one could try to create match keys based on parts of the name and address.

However, if the orders are placed by a purchasing department with delivery to yet another department or address, and the sales contacts in the separate sales lead system have been created independently with no address standardisation, very often a much more complex and part fuzzy-logic/derived match key approach will be needed to create the link.  Specialist conversion, merge/purge and optional address standardisation software would be used in conjunction with tailored methodologies developed specifically for the task and the data.

Once data has been cleaned and unique link numbers created, the files would be imported into a statistical analysis programme such as SPSS and rules determined for the merge and also how best to structure the data for analysis - given that most analytical procedures require data to be held in a single record where relationships between each field can be examined. The merged data would then be appropriately coded and restructured to create the new summary data fields that are needed for marketing.

Next, techniques such as factor, cluster, regression, CHAID (part of SPSS AnswerTree, see picture to right), etc., would be used to undertake segmentation, predictive and response modelling. Alternative approaches could also be used for data exploration and modelling using software such as 4Thought (neural network) or KXEN (SRM Learning Machine).

For database change analysis and trading behaviour tracking, summary snapshots of the data would be taken at agreed intervals and compared to derive new percentage change indicators. Additional customer understanding could be derived though market research using software such as QPSMR.

Finally, the unique reference numbers and cluster ID's, predictive scores, profile data, etc., could be input into operational marketing, sales and call centre databases to optimise the targeting process.  For a really powerful targeting solution, the transformed data could form part of the load into a high speed campaign planning, selection and targeting system such as KbaseT which will provide the marketer with a complete and single customer view.  Click here for more details.

During the data restructuring and analytical processes, the syntax (computer code) used to run the tasks would be saved so that each process could be automatically repeated as often as required.