Analytics and Insights in Product Management

Of all the sectors where there's most fun to be had if you love data, the Retail, Telecommunications and Banking sectors have to be right on top of the pile. It is sure to be the space where the most variety of customer insights and analytics are to be uncovered

While this particular case relates to banking, the example could just as well be applied to the telecommunications and retail industries, indeed, anywhere where a product or service is sold to many customers, B2B or B2C, and appropriate data is collected. 

---

The client: A top global bank
The problem: Demonstrate that the value of the data warehouse was beyond mere reporting
The solution: Demonstrate how data could be used to drive product-level marketing strategy

The data warehouse: Oracle
Data quality: Balanced and audited, with an analysis for completeness and cleanliness
The data set: Five years of product acquisition and transaction data for six millions customers

Analysis tools: Custom

The process and findings:
Given that marketing can generally be considered a demand-generating activity, the proposal was to consider product acquisition in more detail. Consider Product A.

Figure 1: Product acquisition for the client bank for the period under review

Now the regular way of considering product acquisition performance is by considering new product volumes against the product volumes acquired during the previous period, where a period could be month, quarter, half year or even year.

This type of analysis is generally used simply to track measures such as performance against budget, but it doesn't really give you much insight beyond this.

As can be seen from Figure 1, it seems the product sales are growing year on year, which is generally considered a good thing. However, is the picture as rosy as this? Are there any potential points of marketing optimisation along this growth trajectory?

One type of simple analysis, intuitive with many marketers, could consider the entire growth trajectory over five years as illustrated in Figure 1, and strip out the trend in order to see whether cycles in the sales performance of the product exist. This analysis technique is popular with econometrics practioners.

Figure 2: The seasonal sales performance of the product

By doing this, we can create an index of the performance of sales over each month of the year, such as illustrated in Figure 2, where month 1 equates to January, and an Index measure of 1.0 represents average annual performance. So for example, month 5 (May) has a sales level, as measured over five years, of the average monthly sales of the entire year. In this case, the marketers knew about the sales trough occuring around Easter, and the sales peak occurring in the first three months of the year, but few knew about the statistically significant sales peak occurring in July and August every year.

Recommendation: There was sufficient argument for marketing to change their marketing calendars in order to try and increase the value that could be gained from marketing campaigns during July and August. Marketing had traditionally been low key during the summer holidays, until this analysis demonstrated that there was above-normal sales activity during this period which could likely be increased with targeted campaigns.

Figure 3: Establishing the presence of a long term acquisition cycle

After a little more analysis conducted in the attempt to determine whether a longer cycle exists than the 12-month cycle specified above, we see there exists a distinctly longer term cycle peaking approximately every two years with this product, once at the beginning of 2009, the next at the beginning of 2011, and a third forming at the beginning of 2013.

While this is a fascinating discovery, no satisfactory reason could be found for the existence of the cycle even with the assistance of the client. It is probably linked to something in the macroeconomic environment, but it has not been isolated yet. It remains an enigma at this time. One a link is found, it will need to be tested for causality to ensure the link indeed has a cause and effect relationship with the acquisition of the product.

Recommendation: Certainly an interesting cycle that is clearly visible, but useless in a business context until the reason for the cycle can be established.


This blog article is getting pretty long. More analysis could be conducted at a per-product level including product life cycle analysis, but that will be documented in a separate article.

Finally, note that this analysis was conducted for every single product type and sub-type for the client, providing valuable marketing insights for mortgages, credit cards, savings accounts, unsecured credit and secured credit products. Of course, there may be differences in the outcomes per market segment, per geography and surely for many other attributes, which demonstrates why it is so important that analytics processes are mechanised, not only for the volume of work that this enables, but also to enable you to experiment and to test many different hypotheses within the business.

---

This article has been a simple demonstration of the power of analytics for providing product insights that help drive strategic business decisions. Of course, there are plenty of assumptions that precede any decision-making, and these need to be made explicit at the point of presentation to ensure that any decision-making is made in an appropriate context. It is so easy to make the wrong decisions if the context of the data is not properly understood.