Events versus transactions in marketing analytics

Information management strategy lags the development of technology in most industries. We have seen the slow march of progress towards customer intelligence progress from initial customer identity management in the late 1980s through consolidated customer position snapshots (Customer Information Files or CIFs), crude velocity metrics (recency, frequency monetary or RFM), behaviour based customer value | profitability metrics, to monitoring customer behaviour.

During this procession of learning there have been many diversions of effort into unfruitful investment because the technology was pushing ahead of management thinking. Striking examples are not hard to find: database technologies drove investments in ERP, SCM and similar technology-enabled management methods, rarely with any discernable return to shareholders. Similarly the introduction of Enterprise scaled data warehouses enabled development of the CRM boom, first in the form of contact management later as customer experience based interaction engagement models.

Most projects failed to deliver promised benefits to customers or shareholders and for good reason: the rarified atmosphere of an overheated economy allowed managers to buy into the visionary states being promoted by vendors of technology. Unfortunately technology vendors are mainly interested in selling data storage, processing and analysis tools rather than actually managing your business. They don[t really know what will work for you and why, because they don't know enough about your business - which is perfectly reasonable.

One of the later entrants on the scene has been Transaction Trigger Analytics. First champinoed by banks in Australia (notably NAB,) it was discovered that parsing through transaction files overnight could result in identification of significant changes in customer accounts which, when acted on within 24 hours, could change customer behaviour. By detecting a significant deposit, for example, the bank could contact the customer to ensure all is well and offer any new services that the customer might require, such as investment advice. This technique is used primarily to keep new money in the bank or to keep old money from leaving.

Their experience proved the businesscase for transaction trigger detection - ROI was very high. The technology vendors were delighted - now banks had a good reason to store all their trnasaction files and load their databases up with new data every day instead of periodically as had been the norm. This meant lots of new extract, transformation and loading processes, lots of new storage requirements and lots of new processing power requirements to grind through massess of data every night.... a vendor's dream if there ever was one !

The only problem is that transactions are not a good representation of customer behaviour. Yes they are what what changes accounts, but this is from a company perspective (or more accurately an account management system perspective) which is not the same as customer perspective. Customer behaviour can bve far more comples than "significant deposit" showing in a transaction file. For example, that transaction could arise from a tax refund; sale of a property or business; transfer of an investment account; liquidation of investments; relocation of an account between locations and the list goes on. We have discovered that over 1/4 of banking balance changes result from internal flows of money within a customer's existing relationship.

This means that the transaction triggers will be false positives nearly half the time. Why ? because for every significant internal "plus" there is a corresponding "minus" so each side of an internal transaction appears to be a signirficant transaction event trigger. Transaction triggers can generate false leads about half of the time, draining staff time, program credibility and, worst of all, annoying customers with pointless dialogue.

What banks and other organizations need is to better define customer behaviours in the context of their business relationships. Know what customers really do and model these customer behavioural events . Then aply detection mechanisms to find and route real customer behaviour changes to your customer service staff. Better quality leads to improvements in efficiency, effectiveness and satisfaction for customers, employees and shareholders simultaneously. Stop wasting time with transaction detection - it is too primitive a tool to be relevant in today's customer management environment.

- David B. McNab

Events versus transactions in marketing analytics