Don't let CRM data make MIS into misinformation

David Taber

CRM systems are fundamentally different from the rest of enterprise software. This isn't an issue of technology but, rather, of people, process and policy:

  • CRM systems are political, far more so than any other part of MIS. The main drivers are the personalities and decision habits of sales executives - particularly their lack of longevity in specific positions on the org chart.
  • CRM systems have much deeper and more pervasive data quality problems than any other part of MIS. This is true both in individual records (plain old bad entries) and at the metadata level (blurred semantics, uncoordinated changes across different tables and so on).
  • CRM systems (and customer relationships) evolve for their entire lifespan; the no-change-at-all interval for any particular function point is typically measured in months. Legitimate business need (the competitive and channel worlds change quickly) and arbitrary politics ("I'm in charge now; do it my way") drive this.
  • CRM systems need to be integrated with more other business processes and IT assets than just about anything else. This ranges from phone systems and automatic call distributors to accounting to distribution to reverse-logistics. This issue, particularly for sales and customer service, exaggerates the issues of the first three bullets.

These four issues are facts of life, but none, in and of itself, is particularly dangerous to IT and MIS credibility. So what's the problem here?

CRM Data, Good or Bad, Ends Up Everywhere
Reports. Dashboards. Extracts. These are all distillations of the data in your CRM system. In this case, the distillations exaggerate the impurities, rather than removing them.

The slightest semantic or data-quality issue in the underlying records gets magnified into its own group-by category, roll-up cell or pivot table column. Instead of having 687 customers in the United States, you have 325 in the US, 282 in the USA, 77 in the United States of America and three in a country called "60609."

Let's play this forward into the context of the enterprise. By default, most CRM systems give users the permission to create their own reports; some make it really easy with nice drag-and-drop editors. Those pesky users will run off and make reports without first understanding the object model, semantic distinctions or required filtering. Users have a very low chance of properly applying Boolean logic, either. So most of the reports are junk when written - and, thanks to the constant semantic drift characteristic of CRM systems, results degrade over time.

Our firm informally surveyed more than 100 CRM installations and found that, on average, users create about one new report per quarter for the life of the system. If you've had 100 users on your system for five years, expect something like 2,000 reports.

1  2  Next Page