Define Terms, or Doom Your Project
In the world of information management, we think a lot about the critical connection between good information and good decisions. After all, we know that the very best analytics, when applied to information that is inaccurate and out of date, can lead to decisions that send a business in the wrong direction.
To help organizations have a better basis for decision making, we focus our efforts on continuously improving our tools that consolidate, integrate, transform and cleanse the data that feeds the warehouses where analytics get applied. And there’s no doubt this whole process is important.
Now along comes an article in Harvard Business Review to remind us that the process doesn’t end there. We can take our disparate, unruly data and scrub it for squeaky-clean presentation to a data warehouse, and we can apply the best analytics to the information. But ultimately, as Shvetank Shah, Andrew Horne, and Jaime Capellá of Corporate Executive Board remind us, there’s a person who must review the results and make a decision. And that person may not have the requisite skills to make the right decision.
The CEB authors go on to highlight key problems their research has identified as preventing organizations from realizing better returns on big data. Particularly noteworthy is their conclusion: “Managers need to wake up to the fact that their data investments are providing limited returns because their organization is underinvested in understanding the information.”
There are many ways of increasing the understanding of information in an organization. For example, as the authors of this article point out, organizations can invest in training knowledge workers in the fine art of using information to make decisions.
I would submit that another approach that’s also worthy of investment is facilitating – and then increasing – the collaboration of business and IT on defining and sharing the terms that are used in running the business.
If an organization has a goal of increasing the number of customers, for example, what does “customer” mean? Is it a global corporate entity? Is it a particular division of a corporate entity? Is it a particular location of that division? Does a unique department count? Is it an individual user of a product? Must the individual meet any criteria—age, perhaps—to be considered a valid customer?
If business and IT can collaborate and agree on critical definitions like these, they’re off to a good start in the process of capturing, presenting and analyzing information to drive better business decisions.
On the other hand, failure to solidify agreement on critical definitions can doom an information-intensive project.
Employee ID? It All Depends . . . .
If you’ve walked in the shoes of a company like one I describe below, though, you may realize that this is often easier said than done.
A large manufacturer had 37 different definitions of the term “Employee ID” across multiple divisions. Using spreadsheets to record and compare the representations, the company discovered at least 15 different data types and formats, with different characteristics and usages — a result of multiple legacy systems, acquired companies with their own systems, and home-grown applications.
So what difference did it make? People in IT — the ones who needed to deliver the reports — required a tremendous amount of experience and undocumented knowledge to determine which “Employee ID” data to use in which circumstances.
The data quality issues became unmanageable, with the wrong data sources being used in various applications and reports. Developers and analysts spent enormous amounts of time trying to determine what data to use where. Of course, the data quality issues then affected the business people trying to make decisions based on questionable data.
IBM is delivering tools that facilitate the critical collaboration that’s needed. But more on that another time. In the meantime, if you’ve had a similar experience, I’d love to hear about it.