Turning Data into Knowledge

When you think about it, many concepts in our world are misnamed. Chief Information Officers are really Chief Technology Officers; business analysts are usually just systems analysts. Our entire field is a misnomer. It’s called information systems or management information systems, but I would argue that the old term `data processing’ is still a better fit for how we work most of the time. And the Information Age has really been the Data Age.

It’s unfortunate, but we have focused too much on mastering transaction data and not enough on turning it into information and knowledge. We’ve talked about it for years, but it doesn’t happen very often. The situation is reminiscent of 16th-century England, when Sir Thomas Gresham, an English financier, popularized Gresham’s law. It’s usually stated as, “Bad money drives out good,” meaning that worthless or debased currency drives more valuable money out of circulation.

Today we might refer to Gresham’s information law: bad transaction data drives out good knowledge. Not that transaction data itself is bad – it’s a necessary precondition for turning data into knowledge – but the continued focus on data at the exclusion of higher-value forms of information is not doing us much good. In the rush to ensure that employees and suppliers are paid, orders are taken accurately, and debits and credits are posted to the right ledgers, we have neglected the realm of analyzing and interpreting trends in the data and acting on these insights.

We have the technology to do this, but something else is missing.


What’s the evidence for this outlandish hypothesis? I’ll give you several small bits of data and two large ones. The first big bit relates to ERP. I have just finished a research project funded by SAP AG in which I was trying to find evidence that the great new data supplied by that package – integrated, cross-functional, real-time and global – is being used to change management and decision-making processes. In other words, I wanted to see if the data was being turned into information and knowledge. And it was at a few places such as Amerada Hess, Dow Chemical and Microsoft. These companies are making great progress in educating workers and managers about what data is available, how to build data warehouses or marts, and how to create organizational structures to do analysis and act on the results.

But there weren’t enough of these success stories. In many cases, when I asked managers what they were doing with SAP data, they said, “Talk to me in a couple of years,” or “We’re focusing on getting the basic processes and transactions right.” I’m worried that by the time they get around to focusing on using the data it will be time to install R/4 or R/5. Or to put in that new R/3 module to manage janitorial services.

The second big bit of evidence involves customer data. Companies gather a lot of it in transaction systems, but they don’t do much with it. Many companies still find great difficulty in the number of systems across which customer data is spread, the poor quality of customer data, and the organizational politics involved in addressing the issue. The other problem with customer data is that organizations gather too much of it.

A team that’s developing a customer data warehouse asks everyone what customer data they want and throws it all into a big warehouse. It’s then often too big for people to find and understand the right data. The customer-data equivalent of the SAP story described above occurs when companies install customer asset management systems to support sales and service transactions. These systems could also be used to support knowledge management initiatives such as customer problem-resolution, forwarding customer and product knowledge to other parts of the organization for trend analysis. Instead, they’re used for tracking trouble tickets.

Now for some piddling little examples of the data-to-knowledge deficit. Take scanner data in retail stores. The CIO of a grocery chain that is highly regarded for its IT use once confided to me that his company analyzed at most only 2 percent of the data it collected. I heard recently that a Midwestern grocery chain had decided to throw away its scanner data. Previously, the data had been saved for years in the hope that someday it would be analyzed, but it never was.

Web transaction data is increasingly following the same pattern. I spoke with the designer of a large information technology firm’s Web page. He described a series of changes to the page over time that involved such fashionable features as streaming audio and video, and lots of white space. I asked him whether anyone had ever analyzed the site transaction data to see what visitors actually did there. He said, “Oh, sure, we found out that what most people did on the top-level page was to initiate a search, so we made the search button larger.” And that was the extent of Web transaction knowledge.

Then there’s basic financial and HR data. The first IT application in business in 1953 at General Electric involved payroll. We’ve had variations on financial and HR data processing ever since.

Yet as a recent McKinsey & Co. study points out, most companies do a very poor job of managing their talent. They don’t know what skills they need or what their people have. This situation doesn’t have to prevail. A manager at Ford said the company had bought the PeopleSoft package not to handle basic HR transactions but to capture and manage HR knowledge for purposes of talent management, succession planning and general HR analysis and decision making. Why don’t more companies do that?


We have the technical capability to transform data into knowledge. In fact, in the 20 years or so that I’ve been paying attention to the IT management field, we’ve had several generations of analytical technology. Sometimes I suspect that only the names have changed over time. Let’s see, there was decision support systems, then executive information systems, then Olap/Rolap/Molap, then data warehousing and mining. Have I missed anything?

Peter Keen, one of the founders of the decision support movement, noted recently that decision support felt an awful lot like data warehousing to him. I heard someone say recently that they were doing some data mining in her company, and I asked what technology they were using. “It’s called SPSS,” she said. That meant I had started my data mining career in 1975! So we’ve not been lacking in technology; but something else has been missing.

What is it? As with so many other things in the realm of IT, the missing ingredient is people. You need fast hardware and capable software, but what most companies overlook is the “wetware.” What, you might ask, is the wetware architecture for successful data-to-knowledge transformation? What kinds of people, skills and organizational structures are necessary to pull it off? To be honest, I don’t really know for sure yet, since I am studying that very topic in a new research project. But I do already know that whenever I have come upon an example of the successful use of data for decision making and management, the following types of people seemed to be around: