By Ed Sperling
Just because you have an abundance of data and expensive tools doesn’t mean you’re making good use of them.
The amount of data inside corporations is exploding. In fact, there has never been such a wealth of information available in the history of business, and there has never been the vast array of tools to dissect it.
CIOs are generating reports for business leaders that slice the data horizontally and vertically, project growth, calculate productivity and profitability, and compare all of this historically and competitively. They are even pulling out tidbits of data that may appear randomly and building models based upon recurrences that escaped notice by even the most astute teams of experts. But is all this stuff right?
Forbes caught up with Marcus Schaper, principal at McKinsey & Co.’s Business Technology Office in Hamburg, Germany, to talk about how data is being used and where the pitfalls are.
Forbes: How do we keep up with an explosion of data and turn it into something truly valuable?
Schaper: This is a real challenge for many CIOs and many business units. We are all swamped by data. The availability of data is not the issue. It’s how we can extract information from all of that data and make good business decisions from it. That’s where many companies fail, despite the fact that many of them have very expensive and relatively good tools.
How do they solve that?
First of all, they have to ask the right questions. Then they need to derive the right structure to bring the data into. And finally they have to present it in a way that people can understand and trust. Being able to trust it is even more difficult because the data has to be in sync with the world as we know it. If the reports you get on a daily or weekly basis are different than what you know from your daily life, then you don’t trust them and you don’t trust the system.
You’re talking about creating a data model, but aren’t models inherently flawed? Look at what happened to Wall Street in 2007.
The problem with models is not so much that they’re difficult to create. It’s whether everyone involved agrees with the same semantics. If you want a revenue report or a profitability report, you need to figure out what should be included. Once you have clarity on that, the other steps are much easier.
Do these skills exist?
They’re not very common. We need a completely new skill set for the CIO and the IT department. They have to speak the language of requirements for the application as well as the business language of reports.
But you’re also now dealing with organizations that change quickly. How do you program that into the models?
It’s a question of two things. One is the underlying data architecture, which is an IT topic. You have to be able to find the modularity in the data so you can access all different cuts and views. If the company wants productivity data this year that will be different from profitability data next year, it all may be different from the growth program in year three. You need a data architecture that allows all the different cuts and views and which enables quick access. That’s one piece, and it’s a key piece. The more difficult piece is the interface between IT and the business side. If you really want to have views and cuts of the different pieces, then the business side needs to sit down with IT and agree on the structure and the semantics and the business structure. The notion of sitting together every now and then and reviewing where the business is going and what that means in terms of reports–that’s something that people don’t do on the business side. That needs to change.
Is there a risk of relying too much on data and not enough on people’s observations and insights?
People already rely on data quite a lot. All my clients get reports and they draw conclusions based upon those reports. In discussions with other people they are either validated or not. Numbers have a tendency to anchor in your head more than beliefs or opinions, but there is also a need to correct the data if it is wrong to make sure the number of anchors you have is correct as possible.
Here’s an example: One logistics company that’s a client of ours was trying to understand the profitability of its products. They get bulk part numbers, but when you try to boil it down, the average profitability does not apply to a large customer because they get different rates. The products are not comparable over different regions either, because there are different service levels and costs. So looking at the same products in different parts of the world, the overall data you have is probably wrong. Too often we use incomplete data or average data and you come to the wrong conclusions.
As companies outsource applications into clouds, does it become more difficult to gather data?
Data gathering isn’t the issue. Data is lying around everywhere and it’s not technically a problem to get together. But it does get more difficult to agree on the right indicators and the definitions behind them. If you have a small board of five people, you can sit down and agree on everything in one board meeting. If you have a global company, it’s much tougher. If it’s in a cloud it’s even harder. You might not even know your counterparts, and the risk is higher that they do not have the same interpretation of certain terms you are using. If you are using lots of different data and content providers and they give you pre-process information, it’s nearly impossible in the cloud environment to agree on the same semantics.
So where do you start?
You basically have to re-invent your language. You have to start with your own definitions and language. And you have to get everyone involved so they understand it the same way. This is a pure business issue. It has nothing to do with IT. After that you can establish your data structures.
How do you add in the necessary flexibility, such as when you acquire a new company?
You don’t have to change it as much when you buy a company as when your own company changes direction a little bit. When companies switch from cost-cutting to growth, all of your indicators for cost-cutting are no longer relevant for growth. You need to build intelligence into your glossary so you have different missions or directions. After that it depends on where your data is coming from. You might find different companies have completely different data schemes.
Is all the data necessary?
No, and some companies might do better if they throw away some data they don’t need. Sometimes the data they have is incompatible and it can make people misunderstand connections between different items. Most of the data is very valuable, but there is an art to defining what you should trust and what you should ignore.
Do people understand what they’re getting in the data, even if it’s done right?
In my experience they understand half of the key performance indicators. The business guys need to really understand the business from a number of perspectives. You may need a data translator who can translate the data into business intelligence. There are very, very few people who can do this.
Ed Sperling is the editor of several technology trade publications and has covered technology for more than 20 years. Contact him at firstname.lastname@example.org.