“Be approximately right rather than exactly wrong.” – John W. Turkey
“Approximately” is a wonderful term – and an essential concept for everyone working with analytics.
First, I want to state a presumption. Data is valuable. Every bit of information your organization captures, copies, or keeps has an intrinsic value to the organization. But, that value is only realized when people leverage the information for some purpose that creates value.
I’ve worked with many companies; and, quite naturally, those tasked with providing data for analytics have a natural drive to make their data perfect – to remove any approximation. And while this aspiration is great, the reality typically falls far short. Data quality is more like gardening than construction. No matter how often you have perfected your lawn, in a few weeks, you’ll have to mow it again.
Too many businesses, when faced with the challenges of data quality, chose to hold back less-than-perfect data from business users. This guarantees no value will be gained.
We have to recognize that all data is valuable and make as much of it available to our users as possible – as quickly as possible. If that information is approximately correct, state how close. Decision makers work from approximates all the time. Most of us have information that, at best, suggests trends or directions that our business is progressing, even if the individual values aren’t 100 percent right.
As we bring more information to more people, as we recognize that value is some function of the amount of information and the number of people employing that information, we need to accept that we can’t afford to remove all ambiguity and approximation.
On the bonus side, analytics is always about people, and people are really good at using and dealing with approximates. They’re really good at finding and leveraging trends and filling in the gaps to discover patterns. And they’re good at synthesizing divergent pieces of information to create new value.
We can cope with approximates as long as we can get to the data.