I was going to write this article, but a few days ago Ron Ashkenas at Harvard Business Press in effect wrote it for me.

He nails four issues around believing that data = truth:

1. Are we asking the right questions?
2. Does our data tell a story?
3. Does our data help us look ahead rather than behind?
4. Do we have a good mix of qualitative and quantitative data?

The Three Data Points That Matter

At the heart of any organization, there are really three data points that determine success:

1. Profitability (present dollars — or yen, Euros, etc.)
2. Intent to repurchase (future dollars)
3. Key-employee commitment (the future of the company)

I find it odd that in the various corporate data “dumps” I’ve seen, #2 and #3 are often absent. In fact, I don’t recall the last time I saw repurchase intent included as a key metric; usually, it’s the misleading “customer satisfaction” that serves as a (poor) substitute metric.

Even #1 is poorly represented in many corporate metrics, especially within departments. Sure, the company has to publish a P&L (profit and loss statement) if it’s a publicly traded corporation, but departments within it do not. Yet how can employees at all levels make the right decisions without this information?

Internal departments certainly can build data around #1 and #3, and most have a version of #2 as well. Even if you’re a sole-source internal supplier such as IT, it’s important to understand your long-term support within the businesses you serve. (If you’re afraid to ask, you already know the answer.)

If you don’t have all three of these data points, you’re not a data driven organization. Period.

Now, around each of these, add Ron Ashkenas’s four questions. Answer them sensibly and honestly. Now you’re in business.

Why Qualitative (Non-Numerical) Data Matters

Here’s a hypothetical. You are a manufacturer with great quality control procedures. You hear from a very few customers that they have concerns about the quality of one of your parts. However, there are always complainers, and your own quality checks show no untoward problems.

The data says you’re good, right? Sure, maybe you want to rerun a few quality checks, but when they don’t turn up any new issues, you don’t change a well run business.

Now let’s put a couple of specifics in there. Let’s call the manufacturer Toyota, and the part in question is the accelerator system on something called, say, a Prius.

Still all good?

As it looks today — from an outside perspective, based on what’s been in the news — their quite stringent quality control checks turned up no issues. But of course, if a part were mis-designed and built exactly to that spec, it wouldn’t show any issues in manufacturing. Or what if it does turn out to be the floor mats — user error, right?

Quantitatively, right.

Qualitatively, dead wrong — and the pun, unfortunately, applies. Listening to the qualitative data — people are dying in runaway Toyotas — would have, I believe, brought everyone running to focus on figuring out what kind of wild-a** possibilities could cause this to happen. And then they could think about how the company might respond even when unable to figure out what, if anything, was going wrong.

It takes math-smarts to process quantitative data. It takes real-world smarts to process qualitative data.

Math’s easy. Real-world stuff is hard. But you’ve got to do the hard stuff, perhaps even more so than the easy stuff.

You can’t be data-driven by looking only at quantitative data. Or easily collected data. Or incorrect data.