the doctor On Dashboards: They’re Dangerous and Dysfunctional

A colleague of mine recently remarked that the whole dashboard concept is stupid beyond belief and just a lot of malarkey and I have to agree. “Dashboards” as they exist today are completely useless, and, in general, the concept is mostly useless.

As my enlightened colleague would say, a single metric on a dashboard is like saying one idiot light is better than the combination of oil pressure, water temperature, and battery voltage gauges. Sorry, it’s not. And, guess what, you can’t see the relationship between oil pressure and water temperature, so you could have a “TEMP” light that’s going off because oil pressure is nonexistent and you’re scraping the metal off the cylinder walls! It’s useless.

The fundamental flaw is that today’s dashboards are marketed as a snapshot view of how well your organization is doing. A dashboard can not tell you how well you’re doing. To do that, it would have to know everything you’re not doing well, determine how much that is impacting your overall performance, and report that. However, the best it can do is capture the data it’s been programmed to capture, roll-up the metrics it’s been programmed to roll up, and do the built in calculations of efficiency based on those roll-ups.

Just because it reports 90% of spend “on contract” does not mean 90% of your spend is “on contract”. Maybe 10% of your total spend on a commodity under contract has been misclassified under the wrong commodity code, and all of this spend is off contract, meaning that only 82% of your spend is actually on contract. Just because it says on-time supplier delivery is 95%, does not mean that you’re doing a fantastic job of managing your suppliers and that only 5% of shipments are late. Maybe it’s only recording shipments late if they don’t arrive on the designated day and not taking into account the time of arrival – which could be a consistent 2 to 4 hours late. Since this could require a lot of overtime by your warehouse crews who arrive early with nothing to do for the first two hours, this costs you. And so on.

A dashboard can only provide an upper bound on how well you’re doing, and this is useless. Saying my efficiency is at most 98% when it is in fact 92% is useless and unactionable. I can say your efficiency is at most 100% and always be correct – and I don’t need an overpriced software hack to tell you that!

The most a well designed “dashboard” could do, if the goal was reversed from trying to tell you how well you are doing, which it cannot do, to how poor you are doing, is give you a lower bound on how poor you are currently performing. Whereas “my efficiency is at most 96%” is not useful, “my inefficiency is at least 4%” is useful. That tells you that not only are you not performing at 100%, but that the system, even though it’s unable to identify all sources of inefficiency, has found 4% inefficiency that is immediately actionable. Whereas “at most 88% of spend is on contract” is not useful, “at least 12% of spend is off contract” is useful because it identifies some low hanging fruit that should be immediately tackled to improve spend compliance within your company. Of course, this is assuming that the process used to compile the data and calculate the metrics isn’t fundamentally flawed (which it very well could be in some of the dashboards out there).

Of course, this is still nowhere as useful as a good spend analysis or business intelligence tool that allows a seasoned analyst to construct some well formed cubes and drill around as she desires – as such an analyst would find anything the “dashboard” would find in about five minutes of drilling into the “spend cube” or “efficiency cube”. Furthermore, even if you covered an entire wall with dashboard displays, you still wouldn’t get close to all the perspectives you could come up with by drilling around a couple of well formed data sets.

Basically, all today’s dashboards do is present a pretty picture of a rather useless report. And I don’t want to hear any arguments that they’re “flexible” or “configurable” or “customizable” and that they can do whatever you want them to do – they can’t – they can only be “customized” or “configured” to the extent that they were built to be “customized” or “configured” by the development team – who probably had little understanding of your business, your data collection methodologies, your processes, and the information you really need to understand your business – and, in my experience, that’s usually not nearly as “configurable” or “customizable” as they would need to be to be useful even to the limited extent they could be if they were designed properly.

So next time someone tries to sell you a “dashboard”, thank them for the offer, and instead ask them about their analytics engine. That’s what you really need!