Choosing Metrics

Choosing Metrics (Blog)

Every operation monitors metrics. Choosing which metrics to review can be remarkably difficult. This post provides a few principles for selecting metrics to monitor.

Inputs and outputs

Most teams start by monitoring output metrics. These are the high-impact metrics you’re likely to see reported to top executives or external investors. Common examples: user satisfaction; cost of goods sold (COGS); revenue; contacts per unit sold. These are important gauges to keep an eye on. However, they are not good enough on their own to run an operation. Why? Because they are outputs. To extend the gauge analogy, they might tell you that a plane is losing elevation and headed for a fiery crash, but they will provide no insight into why that is happening or what you should do to adjust.

To run and improve an operation, you need to monitor and control input metrics. What distinguishes input vs. output in metrics? An input metric can be directly impacted by action levers your organization controls. An output metric, on the other hand, is influenced by multiple input metrics (and possibly other external factors) and the link between cause and effect is more murky. If you find yourself celebrating a metric improvement that was caused by something external to your org, or trying to explain a goal miss that was entirely out of your control, you’re probably dealing with output metrics. As an operator, you want to deal with input metrics whenever possible.1

Identifying inputs

Choosing input metrics is necessary work if you want to control your direction. But it’s hard work. Output metrics present themselves to you. They scream in your face. Revenue, customer satisfaction, contacts—you can’t ignore them if you tried. How do you zero in on the inputs you can control?

You can start by decomposing recent trends in an output metric of interest. Say that customer service contacts received per sold unit has been creeping upwards. As mentioned in Effective Operational Reviews, you should ask why that is happening over and over until you hit a reason you can act on. Here’s the key: that actionable reason is an input metric.

Let’s keep running with the customer service contacts example above. Perhaps you find that furniture is increasing as a percentage of total sales, and that damage contacts are rising as a rate within the furniture category. Is the increase in furniture as a percentage of total sales actionable? Maybe; it depends on your overall strategy. You have to know your business in order to convert principles to rules.

But back to that second cause in our example: increase in damage contacts within furniture. That’s something you want to understand. Let’s say you dig another level deeper and find that the increase is driven by one fulfillment origin. That’s starting to sound like a potential input metric: all else being equal2, you’ve identified a metric that you control or influence, and that can in turn improve the output you want to see move: damage contact rate within a single product category by fulfillment origin.

Filtering and diving deep

As you get into the hang of identifying input metrics, you may find that you have more than you can keep a close eye on. This becomes more true as you move up the management chain and there are increasing numbers of potential input metrics. Which should you monitor?

Three principles to keep in mind. First, aim for leverage. Input metrics that have a large impact on your organization’s priorities need to be at the top of your list. Second, this is a moving target. At whatever level you are monitoring, you need a regular cadence to review the list of metrics you and your team are monitoring. As conditions and priorities change, your list of input metrics also needs to change. And third, make sure you regularly dive deep to the ground truth. Not all the time, but often enough that you remain connected and you don’t lose the ability to find out what’s really happening. You will need to figure out the right cadence. This is more important the more senior you become, and the more separated from the front lines you are. Don’t limit yourself to the reports that are being brought to your desk. See for yourself.

Future topics

There’s plenty more to discuss about metrics. How should we set goals? How do we think about normal variation—what’s real and what’s an artifact? We’ll return to this topic on a future date. If you’d like to hear about this and other similar topics, subscribe for notifications.


1 Another set of terms for "input" and "output" metrics is "leading" and "lagging" metrics or indicators. In this context these sets of terms are synonymous.
2 “All else being equal” is important, and will require additional digging in the real-world version of this simplified example. For example, product mix between fulfillment centers or distance travelled to the customer could vary significantly between origins. Keep going until you get to the truly controllable inputs!