Metrics Lie. Here Is What Yours Is Hiding.

A metric that goes up is not evidence that the underlying thing improved. That distinction sounds obvious. Most commerce dashboards are built as if it does not exist. MIT Technology Review's recent examination of measurement culture makes a calibrated argument: the longer you track something in fine detail, the more clearly you can see what the number was never actually capturing. For executives who have spent the last three years optimizing toward return on ad spend, contribution margin, and customer acquisition cost, that is probably an uncomfortable inference to sit with.

The Average Brand Tracks Performance. The Top 10% Track Metric Validity.

Here is roughly where the separation happens. Average operators inherit a dashboard from a previous team, a vendor setup, or a platform default. They optimize toward whatever those numbers say. Top-decile brands do something different. They periodically ask whether the metric itself is measuring the right thing, whether the measurement method introduced distortion, and whether hitting the target has started to corrupt the behavior it was supposed to encourage. That last question is the one most leadership teams skip entirely.

The pattern has a name in academic circles: Goodhart's Law. When a measure becomes a target, it ceases to be a good measure. You have probably seen it play out without labeling it. A team optimizes email open rates and starts writing subject lines that trick the click. A media buyer chases ROAS and pulls budget from brand-building channels that do not attribute cleanly. A loyalty program gets tuned for repeat purchase frequency and starts training customers to wait for discounts. The metric moves in the right direction. The business does not.

What Best-in-Class Looks Like, Specifically

Best-in-class operators, roughly the top 5% by measurement discipline, run what some internal teams call metric audits. Not a rebrand of the quarterly business review. A structured process where someone is assigned to argue that each core KPI is the wrong KPI. The goal is not to tear down the measurement stack. The goal is to find where the proxy has drifted from the underlying truth it was supposed to represent. This happens more often than most dashboards will tell you, because dashboards are built to display numbers, not to question them.

The structural difference between average and best-in-class here is not budget. It is not headcount. It is a habit of institutional skepticism toward the numbers your own teams produce. That habit is harder to build than a new attribution model. It requires someone in the room with enough standing to say the metric is probably wrong, and enough specificity to explain why, without that observation being read as an attack on the team that built it.

Three Moves Worth Making Before Q3 Planning

First: map each core metric back to the business outcome it is supposed to represent, then write down at least one plausible way optimizing the metric could make the outcome worse. If you cannot write that sentence, you do not understand the metric well enough to trust it. Second: identify which of your top five KPIs were defined by a vendor or a platform rather than by your team. Vendor-defined metrics have a structural incentive problem. They tend to measure what the vendor can influence, not what your business needs. Third: pick one metric that has been moving in the right direction for more than six months and ask whether team behavior has changed in ways that explain the movement without the underlying thing actually improving. Promotion timing, creative testing pauses, audience exclusions added quietly. The answer is probably yes at least once.

Three Questions to Pressure-Test Your Measurement Stack

Could your single most-watched KPI be going up while the business it represents is going sideways? Name the specific mechanism, not just the theoretical possibility. When did someone last get paid, promoted, or praised for proving a metric was wrong rather than for hitting it? If the honest answer is never, that is a structural incentive problem worth addressing before it compounds. And finally: if a competitor were using a more accurate version of the same metric, what decision would they make in the next 90 days that your current dashboard would prevent you from seeing? One uncertainty worth naming: it is genuinely hard to know how widespread rigorous metric auditing is among top e-commerce operators, because the brands doing it well tend not to publish the methodology. If credible benchmarking data on measurement practice frequency surfaces, it would change the calibration here considerably.