Measuring what matters (and ignoring what does not)

Every so often a client walks us through their analytics setup with genuine pride, and it is genuinely impressive - forty-tile dashboards, event tracking on everything that moves, weekly reports circulated to a distribution list of thirty. Then we ask the only question that matters ('which of these numbers, if it moved, would change what you do next week?') and the room goes quiet, because the honest answer is usually none of them. The company is data-rich and decision-poor, which is the standard failure mode of modern measurement, and it doesn't come from measuring too little.

Measurement has a cost curve that nobody talks about. Collecting a number is nearly free, which is how you end up with forty tiles, but every number on a dashboard also charges rent - in attention, in meetings spent explaining its wiggles, in the false comfort of feeling informed - and most numbers never pay it back. The discipline isn't measuring more, it's choosing, and choosing means being willing to ignore things on purpose.

One number that means the business is working

We start every measurement conversation by hunting for the product's defining metric - the single number that best captures value actually being delivered, not activity occurring nearby. For a booking platform it's completed stays, not sessions; for a B2B tool it's the workflow completed weekly, not logins; for a content business it's returning readers, not pageviews. The test is always the same: could this number go up while the business quietly dies? Pageviews can. Registered users famously can. Completed bookings can't, and that's what makes it a real metric rather than a vanity one.

The defining metric earns its seat by being an argument-settler. Roadmap debates, design reviews, the quarterly what-should-we-do-next conversation - all of them get shorter when there's one agreed number the work is supposed to move, because 'I like it' and 'the CEO wants it' both lose politely to 'it moved the metric'. Teams don't drown in data because they love data; they drown because nobody agreed which number is the boss.

Could the number go up while the business quietly dies? Then it's not the number.

Two or three that explain why it moved

A single headline number tells you that something happened, never why, so around it we instrument a small supporting cast - two or three drivers, chosen because the team can actually act on them. If completed bookings is the star, the cast might be search-to-detail rate, detail-to-checkout rate and checkout completion: each one owns a stretch of the journey, each one has a team that can move it, and together they turn 'bookings dipped' from a mystery into an address.

We also insist on a guardrail metric or two, because any number pursued hard enough starts lying. Chase activation and you'll build dark patterns; chase engagement and you'll build a casino; chase conversion alone and support tickets become someone else's quarter's problem. A guardrail (refund rate, support volume, unsubscribes - whatever the failure smell is) sits beside the target metric precisely so that winning can't be faked by cheating somewhere less visible.

Instrument the questions, not the pages

With the metrics chosen, instrumentation stops being a tagging exercise and becomes design work, which is why designers belong in the conversation. The moments that matter get named while the flows are being drawn - the fork where users commit, the step where the old journey bled, the action that defines 'this person got the value' - and events are written in the language of user intent (booking_completed, filter_applied) rather than the language of the DOM (button_click_47), so the data reads like a story instead of a stack trace.

The payoff shows up the first time someone asks a real question. 'Do people who use search convert better than people who browse' is answerable in minutes on intent-shaped events and answerable never on click soup, and the difference compounds with every question after that. Analytics you can interrogate becomes part of how the team thinks; analytics you can only report becomes a PDF nobody opens twice.

The quiet skill is ignoring things

None of this works without the second half of the title. Choosing three numbers means declining forty, and declining them out loud - agreeing as a team which metrics are deliberately unwatched, so their absence is a decision rather than an oversight. We'll happily keep the long tail collectable (storage is cheap, and next quarter's question might need it), but collectable and watched are different states, and the watching list is where the discipline lives.

There's a version of this article that ends with a framework and an acronym, and we've resisted it, because the whole point is smaller than that. Pick the number that means the business is working, surround it with the two or three that explain it, guard it against its own dark side, instrument the questions you'll actually ask - and then have the nerve to ignore everything else. Dashboards are cheap. Attention isn't, and the products that win tend to belong to teams that spent theirs on purpose.