The dashboard that nobody looks at any more
Every dashboard I’ve ever built has a graveyard section. A tile that used to matter, that somebody asked for eighteen months ago, that still updates every night, and that nobody has clicked on since. If you run your own numbers, you probably have one too. Maybe it’s “total signups” from back when you tracked every account instead of active ones. Maybe it’s an old funnel step for a checkout flow you redesigned. Maybe it’s a churn calculation that used a definition of “active customer” you changed a year ago without ever renaming the metric.
Deprecating a metric just means formally retiring it: marking it as no longer maintained, telling people to stop relying on it, and deciding what happens to the history behind it. It sounds like housekeeping, and mostly it is. But doing it badly causes real problems, and doing it well takes more thought than “just delete the tile.”
Why metrics pile up in the first place
Nobody sits down and decides to build a cluttered dashboard. It happens one reasonable decision at a time.
You add a metric because a specific question comes up. Maybe you want to know if a new landing page is converting better than the old one, so you build a conversion metric scoped to that campaign. Six months later the campaign is over, but the metric is still running, still eating a query, still sitting in your weekly report.
Or your business model shifts. You started tracking monthly recurring revenue when everything was subscriptions. Then you added a one-time product line. The MRR number technically still works, but it’s now describing a shrinking slice of your business, and if you’re not careful it starts implying things that aren’t true, like your revenue holding steady when it’s actually your subscription revenue holding steady while a different revenue stream grows next to it unmeasured.
Or the definition underneath a metric changes and nobody updates the label. Cohort retention calculated on a 30-day window quietly becomes a 60-day window because someone adjusted the query for a different question, and now the number on the dashboard means something different than the name suggests.
None of these are mistakes exactly. They’re just what happens when a business runs long enough. The metric was built for a real question at a real point in time. The question just stopped being the question.
The actual risk of leaving a dead metric in place
The tempting response is to shrug it off. It’s just a tile, right, it’s not hurting anything.
It’s hurting something quietly. A metric that no longer reflects how the business actually works is worse than no metric at all, because it looks authoritative. If you glance at a churn number every Monday and it’s been calculated the same way for two years, you trust it by habit. If the underlying definition of “customer” changed six months ago and nobody flagged it, you’re now making decisions off a number that answers a slightly different question than the one you think you’re asking. That’s a data governance problem hiding inside a dashboard problem.
There’s also a compounding cost. Every metric on a dashboard is something a reader has to mentally sort through before they get to the ones that matter. If half your dashboard is dead weight, the live half gets less attention than it deserves, simply because it’s competing with noise. For a solo operator, that noise tax matters more, not less, because you don’t have a team to double check what you glossed over.
Signs a metric is ready to retire
A few honest questions tend to surface the dead ones:
Has anyone made a decision off this number in the last quarter? Not glanced at it, decided something because of it. If the answer is no, that’s a strong signal.
Does the definition still match reality? If your CLV calculation assumes a subscription model and you’ve since added a one-time product, the number is technically still computing, but it’s answering a question you no longer have.
Is it duplicated by something better? A lot of dead metrics aren’t wrong, they’re just outranked. If you built a simple conversion rate metric early on and later added a proper funnel breakdown that shows the same conversion plus where people actually drop off, the simple version usually isn’t adding anything any more.
Would losing it break an A/B test or attribution model you’re mid-way through? This is the one people skip, and it’s the one that causes actual damage. A metric can look unused on the surface while quietly feeding a test or a report that runs less often than weekly.
Why “just delete it” is the wrong move
Once you’ve decided a metric is dead, the instinct is to remove the tile and move on. Resist that for a beat.
The problem is that deleting a metric outright erases the history along with the clutter. If you ever need to look back, say a partner asks how a campaign performed two years ago, or you’re trying to understand a trend that only makes sense with the old number in the picture, you want that history retrievable even if it’s not on your active dashboard.
There’s also the attribution risk. If a metric feeds into any modeling you do, even loosely, like a rough attribution view of which channel drove a signup, deleting the underlying metric can silently break that model without an obvious error message. It just starts returning worse or blank results, and because nothing crashed, you might not notice for a while.
A simple process for retiring a metric
I don’t have a fancy system for this, but I do follow the same steps every time, because skipping one of them is exactly how you end up either keeping dead weight around forever or breaking something downstream.
Write down why it’s being retired. One sentence is enough: “Replaced by the funnel breakdown, redundant since March.” This matters more later than it seems like it should, because six months from now you or anyone else looking at the dashboard history won’t remember the reasoning, and a one-line note saves you from re-litigating the decision.
Check what else touches it. Before removing anything, grep through your reports, your A/B test configs, and anything doing attribution. A metric that looks isolated on a dashboard can still be a dependency somewhere less visible.
Archive, don’t delete. Move it out of the active view rather than dropping the underlying data. Storage is cheap. Rebuilding a historical series you deleted is not.
Set a sunset date, not a sunset moment. If other people rely on the metric, even just future-you checking an old report, give it a couple of weeks of visible “deprecated” labeling before it disappears from view entirely. That’s enough time for anyone who actually needs it to say so.
Note the replacement, if there is one. If you’re retiring an old conversion metric in favor of a funnel view, say so right on the old tile before it goes. It saves the next person, probably you, from wondering where the number went.
It’s not about having fewer metrics
The goal isn’t a minimalist dashboard for its own sake. Some businesses genuinely need a dozen live metrics tracked closely. The goal is that every number on your dashboard still means what its label says it means, and that you can trust it enough to actually use it when a decision comes up. A metric that’s technically still calculating but has stopped answering a real question isn’t neutral. It’s taking up space that could hold something you’d actually look at, and it’s sitting there looking trustworthy while quietly not being.
Retiring it isn’t a big event. It’s closer to weeding a garden than rebuilding one. Do it in small, regular passes and it never becomes a project. Let it pile up for a couple of years and you end up with a dashboard where half the numbers are load-bearing and half are just habit.
If you want more on how to keep the numbers you do track honest, from how churn is actually calculated to what a funnel can and can’t tell you, you can find the rest of what we’ve covered on Data Research Analysis Collection.
Get new guides and videos first — join the Telegram channel.