Vanity Metric

Pulling a number out of a database and putting it on a screen got cheap, and reports and dashboards filled up. Some of those numbers are up every time anyone looks. Total signups, total documents ever created, visits to the marketing site. Once a count has been added to a running total it never comes back out, so these numbers keep climbing in the months when the product is getting worse.

Pagedeck's monthly review opened with exactly that. Total signups of 120,000, two million documents created, 80,000 visits to the marketing site, spread across three slides, every one of them larger than last month. Twenty minutes went into reading them, and the meeting ended without a single decision about what to build. Over the same period the number of paying teams had been flat for two months, and that number was on slide four.

The numbers weren't wrong. 120,000 was accurate. Nothing followed from knowing it. So a name was needed for a number that is correct and still useless — a separate category from a number that is simply miscounted.

A vanity metric isn't a type of number, it's a state a number falls into when two things line up: it moves in essentially one direction, and nothing the team does next depends on the value it lands on.

1. Running totals don't go down. Total signups rises by ten on a day when 100 people leave and 110 arrive. A line that only climbs feels like progress, and it cannot tell you whether the product got worse.
2. There's no basis for comparison attached. Whether 80,000 visits is a lot depends on last month's figure, or on how many of those visitors signed up. An absolute count with no denominator lets every reader reach a different conclusion: one person sees 80,000 and reads growth, another remembers 95,000 in the spring and reads decline, and both are working from the same slide.
3. Nothing is downstream of it. One question settles this: if this number halved next month, what would you do? No answer means it's a vanity metric. Total signups can't halve from 120,000 to 60,000, so the question doesn't even parse.

This is why the same metric is a vanity metric at one company and not at another. For a publisher selling ad space, page views are close to a synonym for revenue and they drive real decisions. At Pagedeck, visits to the marketing site change nothing by themselves, while visits-to-signup rate decides whether the headline gets rewritten. The underlying thing being measured didn't change; it was reshaped into something a decision can hang on.

The term gets used while choosing metrics, at the moment someone is deciding what to take off the screen. Four places collect them.

1. The first slide of a weekly or monthly report. The front slot invites whatever number looks biggest, and running totals win that contest.
2. The default dashboard. Visits and sessions arrive free with the analytics tool and nobody ever deletes them.
3. Material written to show scale to outsiders — fundraising decks, press announcements, the numbers strip on a marketing page. The goal there is persuasion rather than explanation, so a cumulative figure is a reasonable choice. The trouble starts when that same slide walks into an internal meeting as evidence, because it was built to impress rather than to inform a decision.
4. Quarterly goals. Put a vanity metric in a target and the team moves toward raising it. The way a metric breaks the moment it becomes a target is Goodhart's law; with a vanity metric there's nothing left to break, because the effort was wasted from the start.

What comes off the screen has to be replaced. The usual arrangement puts one north star metric in the middle to point the team in a single direction, a few leading indicators on one side that move early enough to act on, a few lagging indicators on the other that confirm what already happened, and a counter metric underneath watching whatever the target might damage. Rebuilding a metric set in practice is rarely about finding more numbers. It usually ends with half the existing ones removed.

"Total signups is a vanity metric, so it's off slide one and weekly active accounts is there instead."

A number nothing depends on lost the front slot to one that actually moves.

"If this halved, what would we do about it?"

The test for whether a metric reaches a decision — no answer means it comes off the screen.

"Signups are up and paid conversion has been flat for three months."

Pointing out that the number in front is no evidence the business improved.

Reworking Pagedeck's monthly report is what Mia did with this. She attached the same question to each of the three cumulative numbers on the opening slide: if this halved, what would we change next week? None of the three produced an answer. All three came off, replaced by weekly active accounts and the share of new teams that create three or more documents in their first week. At the next review, reading the numbers took five minutes and the rest of the hour went to why the second one had dropped.

The most common mistake is memorising which metrics are vanity metrics. The same examples appear in every explanation, so page views and signup counts get filed as bad numbers on principle. In practice the slot the number sits in decides. Page views are raw material for revenue at a company selling ad inventory, and signup counts are exactly what you watch for three days after changing the signup screen. Treating it as a list throws away usable numbers while keeping numbers that change nothing in your context, on the grounds that they weren't on the list. The test travels between companies; the verdict on any particular metric does not.

The second misreading is that vanity metrics shouldn't be measured at all. They were introduced as the bad ones, so deleting them looks correct. But total signups is still what you need to size server costs or describe scale to an investor. What comes off is not the number, it's the front slot the number was occupying. Keep collecting it and leave it out of the screens people make decisions on. Delete the history instead and you'll be rebuilding the series six months from now with nothing to compare against.