# Vanity metrics vs real clicks: which numbers deserve your budget? > How to separate advertising numbers that predict revenue from numbers that flatter dashboards in 2026 — impressions, reach, opens, clicks and visits, sorted. Published 2026-09-01 · 7 min read · vanity-metrics, advertising, measurement Canonical: https://bidsurvivor.space/blog/vanity-metrics-vs-real-clicks --- Every advertising dashboard leads with its biggest number, and the biggest number is almost never the one that matters. Vanity metrics are the ones that flatter the report without predicting revenue; here's the sorting rule that separates them from the real ones, applied to the metrics you'll actually be shown. (Disclosure: published by [BidSurvivor](https://bidsurvivor.space), whose own two metrics get sorted below by the same rule.) ## Where the phrase came from Eric Ries put the term into circulation in *The Lean Startup*, where a vanity metric is one that goes up and to the right without telling you what caused it or what to do next, and an actionable metric is one tied to a specific change you made ([The Lean Startup principles](https://theleanstartup.com/principles)). His example was cumulative totals: total signups can only ever grow, so a chart of them looks like success whatever is actually happening. The sorting rule below is a way of applying his distinction to advertising reports specifically, where the vanity numbers have names like reach and impressions and the actionable ones are further down the page in a smaller font. ## The sorting rule A metric is real to the degree that a stranger paid a *cost* to produce it. Costs, ascending: a glance costs nothing, a click costs a decision, a visit costs attention, a signup costs identity, a payment costs money. Everything on a dashboard sits somewhere on that ladder, and platforms habitually report from the bottom of it. ## The usual suspects, sorted **Impressions and reach** — the bottom rung. An impression is your ad *existing* near a human. Necessary, predictive of nothing by itself, and the number platforms lead with precisely because it's always large. The Nielsen Norman Group's eye-tracking work on [banner blindness](https://www.nngroup.com/articles/banner-blindness-original-eyetracking/) has shown for nearly three decades that people learn not to look at the places ads live, so an impression is not even a guarantee that anyone glanced. **Engagement** (likes, shares) — a half-rung up. Cheap social gestures that correlate with attention, not intent. Campaigns "optimised for engagement" reliably harvest the people most willing to tap things. **Email opens** — corrupted evidence since privacy proxies began auto-firing open pixels; treat as directional at best. **Clicks** — the first honest rung. A click is a decision to leave somewhere and arrive at you. Still gameable (bots, accidental taps), which is why the next rung matters. Juniper Research puts advertising spend lost to fraud at **$100.2 billion for 2026** ([Juniper Research](https://www.juniperresearch.com/press/digital-advertising-spend-lost-to-fraud-68-billion/)), and most of that is manufactured impressions and clicks — the two cheapest rungs to fake. **Visits that behave** — arrived *and* scrolled, stayed, or acted. The first metric that predicts revenue. The gap between clicks and behaving-visits is also your bot detector — [cheap traffic that clicks but never scrolls isn't traffic](/blog/cheapest-way-to-get-website-traffic). **Signups and payments** — the top. Everything above exists to be divided into these. ## Why email opens broke Opens deserve their own paragraph because they were a decent metric once and a lot of sponsorship pricing still leans on them. An open is counted when a tiny tracking image loads. Since 2021, Apple's Mail Privacy Protection has pre-loaded those images on Apple's servers whether or not the person ever opened the message, so a letter with a largely Apple-reading audience reports opens that nobody performed. The open rate did not fall when this happened; it rose, which is the tell. A metric that improved because measurement got worse is a vanity metric by construction. Ask newsletter sellers for clicks per issue instead — that number still means what it says. ## A worked report, sorted An invented report, round numbers, for a $200 campaign: 120,000 impressions, 2,400 engagements, 380 clicks, 190 visits that lasted more than ten seconds, 9 signups, 1 payment. Read from the top, it looks like a campaign that reached a small city. Sorted by the rule it reads differently. The 120,000 cost strangers nothing. The 380 clicks cost a decision each, at about 53 cents. Half of those clicks — 190 — became visits that behaved, which says the other half were misclicks, bots or people who bounced on the headline; that ratio is the report's most useful fact and it appears nowhere in the platform's summary. The 9 signups cost roughly $22 each, and the single payment is one, which proves nothing yet. Which of these numbers deserves next month's budget? Only the ones from the click rung up, and only once the sample is large enough that 9 is not noise. The top two lines of the report are what the platform wants you to remember; the bottom four are what you are paying for. ## How we sort our own two numbers Our board publishes exactly two metrics per brand, [publicly](https://bidsurvivor.space): card opens and click-throughs to the brand's site. By our own rule, the first is a curiosity metric — a visitor tapping a tile costs one decision, low on the ladder — and the second is the real one, a stranger leaving our page for yours. We publish both, separately, because averaging them would manufacture a vanity metric out of two honest ones. Small numbers, but sorted correctly, and the same two numbers for every brand on [the board](/blog/how-bidsurvivor-works). ## The attribution trap There is a subtler way a real metric becomes a vanity one: measuring it with a tool that quietly stopped counting. Attribution windows do this. Google Ads now ignores offline conversions uploaded more than seven days after the click, which means a business with a two-week sales cycle sees its best customers vanish from the report — the campaign looks worse than it was, and a naive reader cuts it ([the seven-day upload rule, explained](/blog/offline-conversion-upload-rule-explained)). For a small test, the fix is not better attribution but less of it: measure the rung you can actually count, on the day, and stop pretending to know the rungs above. [Fast tests need honest numbers](/blog/fast-tests-need-honest-numbers), and an honest small number beats a modelled large one. ## The counter-argument: sometimes the glance is the point Brand advertising has a real reply to all this. If the job is to be remembered in six months, the glance is the product, and Kantar's work on long-term effects finds that campaigns judged only on immediate response under-invest in exactly the exposure that builds later sales ([Kantar](https://www.kantar.com/inspiration/brands/short-term-growth-long-term-gains)). Fair. But that argument belongs to advertisers with the budget to measure recall by survey, not to a small advertiser reading a weekly dashboard. If you cannot measure the memory, the impression count is not a proxy for it; it is a number standing where the measurement should be. ## The practical discipline Ask of any report: which of these numbers cost a stranger something? Optimise campaigns toward the highest rung you can measure with a decent sample — and mind the sample: one day of small numbers proves nothing on any rung, which is a sizing question, not a metrics one ([what one day can prove](/blog/one-day-ad-campaign)). And when a platform's pitch leads with reach, read it as a confession about the rungs above. If it helps, the [backlink checker](/tools/backlinks) applies the same test to links: exists, versus ever sent anybody. It is free and needs no account.