Vanity metrics and the ones that predict revenue

The reason this matters is not aesthetic. Vanity metrics are how underperforming marketing survives a review meeting. A vendor who cannot show revenue can always show reach, and reach always went up, because reach is a function of effort rather than result.

I sat through years of these in restaurants under a different name. Covers were flat but the promotion reached forty thousand people. Everyone nodded. Nobody asked what happened between the reach and the door.

What actually makes a metric a vanity metric?

Three properties, and it usually has all three.

It only moves in one direction. Followers, total reviews, total pages published, cumulative impressions. Anything cumulative is a vanity metric by construction, because it cannot get worse. A number that cannot go down cannot warn you about anything.

It has no denominator. Two thousand sessions is meaningless without knowing how many converted. Fifty thousand impressions is meaningless without clicks. Whenever a report gives you a numerator alone, the denominator is usually the unflattering half.

No decision changes when it moves. This is the strongest test. If organic sessions rose fifteen percent this month, what do you do differently on Monday? If the honest answer is nothing, the metric is entertainment.

Sessions are the interesting case, because they are not useless. Sessions matter as a leading indicator when your conversion rate is stable and known. Without that second number, traffic growth tells you nothing about the business.

Why is a metric that goes up sometimes evidence that things got worse?

Because the number you are watching is standing in for the thing you actually want, and a stand-in can move in the opposite direction from what it represents. This is the single most valuable idea in measurement and it has a clean experimental proof behind it.

Microsoft tested a redesigned Office Online page and measured clicks on revenue-generating links as a proxy for purchases. The redesign produced a 64% reduction in clicks per user. By the metric on the dashboard, a disaster. The explanation, in the words of Kohavi, Deng, Longbotham and Xu writing in the KDD 2014 proceedings: "The Treatment page, which showed the price of the product, attracted fewer clicks, but those users were better qualified and had a much higher conversion-rate." Publishing the price cut the surrogate metric by nearly two thirds and improved the business outcome.

The same paper contains the inverse case. Bing extended its results page to twenty algorithmic results in a controlled experiment on more than three million triggered users. Pagination fell 18%, which reads like an engagement win, and abandonment did not move at all. Revenue fell 1.8% and the change did not ship. Truncating the page to four results, over more than eight million users, also left abandonment unmoved and shipped because the page got faster. Length itself was not what mattered. What the length cost was.

Two conclusions follow. Any metric that sits between the customer and the money can move against the money. And a dashboard full of surrogates will eventually recommend the wrong decision with complete confidence. Note the domain gap while you use it: this is search engine and large retailer data from experiments involving millions of users, not local business data. What transfers is the logic, not the figures.

Why are impressions worse than they used to be?

Because impressions and clicks have come apart, measurably, in the last two years, which turns a formerly weak metric into an actively misleading one.

SparkToro's panel work with Similarweb on US Google searches from January to April 2026 found 68.01% of searches ending without any click, and reports that per 1,000 US searches only 276 clicks now reach the open web against 374 in 2024, a 26% reduction in two years. That is panel-based vendor research and Google has disputed the methodology, so use the direction rather than the decimal. The direction is not in dispute.

Here is what that does to a report, with the arithmetic.

Last year. 41,000 impressions produced 779 clicks, which is a click-through rate of 1.9%derived.

This year's headline. Impressions doubled to 82,000. At the old rate that would be 1,558 clicks (derived), and the slide says the work is compounding.

Now apply the measured decay. Ahrefs compared 150,000 AI Overview keywords against 150,000 matched non-AIO informational keywords using aggregated Search Console data and found a 34.5% decrease in clicks, revised in a later cut to 58%. SparkToro's panel puts the drop nearer 60% when an AI Overview is present. All vendor research, all time-dependent, and the honest way to use them is as a range.

At the friendly end of that range, a 34.5% lower rate on 82,000 impressions gives 1,020 clicks (derived). Growth, but a third of what the doubled impressions implied.

At the harsh end, a 58% lower rate gives 654 clicks (derived). That is 125 fewer clicks than last year (derived), a decline of 16%derived, on double the impressions.

Same report, same true state of the world, and the impressions tile says the year went twice as well while the business got fewer visitors. That is not a hypothetical failure mode. It is the most common reporting error of 2026.

There is a structural version of the same trap arriving now. Google's Search Console generative-AI performance reports, live since June 2026, expose impressions, pages, countries, devices and dates, and explicitly do not expose clicks, click-through rate, position or the user's prompt. A tile built from that data is a numerator with no available denominator. It is worth watching for direction. It is a vanity metric by construction, and it should be labelled as one on your own dashboard.

Which metrics actually predict revenue for a local business?

The predictive ones sit close to a human decision, and there are fewer of them than you would expect.

Qualified inquiries per week. Not form fills. Inquiries from someone who could plausibly buy. If you strip out spam and wrong-market submissions, this number is the earliest honest signal you have.

Connected calls and their outcomes. For most service businesses the phone carries the highest intent traffic in the business, and the size of the prize is documented. ServiceTitan's platform data from June 2022 puts the typical home services shop at a 42% call booking rate and calculates that every 5% improvement in booking rate is worth roughly $100,000 in additional revenue for a shop with five to fourteen technicians, achievable with less than one additional call booked per weekday. ServiceTitan sells field service software, so label it vendor research; the figures are real platform data. One more booked call a day is a bigger number than most marketing campaigns produce.

Conversion rate by page, on your money pages. A change here moves revenue without any additional traffic, which is the cheapest growth available to a small business.

Repeat rate and time between visits. A restaurant with falling repeat frequency is in trouble months before the revenue line shows it. Most owners find out from the revenue line.

Cost per qualified inquiry by channel. Not cost per click, not cost per lead where lead means form submit. The qualified version is the only one that survives contact with your sales calendar.

Search Console gives you a genuinely predictive pair that most owners never look at: impressions and clicks together for the queries that describe buying intent, rather than for your own brand name. The free tool most businesses ignore is the single best source for that distinction, and it is the one place impressions become useful, because the click column sits right beside them.

How do I tell the difference in my own reporting?

Run each metric through one question. If this number doubled and nothing else changed, would I make more money?

Followers doubling changes nothing. Booked appointments doubling changes everything. Sessions doubling might change something, depending on a conversion rate you may not be tracking.

Then add the surrogate check from earlier. For every metric you keep, name the outcome it stands in for, and ask whether the two could move apart. If they can, the metric needs its outcome reported next to it or not at all.

Then check whether you can attribute it. A metric you cannot trace to a source is a metric you cannot act on, and the tracing is mostly mechanical. Consistent campaign tagging turns a pile of undifferentiated traffic into a table you can make budget decisions from.

The last check is durability. Half the metrics on small business reports break silently during a website rebuild and keep reporting a number that no longer reflects reality. A conversion tracking setup that survives a redesign takes an afternoon to build and saves you a quarter of misreading.

Why do vanity metrics keep coming back?

Because they are easier to produce and more pleasant to receive.

An agency reporting qualified inquiries has to sit with a bad month. An agency reporting impressions never has a bad month. Given the choice, and the client rarely forces one, the second report gets sent.

There is also a version of this that has nothing to do with vendors. Owners choose vanity metrics for themselves when the real number is uncomfortable. Watching followers grow feels like progress in a slow quarter.

Restaurant operators know this pattern from the kitchen side. It is why tracking waste without running a full inventory program works: you pick the small number that hurts to look at, and you look at it weekly. The uncomfortable number is almost always the useful one.

What to do this week

Take your most recent marketing report. Draw two columns. Put every number in one of them, using the doubling test.

If the left column is longer, you are being reported to rather than informed.

For every impressions or reach figure that survives, write its click or action count beside it. If that second number is not available, the tile goes in the vanity column regardless of how good it looks.

Pick one metric from the right column that you are not currently tracking well and fix that one. For most local businesses it is either connected calls or qualified inquiries.

Then write down today's value for every metric in the right column, with the date. Capturing a baseline before you change anything is what turns next quarter's numbers into evidence instead of a story.

Questions about which of your numbers matter?

Email me at eric@seod.com with the one metric you have felt best about this year. Just the metric and roughly what it did. I will tell you honestly whether it predicts revenue in a business like yours or whether it only measures effort, and which number I would watch instead.

I answer these myself and it takes a few minutes. If your metric is a good one, you will get a short email saying so.

The rest of the analytics library covers the mechanics.