Where people quit your funnel tells you what's broken in the ad.

Say you run a $30-a-day campaign for three weeks. You get 40,000 impressions, 300 clicks, and four sales. Something in there is broken. But 'the ad isn't working' is not a diagnosis, it's a shrug. The impression-to-click number and the click-to-sale number are telling you two different stories, and they call for two different fixes.

Most owners see low conversions and rewrite the whole ad: new photo, new headline, new offer. Often that's wrong. If clicks are strong and sales are weak, the ad is doing its job. The problem lives on the landing page, or in the gap between what the ad promised and what people found when they got there.

The four places people leave, and what each one means.

Think of the funnel as four gates: the ad gets shown (impression), someone notices and clicks (click), they land somewhere and keep reading (landing), they decide to buy (conversion). A person can drop out at any gate, and each drop-out point implicates a different part of the creative.

Low impressions relative to spend is usually a targeting or budget problem, not a creative one: the ad isn't reaching enough of the right people to judge anything yet. That's a platform-delivery issue, separate from what this piece is about.

High impressions, low clicks means people are seeing the ad and scrolling past it. The visual isn't earning a second of attention, or the headline isn't making a clear enough promise to anyone in particular. This is a hook problem: something in the first half-second doesn't immediately answer why they should care.

High clicks, low landing-page engagement (people click, then leave within a few seconds instead of scrolling or reading further) usually means the ad and the page don't match. The ad said one thing, the page said something else, and the mismatch reads as a bait-and-switch even when nobody intended it that way.

High engagement, low conversion means people are interested but not convinced. This is usually a proof problem: no price, no specifics, no reason to act now rather than later. Spotting that gap in your own copy is hard when you already know what you meant to say, and Knowing shows whether the weak spot is the first glance, the read or the click.

Why an ad with no clicks might still be working.

Here's where most owners get the diagnosis backward. An ad that gets seen but not clicked looks like a failure if you're only counting clicks and sales that happened right after. But a 2018 study by Vibhanshu Abhishek, Peter Fader, and Kartik Hosanagar modeled individual consumer behavior across the full path to purchase, instead of crediting only the last thing someone clicked before buying. They found that display ads, the kind most small business owners run on Facebook or Instagram, mostly do their work early: moving someone from not paying attention at all into the first stage of noticing a brand exists. The click, or the sale, often comes later, from a different touchpoint entirely.

Their finding pushes back on a common assumption: that a low-click ad is a wasted ad. As the researchers put it, "Contrary to the common belief that display ads are not useful, our results show that display ads affect early stages of the conversion process." If you judge that ad purely on click-through, you'll conclude it failed. The researchers' model suggests it may have been doing exactly what an early-stage ad is supposed to do: getting you noticed, not getting you clicked.

The same paper offers a useful caution in the other direction, too: "only a fraction of online conversions are driven by online ads" at all. Some of what looks like an ad problem is actually a demand problem, or a word-of-mouth problem, or a coincidence of timing. Not every sale, and not every missed sale, belongs to the last ad someone saw.

An ad that gets seen but not clicked isn't automatically broken. It may be doing early work that the click number can't see.

Different stages need genuinely different ads, not just different targeting.

If different gates have different problems, the creative sitting at each gate should be different too, not just retargeted at a narrower audience with the same message. A 2019 paper by Yichao Zhou, Shaunak Mishra, and Jelena Gligorijević built a model that reads a person's activity trail, what they've searched, visited, and clicked, to infer which stage of the decision they're actually in: unaware, aware, interested, considering, or ready to act. In plain terms, it looks at what someone searched for, visited, and clicked to figure out whether they're just learning the product category exists or already close to buying.

What makes this more than an academic exercise is what happened when it went live. The researchers report that campaigns on Yahoo Gemini using "funnel-stage-specific targeting and creative variants" produced measurable lifts over campaigns that didn't differentiate. Their attention-based model also improved conversion prediction accuracy over simpler models, with the authors noting gains "of 0.9% on a public dataset... and up to 3.6% on three proprietary datasets." On a modest monthly ad budget, a few percentage points of extra conversions adds up to real sales over a year. And the mechanism behind the number matters as much as the number itself: an ad tuned to where someone actually stands in their decision beats an ad written for an average person who doesn't exist.

Most small business owners don't have the volume of data these researchers had, and they don't need a machine-learning model to act on the underlying idea. If your data shows people clicking but not buying, the fix isn't a punchier headline, it's more proof on the page they land on: a clearer price, a specific guarantee, a photo of the actual product instead of a stock image. If people aren't clicking at all, the fix isn't a better offer buried in paragraph three, it's a sharper first line and a clearer subject in the image, because nobody gets far enough to read the offer.

What this changes about how you read a report.

The instinct after a disappointing campaign is to throw the whole ad out and start over. That instinct is expensive, because it treats every drop-off as the same failure and every fix as a full rewrite. It's also frequently wrong: a rewrite aimed at the wrong gate can make a working part of the ad worse while leaving the actual problem untouched.

The more useful habit is to look at the numbers in sequence, find the specific gate where the falloff is steepest, and ask what that gate is asking of the creative that it isn't getting. An ad that gets clicks but few sales doesn't need a new photo. It needs a landing page that keeps the promise the photo already made.