Your Best-Performing Ad Is the One Quietly Costing You Money

When founders hand me an ad account, the first thing they usually show me is the winner. Best return on ad spend in the whole account, scaled a little more every month, and they want help pouring more into it. I have learned to treat that campaign as the first place I look for wasted money. The reason is not an opinion I formed. It has been sitting in the published research for more than a decade, and most ad dashboards are built in a way that keeps it out of view.

1. Why does your best ad look so good?

Your best ad looks good because it is taking credit for demand that the rest of your marketing already created.

Almost every ad platform reports on a last-touch basis, so whichever placement gets the final click before the sale gets the sale in the report. Think about who tends to get that final click. It is your retargeting campaign and your branded search, the ads that appear after someone has already seen you, decided they want you, and typed your name into Google.

Those placements are cheap to serve and they convert at high rates, so their reported numbers look extraordinary.

Economists put this exact question to the test at eBay. In a large-scale experiment published in Econometrica, they switched off eBay's branded search ads and watched what happened. According to the write-up from Chicago Booth, about 99.5 percent of the traffic those ads would have driven arrived at the site anyway through the free organic results. The researchers concluded that brand-keyword ads produced no measurable short-term benefit. eBay had been paying for clicks it was already getting for nothing.

2. What is your best ad actually being paid to do?

Most of your top-performing spend is paid to close buyers who were already on their way to you.

There is a question underneath every line in your ad account that the platform never asks. If this ad had not run, would this person have bought anyway? For a lot of your best-looking spend, the honest answer is yes. Someone who searches your brand name has already decided. Someone you retarget an hour after they left a full cart is coming back with or without the nudge.

The same eBay research made this concrete beyond brand terms. Once the experiment looked at the wider set of keywords, it found that most of the ad budget was landing on frequent shoppers whose buying did not change when the ads ran. New and infrequent shoppers did respond, but they were a small share of the spend, so the average return across the account came out negative. The account looked profitable in the dashboard and was losing money in reality.

3. How much is attribution overstating your winners?

Attribution is often overstating your winners by a wide margin, and you can put a number on it.

Here is a pattern that shows up constantly once you test it. A brand spends $100,000 a month on paid social, the platform reports a 4.0 return and around 2,000 conversions, and everyone treats that as settled fact. Then a proper conversion-lift test runs, and it turns out only about 800 of those conversions were actually caused by the ads. The rest were people who would have converted regardless. When a retargeting campaign shows a 10x return, that number is usually describing how good you are at picking people who were already going to buy.

I watched this play out up close when I ran marketing for a large telehealth company. One of the executives got fixed on click-through rate as the number that mattered, and he decided every ad we ran had to clear eight percent or it was a failure. I told him more than once that I could get him a twenty or thirty percent click-through rate whenever he wanted. All it took was a branded ad, the kind that shows up when someone is already searching for us and already knows they want us. Those clicks would have looked spectacular in the report and told us nothing about whether we had actually grown the business. A number you can manufacture on demand is worth exactly what it costs to manufacture it.

I want to be fair about this, because the story is not that retargeting is worthless. Controlled experiments using holdout designs have shown that retargeting can create genuinely incremental sales in the right conditions. The problem is that nothing in your platform report tells you which situation you are in. The same 10x campaign could be doing real work or simply collecting credit, and the dashboard presents both cases identically.

4. How do you find out what an ad is really worth?

You find out by measuring incrementality, the lift an ad actually causes.

Attribution answers which touchpoint came last. Incrementality answers what changed because the ad ran, and only the second question has anything to do with where your next dollar should go. You can run a holdout, where a defined audience or region is kept from seeing a channel while a comparable group still sees it, then compare the sales between them. A geo experiment does something similar at the market level, raising spend in some regions and cutting it in others. And a marketing mix model takes the widest view, reading months of spend and revenue across every channel at once to estimate how much each one actually moved the top line.

The big platforms have quietly built versions of this into their own tools, with Google's Conversion Lift and Meta's lift studies both designed to measure causal impact against a control group. The fact that they offer it at all should tell you how far their default attribution reports sit from the truth.

5. Why is everyone rebuilding measurement around marketing mix modeling?

Everyone is rebuilding around marketing mix modeling because the tracking that propped up attribution has fallen apart, and MMM never depended on it.

The user-level tracking that multi-touch attribution was built on has been hollowed out by privacy regulation, Apple's tracking limits, the long saga around third-party cookies, and the rise of AI search sitting between you and the buyer. Marketing mix modeling sidesteps all of it because it works from aggregate patterns and never needs to see an individual person, so privacy changes do not degrade it. It reads spend, sales, seasonality, and outside factors together, and it separates the channels that are genuinely driving results from the ones that only look correlated.

This is not a fringe view anymore. In a TransUnion survey reported by eMarketer, close to 47 percent of United States brand and agency marketers said they plan to invest in MMM over the next year, and more than a quarter named it their single most important measurement method. Part of what changed is cost. Google released its Meridian mix-modeling framework openly in early 2025, and Meta maintains its open-source Robyn package, so a capable analyst can now run a model that used to mean a six-figure engagement. The current best practice is to triangulate. A mix model gives you the strategic picture of where growth actually comes from. Incrementality tests pressure-check the individual channels you are unsure about. Attribution still earns its keep for day-to-day optimization inside a channel.

6. How do you know if this is happening in your own account?

You can find out this afternoon by checking whether your best channel's spend and your total revenue actually move together.

Pull the monthly spend for your top-ROAS channel over the last year and set it next to your total company revenue for the same months. If you have been scaling that channel and the top line has climbed with it, good, it may be earning its place. If you have been feeding it more and total revenue has flattened or drifted sideways, that channel is very likely harvesting existing demand it did not create, and its reported return is crediting itself for sales that were always coming.

That test is rough, and a proper mix model will give you the real picture across every channel with far more confidence. But it is enough to know whether the conversation is worth having. Most founders I show this to go quiet for a moment, because they were about to pour their next quarter into exactly the wrong line.

If you want to know what your best-performing ad is really contributing, and where your next dollar should actually go, book a marketing strategy call and we will look at it together.