Pay more to win the Meta auction? Your creative outweighs your bid

The reflex is to think: whoever pays the most. That’s the picture the word “auction” brings to mind: to win, you just need to bid higher.

On Meta, that idea doesn’t hold up. I’d like to follow an ad’s journey, from the tens of millions of candidates down to the one that gets shown, then look at the formula Meta uses to pick the winner. You’ll see that you can win auctions without paying more, and that the main lever sits in the creative.

In short

• Before the auction even starts, Meta eliminates the vast majority of candidate ads, first on their relevance to the person scrolling their feed.
• The auction itself is poorly named. Meta ranks ads on a value that multiplies what the conversion is worth to you by the probability that it happens, then adds a quality score.
• An ad willing to pay less can therefore beat a competitor that bids more, if it has a better chance of converting and better quality.
• Your grip on the bid is limited, and it’s precisely the setting you want to leave alone. That leaves two levers: the probability of conversion and the quality of the ad.
• Both are worked through the creative, and more precisely through its concept: the persona targeted, the angle, the format and the offer. A more specific creative wins more auctions, which lowers your CPM and improves your ROAS or your cost per lead.

Why the highest-bidder idea is so appealing

It’s intuitive. In everyday life, an auction is a salesroom: buyers raise their hands, the price goes up, and the item goes to the highest bidder, or at least to the one ready to spend the most.

It’s also consistent with what we know about Meta’s business model. Meta bills you per impression: you pay it to show your ads on the phones and computers of your audience. The price of a thousand impressions, the CPM (cost per thousand impressions), is the variable that has driven its revenue up for several years, since the number of ads users are willing to see has a limit. You could conclude that Meta has every reason to let whoever pays the most win.

The reasoning is logical. It overlooks, however, how Meta protects what gives it its value: the time we spend on its apps.

Reason 1: most ads are eliminated before the auction

For each ad slot, several tens of millions of ads are candidates. That’s an order of magnitude, which you can round to 100 million to fix ideas; Meta itself speaks of “tens of millions” of candidates (Meta Engineering, December 2024).

Meta has to go from those 100 million to a single ad, for each person scrolling their feed, across billions of users, in a very short time: about 100 to 200 milliseconds, by the order of magnitude I use. It’s a race against the clock. Even the best servers can’t analyze 100 million ads in depth in so little time.

To simplify, Meta sends ads through three successive gates.

The first gate shrinks the field. Meta calls it retrieval. Its goal is to cut deep: to know whether an ad is relevant to this person, simply relevant. To do that, Meta computes a distance, a measure of the match between the person and the ad. Using everything it knows about the person would take too much computation: it takes a few hundred signals, and a few hundred signals about the creative. This first calculation ranks the candidates and keeps only a small fraction. According to Meta, this stage goes from tens of millions of candidates to a few thousand, and processes about a thousand times more ads than the following stages (Meta Engineering, December 2024). This is where Andromeda comes in, the retrieval engine Meta introduced at the end of 2024.

The second gate ranks the remaining candidates. This is ranking. On a much smaller number of ads, Meta can afford to go further. It computes a second distance between the person and the ad, this time from thousands of signals, and runs a full analysis of the creative. According to Meta, this ranking estimates the value both for the person and for the advertiser (Meta Engineering, December 2024). Out of it come, to simplify, about a hundred ads: the best candidates to fill the slot.

The third gate is the auction. Among those hundred or so, it picks the ad that wins and gets shown.

What this journey changes for you: before the auction decides between the finalists, your ad has already had to clear two gates where its match with the person carries a lot of weight. A creative that doesn’t look relevant to anyone in particular has little chance of making it to the auction.

Reason 2: the auction formula rewards the probability of converting

The auction itself follows a different logic from the salesroom. Meta gives a simplified formula for it, which sets things out clearly at the conceptual level. The value of each ad in the auction depends on three variables:

Auction value = value of the conversion × probability of the conversion + ad quality

In Meta’s official vocabulary, these three terms are called the bid (the advertiser’s bid), the estimated action rate and ad quality; the ad with the highest total value wins (Meta, help pages on ad auctions; Jon Loomer).

The value of the conversion

This is the expected gain the conversion represents for you, say $100. It depends first of all on your business model and your economics: selling mattresses at $5,000 or underwear at $8 gives very different values. Landing pages that convert very well also play a part. Meta keeps the exact calculation to itself; we know these are the main criteria.

The probability of the conversion

This is the probability, as a percentage, that the conversion happens if Meta shows this ad to this person, right now. For example 1%: a one-in-a-hundred chance that the $100 sale takes place.

Ad quality

Meta gives each ad a quality score, which measures the experience it offers the user. Meta keeps its method to itself; it only says that it calculates it.

A worked example

Take three ads in the final selection, A, B and C. The figures are hypothetical; they’re there to show the mechanism.

B wins the auction, A comes second and C third. Yet B was willing to spend clearly less than A and C. It wins because Meta estimates it has three times the chance of converting that A has, and because its quality is better. C, which offered the highest value, comes last: its probability of conversion is too low.

In other words, the ability to pay a high price for the impression or the click matters less than you’d think. Meta takes into account the probability that your ad, your ad set, your account converts, and the quality of what you show.

The consequence matters, and we’ll come back to it below: you can win auctions without raising what you’re willing to pay, and therefore bring your CPM down.

Reason 3: Meta has every interest in letting the best ad win

This formula follows from the same reasoning as the limit on the number of ads.

What gives Facebook and Instagram their value is that their users come, stay and enjoy being there. If Meta showed only poor-quality ads, we’d like its platforms less, we’d spend less time on them, the value of its CPM would fall, and Meta would lose money. The quality score is there to keep mediocre ads from taking over the feeds.

Seen this way, the formula looks for the best of three worlds:

• your economic interest, that is, what you’re willing to pay to sell your sofa, your mattress or your piece of jewelry;
• the probability that the person scrolling their feed buys;
• the platform’s interest, which wants the experience to stay pleasant.

I find this mechanism very clever. It aligns the winning ad with the one most likely to serve the advertiser, the user and Meta at the same time.

What you really control: two levers out of three

Going back to the formula, the question becomes: what can you act on?

The bid, which we leave aside

The value of the conversion is what Meta calls the bid. Your grip on it is limited. You choose your bid optimization settings: ask Meta to get the most results at the lowest cost, or to aim for the highest value. These settings don’t change the value of your conversion, which depends on your business model.

To be precise: Meta also offers strategies that frame the bid, such as the cost per result goal or the bid cap, while the “Highest volume” strategy, the default, lets Meta set the bid itself (Jon Loomer). These strategies are there to keep cost in check; the question here is how to win more auctions without touching the amount you’re willing to pay.

The probability of conversion: relevance

You raise it with the same lever as at the first gate, when Meta sorts candidates on relevance, and the one that also makes automatic targeting work: the correlation between your target and your ad. That correlation comes from the creative’s concept: the persona targeted, the creative angle, the format and the offer.

Take a reporting software for agencies. An ad aimed at media buyers in their thirties, at an agency of about fifty people, who struggle every Monday morning to prepare the numbers for their client, has a very strong correlation with its target. Meta rewards that conceptual work by estimating a higher probability of conversion.

This extends the idea of specificity. A creative that is too broad, presenting the same software as “a tool for digital advertising in the age of AI”, would reach a gigantic audience, spilling beyond the target market, and would convert very little. A precise creative reaches a smaller audience, but signs far more customers: a precise persona, a precise problem, a precise angle. The goal is to cover your market with small, diverse audiences, and to convert as much as possible within each one.

What’s remarkable is that the quality of your creative system plays a role at every stage of the journey: from the first gate, when Meta cuts deep among tens of millions of candidates, all the way to the very last, when it decides between the hundred or so finalists.

Ad quality: the user’s experience

Meta wants to see that the user experience is good, and it has metrics for that. When users hide your ad, leave negative feedback or write negative comments, it’s a bad sign. Meta has long used these hides as a signal that other people may not want to see the ad, and then shows it to fewer people (Meta Newsroom, September 2014). The goal, then, is to reduce every negative interaction with your ad.

How? By raising relevance, once again. People rarely react badly for no reason; faced with content that doesn’t suit them, though, they’re merciless. The more precisely your ads answer their needs, the fewer hides, negative feedback and negative comments you get, and the better the experience. It’s one more reason to stay specific.

One nuance: a quality ad, in Meta’s sense, can stay very simple. UGC (content shot by users, or in their style) and videos filmed on an iPhone are still entirely possible. What counts is that the ad fits the platform well, so the experience stays good.

What this changes for your CPM

The practical consequence is significant. If a more relevant, higher-quality ad wins auctions without you raising what you’re willing to pay, then you can bring down the price of your impressions, your CPM. And a lower CPM, for the same conversions, raises your ROAS (revenue divided by ad spend) or lowers your CPL (cost per lead).

So the initial reflex flips. When your auctions get expensive, the first question to ask is about your creatives, before it’s about your budget.

I’d rather stay careful on one point: Meta publishes neither the exact calculation of the value of the conversion, nor that of the quality score, nor how it estimates the probability of conversion. The formula is a simplification, and the figures in the example are only an illustration. It’s still enough to show the direction: what’s in your creative weighs on each of the three gates.

The automation of targeting is often summed up by a phrase attributed to Meta: “Creative is the new targeting.” The auction leads us to complete it: “Creative is the new targeting and bidding.” The creative has become the biggest lever for winning cheap auctions, and so a lower CPM and better profitability. Put another way: “Creative is the new media buying.”

Three questions to ask yourself

1. 1For each of your active creatives, can you name the persona targeted, their problem, the angle chosen, the format and the offer?
2. 2When your CPM goes up, is your first reflex about budget and bid settings, or about the relevance of your creatives?
3. 3Do you track your ads’ negative signals (hides, negative feedback, negative comments), and do you know which creatives trigger them?

If you answer no to any of them, start there.

On Meta, the auction is won first in the creative.