Skip to content
Bach.ai

How Does the Facebook Ad Auction Work? (2026)

Published
Drafted with AI; facts checked against the linked sources on . How we write

How does the Facebook ad auction decide which ad wins?

Meta ranks each eligible ad by total value: the advertiser's bid multiplied by the estimated action rate, plus an ad quality score. The ad with the highest total value is shown. Meta says ads with the highest bid don't always win, because a lower bid can beat it with a higher predicted response or quality.

Meta does not hand an impression to the biggest bidder. It runs an auction every time there is a chance to show an ad to someone, and scores each eligible ad on three inputs: your bid, how likely this person is to take your chosen action, and the ad’s quality. The ad with the best combined score wins that single impression.

That is why two advertisers can bid the same amount and pay very different prices, and why a better ad can beat a bigger budget. The rest of this page explains each input and the two you can change.

What is total value in the Meta ad auction?

Meta’s own explanation says it selects the top ads to show a person by their total value score, which combines advertiser value and ad quality. It finds advertiser value by multiplying an ad’s bid by the estimated action rate, then adds the ad quality score (Meta for Business, 11 Jun 2020).

In plain form: total value = (bid x estimated action rate) + ad quality. Meta’s current auction page describes the same idea more loosely: unlike traditional auctions that pick a winner by the highest bid, its system looks at a number of factors equally, and estimated action rates and ad quality together measure ad relevance (Facebook Ad Auctions Explained, as of 14 Nov 2026). Meta does not publish the weights or exact numbers.

Input What Meta says it is Can you change it?
Bid What the advertiser is willing to pay for the result Yes: bid strategy and cost controls
Estimated action rate A prediction of how likely this person is to take your chosen action Indirectly: objective, event, audience, creative
Ad quality Viewer feedback, hides, and low-quality traits such as sensational language Yes: the ad itself
Total value The three combined into one score per impression Only through the three above

What is the estimated action rate?

It is Meta’s machine-learning prediction of how likely a particular person is to take the action you chose, based on your objective, for example visits to your website or purchases. The models consider the person’s behaviour on and off Facebook, the content of the ad, the time of day and interactions between people and ads. Off-Facebook behaviour means things businesses share through Meta’s Business Tools, such as a purchase (Meta for Business, 11 Jun 2020).

This is why signal quality matters. If your pixel or Conversions API sends clean purchase events, the model has something real to predict. If it sends nothing useful, the prediction is weak.

What counts as ad quality?

Meta says the quality score comes from the feedback of people viewing or hiding the ad, plus assessments of low-quality attributes such as too much text in the image, sensationalized language and engagement bait (Meta for Business, 11 Jun 2020). Its current page lists withholding information, sensationalized language and engagement bait.

Quality is not a number you can read in Ads Manager. The nearest visible signals are the diagnostics covered in how to read Meta’s quality and engagement rankings, where Meta shows them.

Does the highest bid always win?

No. Meta says ads with the highest bid don’t always win. Ads with lower bids often win if the system predicts a person is more likely to respond, or finds them higher quality (Meta for Business, 11 Jun 2020). A higher bid raises one term of the score; it cannot rescue an ad people hide or ignore.

How do I improve my odds of winning?

Work on the two inputs Meta says it measures about you, not on the bid alone.

  1. Choose the objective and optimization event that match the result you want, in Ads Manager under the ad set’s Optimization and delivery section, so the prediction is about the right action.
  2. Check that the event you optimize for fires correctly in Events Manager before blaming the auction.
  3. Replace ads that attract hides or low engagement; read the delivery signals in policy or auction: why delivery is limited.
  4. Remove text-heavy images and sensational wording from the creative.
  5. Change the bid strategy last, using which Meta bid strategy to use.

If your CPM jumps, the cause may be auction pressure rather than a fault in your account; the check is in Meta CPM spike: auction pressure or account fault. Terms are defined in the Meta ads glossary.

What does Bach.ai do here?

Bach.ai is our product. It audits Meta and Google Ads accounts and flags tracking and signal-health gaps, creative fatigue and learning-limited ad sets, each with an estimated recoverable spend. Those are estimates, not guaranteed results. See pricing.

FAQ

Does a bigger budget win more auctions?

Not by itself. Meta’s current page lists a sufficient budget among five components of a strong ad, but the total value formula has no budget term. Meta says its system lets businesses of all sizes compete and reach customers on any budget, because each impression is scored on bid, estimated action rate and quality.

Is the formula multiplication or addition?

Meta’s explanation uses both: the bid is multiplied by the estimated action rate, then the ad quality score is added. Some third-party guides write it as one product or sum, so rely on Meta’s wording and its date, 11 June 2020, because Meta’s current page is less specific.

Can I see my estimated action rate?

No. Meta describes it as a machine-learning prediction for a particular person and does not publish it per ad. You can only influence it through your objective, tracking, audience and creative, and judge it by the results you see in reporting.

Does Meta show my ad to the person who values it most?

Meta says the goal is to pair your ad with people most likely to be interested in it. The auction compares ads competing for the same person at that moment, so who wins depends on competing ads, not only on yours.

Sources

Sources: Meta for Business, How Does Facebook Use Machine Learning to Deliver Ads? (11 Jun 2020) and Facebook Ad Auctions Explained (checked 14 Nov 2026).

See what your Meta ads are really costing you.

Connect your account and Bach ranks potential revenue leaks with estimates and actions you can review. Free for 7 days, no credit card.

Start Free Audit
Start your free audit