What Is Estimated Action Rate on Meta Ads (And Why One Ad Takes All the Budget)

You put four creatives in an ad set. By day three, one of them has 90% of the spend and two have essentially none. Nobody touched anything.

That is Estimated Action Rate doing its job, and understanding it changes how you build campaigns — because two of its three inputs are things you control and routinely get wrong.

Table of Contents

• What EAR actually is
• The three inputs
• Why one ad takes all the budget
• The input everyone forgets
• What we spent $400 learning
• What this changes about how you build

Key Takeaways

What EAR actually is

Meta's auction does not simply award impressions to the highest bidder. It awards them to the highest total value, and a large part of total value is Meta's estimate of how likely this particular person is to take your optimisation event if they see this particular ad.

That estimate is the Estimated Action Rate. It decides where your ads serve and to whom.

Two consequences follow, and they are the whole practical payload:

1. A cheaper bid with a higher estimated action rate can beat a higher bid. You are not purely buying impressions, you are buying Meta's confidence.
2. Inside an ad set, Meta is not splitting budget fairly between your creatives. It is backing its estimate. That is not a bug in your setup.

The three inputs

EAR is driven by your bid, your creative, and your URL destination.

Bid is the least interesting of the three for most advertisers. It is a constraint you set from margin, not a lever you optimise daily.

Creative is where almost all attention goes, correctly. It is the biggest input you control, and creative volume is the main thing separating accounts that keep finding winners from accounts that ride one ad until it dies.

URL destination is the one that gets ignored, and it is the reason for this post.

Why one ad takes all the budget

Because that is what you asked for.

When Meta concentrates spend on one creative in an ad set, it has estimated that creative converts best for the people currently available. Redistributing budget to your other three creatives would, by its own model, cost you conversions.

This is why we don't switch ads off inside a testing ad set. Allocation is the test. An ad that stops getting delivery has been judged by a system with far better information than you have, and the correct response is to read the ad set's CPA rather than to intervene at the ad level.

It is also why new creative must not share a campaign with a proven winner unless spend limits protect it. Your unproven ad has no history to support a confident estimate. It will lose, every time, to the ad Meta already trusts — and you will misread starvation as a bad creative. That failure mode is the entire reason for the three-stage structure.

The input everyone forgets

Two identical creatives pointing at two different pages do not have the same EAR.

Meta is estimating the likelihood of your optimisation event — a purchase, a lead, a booking. That event happens on your site, not in the feed. So the destination is not neutral: page speed, whether the page delivers what the ad promised, and whether the conversion action is actually completable all feed the estimate.

The practical implication is uncomfortable if you have been treating the landing page as a separate workstream from the ad: the page is part of the ad. A slow page or an off-message page suppresses delivery before a single human decides whether they like your offer. You will see it as a creative problem — high CPM, weak delivery, poor CPA — and you will iterate on creative, and it will not help.

This is why we build landing pages and ad creative as one job. The method is in how we build Meta ads landing pages with Claude, and the angle-generation side is in Claude prompts for landing page ideas.

What we spent $400 learning

We ran our own Meta traffic to two purpose-built landing pages. Creative was fine. Structure was fine. Bookings: zero.

The problems were all on the destination side, and they were all invisible in Ads Manager:

• The Calendly embed never mounted on a chunk of sessions. It was lazy-loaded behind an IntersectionObserver, and IntersectionObserver silently never fires inside the Facebook and Instagram in-app browsers — which is where essentially all of that traffic was. People arrived at a page whose booking widget simply was not there.
• A cookie banner sat on top of the calendar, so the visitors who did get the embed could not use it.
• The email input rendered white-on-white on mobile, so anyone typing into the form could not see what they had typed.
• The page had too many competing calls to action for something with a single objective.

Every one of those degrades the URL-destination input to EAR, and none of them show up as anything other than "the ads didn't work."

That is the honest version of why this input matters. We did not learn it from documentation.

What this changes about how you build

Ship the page before the creative. If the destination is weak, the creative cannot rescue it and you will spend your budget diagnosing the wrong thing.

Test the page in the environment the traffic actually arrives in. Meta traffic arrives inside the Facebook and Instagram in-app browsers. Those are not Chrome. Behaviour you rely on may silently not run there — as we found the expensive way.

Treat page changes as ad changes. Changing the destination changes the ad's EAR. If you swap the page under a scaling campaign, expect delivery to move, and do not attribute it to creative fatigue.

Judge creative only when the page is not a variable. Otherwise every creative test is really a creative-plus-page test, and you cannot separate the two after the fact.

We build the ad, the page and the tracking as one system rather than three handoffs — Brooklyn Transfers is what that looked like across CRO, local SEO, Meta and Google in a single quarter. Book a strategy call.