Opportunity Score: Which Meta Recommendations to Ignore

There is a number sitting in your Ads Manager that looks a lot like a grade.

It goes from 0 to 100, it updates as you make changes, and it comes with a list of things Meta thinks you should do next. Most advertisers react the way people react to any grade: they want it higher.

My reaction is different. Before I touch a recommendation, I ask who wrote the answer key. The score grades you against Meta's preferences, which are sometimes your preferences too, and sometimes not.

What is Meta's Opportunity Score, exactly?

Meta's developer documentation describes it as "a tool for understanding how well-optimized an ad account is," scored from 0 to 100, where a higher score "indicates better optimality and a greater likelihood of improved performance over time."¹ It updates in near real time as you change campaigns or apply recommendations.¹

Each recommendation carries a point value, which is the lift in score Meta expects you to get by applying it.¹ Meta describes the recommendations as experimentally proven best practices personalized to each account, and says they can cover goals, audience, automation, creative, placements, budget and bidding, or signals.¹

Where it came from

Meta tested the score earlier this year. In February, it said advertisers who followed the recommendations in early tests saw a 5% median decrease in cost per result.² I covered that announcement in Advantage+ Sales Campaigns. When it rolled the tool out to all advertisers on June 9, 2025, Meta reported a 12% median decrease in cost per result among advertisers who adopted recommendations in testing.³

That is a real signal worth taking seriously. It is also a company reporting on its own product, and I have not seen the sample or the setups behind it. I read it as "there is probably something useful in here," not "do everything it says."

Which recommendations are worth applying?

The ones that fix a gap without changing who the system targets or what it optimizes toward. Think of them as housekeeping.

Apply these, after a quick read

• Creative formats you are missing. If the system says you have no vertical video for Reels and Stories, it is usually right. You are not giving up control, you are giving the system more to work with.
• Obvious setup gaps. Missing tracking parameters, an unfinished conversion setup or an ad set with a single ad all limit what the system can learn.
• Signal improvements (the events themselves are covered in Pixel Events for Coaches). A recommendation that asks you to send better or more complete conversion data is almost always worth doing, as long as the data is accurate and you are allowed to send it.

Test these, never just click

Some recommendations change the targeting or the automation level: widening an audience, switching to a more automated campaign type or merging budgets. Meta has been pushing in that direction all year, from streamlined Advantage+ setup to renaming shopping campaigns as Advantage+ sales campaigns.²

Those moves can work well when the system has strong conversion data to learn from. They can work badly when it does not, because a broader audience guided by a weak signal is just a bigger haystack. The recommendation will not know which situation you are in.

Which recommendations should you ignore?

Ignore any recommendation you cannot explain back in one plain sentence. If you cannot say what changes, what it costs you and how you would know it worked, you are not making a decision. You are following an instruction.

Ignore the ones that give up control without earning it

A few patterns deserve suspicion:

• Consolidation that mixes different offers (consolidation has its place, as in Scaling Meta Budgets Without Resetting Learning). If two products have different buyers, different margins or different sales cycles, merging them to save learning budget can bury the one that matters.
• Automation upgrades on thin data. The system optimizes toward the events you give it. If those events are leads of mixed quality, more automation just finds more of the same mixed quality, faster.
• Anything that quietly changes your goal. Read the details. A recommendation that shifts what the campaign optimizes for can change your results more than any creative will.

How do you test a recommendation without wrecking your account?

Treat it like any other change: one at a time, with a baseline.

What should actually drive the decision?

Your conversion data. The score is a proxy for good setup, and good setup is only worth something when it moves the numbers you are paid on.

The reason I care so much about signal is simple. The score measures configuration, but the delivery system learns from outcomes. If the outcomes you send are shallow, a perfect score on top of them is a polished version of the wrong thing.

So here is my own order of operations. First the offer and the page. Then clean, honest conversion tracking. Then creative that gives the system something worth showing. The score comes last, as a checklist I skim for things I missed.

If you want a rule of thumb: let Opportunity Score find the potholes, and let your own numbers choose the route. That order, offer then tracking then creative, is how the system I build for clients is organized.

Keep building.Ekai