Short answer: A runoff Ad Set is a new Ad Set built to give a second, fair test to ads that sold profitably in creative testing but got very little budget because another ad in the same Ad Set took most of the spend. You pool several of these starved runners up together, give them their own budget, and judge them on the same rules as any test. Use it sparingly: only for ads that clearly earned it.
Every media buyer has seen this. An Ad Set launches with five ads. Meta picks one favorite on day one and pours most of the money into it. One of the other ads gets a small slice and still makes a sale at a great cost, with signs that it is reaching new people. Then the test ends, the favorite is judged, and the runner up is forgotten. That forgotten ad might have been your next winner.
Why do good ads get starved in creative testing?
Because Meta does not split an Ad Set's budget evenly. It picks a favorite early and feeds it, and the other ads get leftovers.
This is not a flaw. It is how Meta's delivery system works. It makes a quick prediction about which ad will hit the goal and leans into it. Most of the time that prediction is reasonable. But it is a prediction made on very little data, and it is not always right. The ad Meta favored on day one is not always the best ad in the group. It is the ad Meta guessed would be best.
So when an ad sells profitably on a small slice of spend, you have an interesting signal. It won people over despite getting very little delivery. That deserves a closer look.
What is a runoff, in one sentence?
A runoff is a fair rematch for runner up ads that proved they can sell at a good cost but lost the budget fight to a sibling.
The word comes from elections. When no one clearly wins the first round, the top contenders face off again. In creative testing, the "first round" was the original Ad Set, and the runner ups never really got to compete. The runoff fixes that.
Which ads qualify for a runoff?
An ad qualifies when it sold at a profitable cost, got only a small share of its Ad Set's budget, is reaching new people, and has not already cleared the bar for promotion to Scaling.
Here is how we think about each condition:
- It is still in creative testing. If the ad is already in a Scaling Campaign or the Graveyard, it has a home. Check every relevant campaign by exact ad name before you list it.
- It is not already in a runoff. Do not keep listing ads that already got their rematch.
- It sold, but did not clear the promotion bar. If it already earned promotion, promote it. Runoffs are for the gap between "made a profitable sale" and "proved itself enough to scale." There should be no gap between the two lists. Every profitable seller lands on one or the other.
- It was profitable. Both cost per new customer and return on ad spend at or better than your goals, over a recent window. Check your third party attribution tool, not only Meta.
- It was starved. It got a small share of its Ad Set's spend, or spent less than about one target CPA (cost per acquisition). The whole point is that it never got a real test.
- It reached new people. Its daily frequency (how many times the average person saw it that day) stayed close to 1, especially on the day it sold. This is the one place we treat frequency as a gate rather than a flag.
The frequency condition deserves a moment. In a creative testing campaign with strict exclusions, an ad with low daily frequency is reaching people who have not seen your ads before. A sale there is a real prospecting win. An ad that sold at high daily frequency was likely selling to people who had already seen your other ads. That ad behaves more like a remarketing ad, and a runoff aimed at new people will not help it.
Also check for the large order trap. If one big order is doing most of the work in an ad's return, judge it without that order. If it only passes because of one lucky basket, leave it out.
What does not qualify?
One unprofitable sale, or zero sales, never earns a runoff. Those ads are just killed.
It is tempting to give every starved ad another chance. Resist that. Runoffs are the exception, not the routine. Creative testing moves fast, and if you rerun everything that did not get much spend, you will spend most of your budget on rematches and very little on new ideas. The bar is a real profitable sale at low frequency. Without that, there is nothing to rematch.
When should you build a runoff?
Build it when you have several qualified candidates at once, not one at a time. Pooling runner ups gives the runoff a real shape and makes it a fair contest.
Our general practice is to keep a running list of candidates on every review and wait until enough have built up across recent tests. A runoff with a single ad is just a new Ad Set for that ad, and it loses the point of pitting close contenders against each other. A handful of candidates, each from a different original test, makes a better rematch.
That said, waiting forever is its own mistake. If you have a few candidates and they have been sitting on the list for a while, it is reasonable to build anyway. The longer a good creative waits, the more its moment can pass.
Two more timing notes:
- If the original Ad Set is still young, it may still give the runner up more spend. Leave it in until the original test's checkpoint.
- If the runner up is now getting its Ad Set's money, it does not need a runoff. Only break it out if the original Ad Set gets paused.
How do you build a runoff Ad Set?
Duplicate the qualified ads into one new Ad Set in creative testing, with the same exclusions as your other tests, a modest daily budget, and everything left paused until a person reviews and turns it on.
A practical checklist:
- Use the same audience rules as your other tests. Creative testing should always run to cold audiences with identical exclusions, or the results are not comparable. See why creative testing must run to cold audiences.
- Pick a budget that matches a normal test. The runoff is a test, so give it a test sized budget, not a scaling budget.
- Name it clearly. A clear name tells everyone on the team this is a runoff, which round it is, and when it was built. Strip any old status tags from the ad names so the new test starts clean.
- Build it paused and review it. At our agency, nothing goes live without a person checking it. Building paused avoids launching a half finished Ad Set by accident.
- Break the ads out before pausing their old Ad Sets. If an original Ad Set is about to be paused, pull the runoff ad out first so it is not lost.
How do you judge a runoff?
Exactly like any other creative test. The runoff gets no special treatment once it is live.
- An ad that spends past your kill threshold with zero sales is paused. See kill thresholds.
- An ad that gets enough sales at goal earns promotion. See promoting a winner to Scaling.
- An ad that sells but below goal moves to the Graveyard.
- Watch how the budget splits. A runoff is still an Ad Set, so Meta will still pick a favorite. If it spreads the money thin, the spread vs concentrated check still applies.
The runoff is the ad's second and final chance in testing. If it does not prove itself here, let it go.
Is there a runoff for ads already in Scaling?
Yes, and it solves a related problem. Sometimes an ad wins in creative testing, gets promoted, and then gets almost no spend in the Scaling Campaign, because the established winners there hold onto the budget.
The remedy is similar: put the newly promoted ads into their own Ad Set inside the Scaling Campaign, and use spend limits (a small minimum and a cap) so they are guaranteed a little budget without stealing from the proven winners. Keep the minimum small relative to the campaign, because forcing spend limits Meta's ability to optimize. The goal is to open a door, not steer the campaign.
This version only makes sense for ads that reach new people. A promoted ad with high daily frequency in testing tends to sell like a remarketing ad, and a Scaling Campaign aimed at new people will not give it much spend regardless. Those ads belong somewhere warmer. We cover the Scaling side in inside the Scaling Campaign.
What are the common mistakes?
- Running runoffs for every starved ad. Only profitable sellers at low frequency qualify.
- Building a runoff with one ad. Pool several candidates for a real contest.
- Using different exclusions from your other tests. Then you cannot compare the result.
- Giving the runoff a scaling budget. It is still a test.
- Forgetting to check whether the ad is already in Scaling or another runoff. Check by exact ad name, every time.
- Letting a runoff run forever. It follows normal test rules and ends like any other test.
Key takeaways
- A runoff Ad Set is a fair rematch for ads that sold profitably but got starved of budget by a sibling ad.
- Qualifying ads sold at goal, got a small share of spend, reached new people at low daily frequency, and have not already earned promotion.
- Zero sales or one unprofitable sale never earns a runoff.
- Pool several candidates before building, use the same exclusions as other tests, and build paused for review.
- Judge the runoff exactly like any other test, and let go of ads that fail it.
- A similar approach, with spend limits, helps promoted ads that are being starved inside Scaling.
FAQ
What is a runoff in Facebook ads?It is a new Ad Set that gives a fair second test to ads that sold at a good cost in creative testing but got very little budget, because Meta favored a different ad in the same Ad Set. Several of these runner ups are pooled and tested together.
Why not just promote a starved ad that made a profitable sale?One sale on a small budget is a promising signal, not proof. Promotion should require enough sales at goal to trust. The runoff is the step between "promising" and "proven."
How many ads should go into a runoff Ad Set?Several, ideally from different original tests, so it works as a real contest. A single ad runoff is just a new Ad Set and loses the point of the comparison.
Why does daily frequency matter for a runoff?Low daily frequency on the sale day means the ad sold to people who had not seen your ads, which is a true prospecting win. High frequency means it was likely selling to warm people, and a test aimed at new people will not reflect that.
Should runoffs be used often?No. They are the exception. Creative testing works best when most of your budget goes to new ideas. Runoffs are for the few runner ups that clearly earned another look.
Wondering if your tests are hiding winners?
Many accounts have profitable runner ups sitting forgotten in old Ad Sets. Our free Meta account audit covers a full account audit plus net new reach and cumulative reach reports, so you can see which creative earned more spend than it got. Request your free audit here.
Keep reading
More on the CTSG system:
What Is the CTSG System? Creative Testing, Scaling and Graveyard for Meta Ads
How to Tell if a Meta Ad Set Got a Real Test (Spread vs Concentrated Spend)
When Should You Turn Off a Meta Ad or Ad Set? How to Set Kill Thresholds
Pausing Ads vs Cleaning Up Ads in Meta: Why They Are Two Different Jobs
When Should You Move a Winning Ad From Creative Testing to Scaling in Meta?
See every article on the CTSG system
Ready for your free Meta audit?
We will review your account and run net new reach and cumulative reach reports, free.
Check what you learned
Tap an answer to see if you got it right.
1. What is a runoff Ad Set?
2. Does an ad with zero sales earn a runoff?
3. How do you judge a runoff?
Ready for the full test? Take the Media Buying exam and get certified. Want this run on your account? Get my free audit.
Module 1: Media Buying · article 11 of 19
Next: What Is a Graveyard Campaign in Meta Ads, and Why Keep Losing Ads?
Previous: How to Stop Holding Losing Meta Ads Too Long (The Prisoner of Hope Problem)
About the author
Jake Mehani, founder of Creative Analytics
Jake has spent 15 years running Meta and paid social ads for consumer brands. Meta published his RBX Active Advantage+ case study, he taught the DMAnc Meta and TikTok ads masterclass, and he spoke at Motion's 2024 Creative Strategy Summit. His senior team runs media buying, creative and measurement like an in house team.