TerradsTerrads
← All posts

Meta has no per-ad budget, and says your new creative getting no spend is intended

A Meta ad has no budget field, in the interface or the API. Meta calls the uneven delivery intended behaviour. Three documented paths, each with a cost.

You add three new creatives to a working ad set. By morning the proven ad has spent the budget and the new ones have a few hundred impressions between them. The question that follows (how do I make Meta actually test these?) has a documented answer that most people don't expect.

Meta documents no per-ad budget, spend floor or spend cap anywhere. Not in Ads Manager, and not in the API: the Marketing API's Ad object has no budget field at all, and the one per-ad bid control it used to have is deprecated in favour of the ad set. Meta describes uneven delivery between ads as the system working, on two separate troubleshooting pages.

There are three documented ways to get a new creative in front of people, and each of them costs you something specific.

Meta says uneven delivery is okay

The clearest sentence is on the page for advertisers whose ads aren't getting results:

If your ad is not receiving enough results but other ads using the same budget are, that's okay. The ad delivery system predicts which ads will generate the best results. Sometimes this means that other ads in the same ad set or campaign won't deliver. If you want your ads to deliver equally, try creating an A/B test.

And on the page dedicated to this symptom, specific ads in your campaign aren't delivering enough:

Remember that it's normal for some ads to have more impressions than others. Over time, our system learns which ads perform the best for you.

That second page is Meta's troubleshooting article for this exact complaint, and its checklist is entirely about faults: is the ad on, is it in review, was it rejected, is the creative eligible for the placement. It offers no budget control, no spend floor, and does not mention the creative-testing tool at all. Its remedy of last resort is "deleting and recreating your ad".

Why the proven ad compounds

The mechanism is documented and it is a feedback loop, not a budget split:

If you have multiple ads in your ad set (or across your account), we'll show the ad that's most likely to achieve the lowest cost per optimization event for the given person. This means that each of your ads won't necessarily be delivered the same number of times.

And:

Each time an ad is shown, the delivery system's predictions of relevance become more accurate. As a result, the more an ad is shown, the better the delivery system becomes at determining where, when and who to show the ad.

Read those two together. Delivery goes to the best predicted cost, and prediction sharpens with impressions. The ad that already has impressions is the ad the system knows most about, which makes it the safest prediction, which earns it more impressions. A brand-new ad is, by construction, the one the system knows least about.

The auction gives you no per-ad setting here either. Meta lists three inputs to the auction (bid, estimated action rates, and ad quality), and only bid is the advertiser's to control; by default Meta sets it for you. Bid is an ad-set setting. The creative feeds the other two, but they are Meta's predictions, and a new ad's predictions are the least informed.

There is nowhere to put a per-ad budget

If you want to know whether a control exists, read the object model rather than the help centre. Meta's Marketing API describes its Ad object like this:

Each ad is associated with an ad set and all ads in a set have the same daily or lifetime budget, schedule, and targeting.

There is no budget, daily_budget, lifetime_budget, spend_cap or daily_min_spend_target field on an ad. Those live on the ad set and the campaign. And the one field that used to let you influence a single ad is gone:

bid_amount — Deprecated. We no longer allow setting the bid_amount value on an ad. Please set bid_amount for the ad set.

Every standing budget control Meta documents (daily, lifetime, Advantage+ campaign budget, minimum and maximum spend limits, ad set budget sharing) is scoped to the ad set or the campaign. The data model has no place for a budget on an ad. (How far a daily budget can run over on one day, and over a week, is in how much Meta can overspend a daily budget.)

There is one exception, and it is a temporary one: the Creative test tool below lets you set a share of an existing budget to spend across a batch of test ads. It is the nearest thing to the control people go looking for, it is bounded to a test, and it still never names a figure for an individual ad.

The trap that punishes you for testing in place

One Meta page makes this worse. Meta lists the edits that reset an ad set's learning phase, and the list includes:

Adding a new ad to your ad set

So dropping a new creative into a stable, performing ad set re-enters the learning phase for the whole ad set, the proven ad included. The learning phase ends "after about 50 results in the week after the ad set's last significant edit", and that count is scoped to the ad set, never to an individual ad. (Every edit Meta counts as significant, and what Learning limited means.)

Now the contrast, from the same page:

Will adding a new ad set to the campaign cause the rest of the ad sets to reenter the learning phase? No, adding an ad set to a campaign with Advantage+ campaign budget won't cause other ad sets within the same campaign to reenter the learning phase.

Adding a new ad set to a campaign-budget campaign costs nothing. Adding a new ad to an existing ad set resets that ad set. Both facts are on the same Meta page, and together they push you towards structuring tests as new ad sets: the opposite of what Meta's ad-volume guidance tells you to do. More on that contradiction below.

The three documented paths, and what each costs

1. Creative test: the only one that ring-fences budget in place

This is the tool most advertisers asking the question have never heard of (how Meta's creative test works). Meta:

In a creative test, you can compare up to 7 creative variants with delivery provided to new test ads. The test is set up in an existing campaign so that high-performing ads can continue to run after the test with delivery system learnings retained.

And the budget control, which is the whole point:

Choose the amount of your existing campaign or ad set budget you'd like to spend across test ads. We suggest using no more than 20% of your existing budget… We'll aim to spend that amount per day on test ads.

That is the closest thing Meta has to "force spend onto my new creative". Four caveats, all from Meta's creative test page:

  • It works with the Highest volume bid strategy only. On any other bid strategy (cost per result goal, ROAS goal, bid cap or highest value) you have to switch the campaign's bid strategy to use it.
  • It makes two to seven copies of one existing ad, whose creative you then replace; it does not promote ads you already made.
  • "We'll aim to spend" is an intention, not a guarantee.
  • "A confidence level is not included", and "the test does not make any automatic changes based on the results". You get a ranking against your chosen metric, and you act on it yourself.

2. A/B test: it splits the audience

Meta's recommendation in that first quote was "try creating an A/B test". Meta defines one like this:

We show each version to a segment of your audience and ensure nobody sees both.

An A/B test splits the audience rather than the budget inside an ad set. The unit of a Meta A/B test is the ad set or the campaign (the flow duplicates an ad set), each with its own ad-set-level budget. So "try creating an A/B test" means, in mechanical terms, give the new creative its own ad set. Which starts a fresh learning phase from zero, and adds to your ad-set count.

It is a real answer to "I want equal delivery". It is not an answer to "I want this creative tested inside my winning ad set without disturbing it".

3. Flexible ad format: stop making them separate ads

Meta's third answer is to dissolve the question:

you can select up to 10 images and videos in a single ad campaign, and the ad delivery system will automatically determine what media or media combination… to show to people.

You also don't need to create more ads per ad set to test different formats.

One ad, up to ten assets, the system chooses. The cost is measurement: your new creative stops being a unit you can report on. There is no documented per-asset spend control and no per-asset delivery breakdown, so you cannot say "this new concept got N impressions and performed like this". It is also limited to traffic, engagement, sales and app promotion objectives, and Meta's page carries the line "This feature may not be available to you."

The near-miss: ad-set spend limits

One more control is the fallback when a creative needs an ad set of its own:

You can set minimum and maximum spending limits for each ad set with Advantage+ campaign budget.

Meta's spend limits page calls a minimum "a target range, not a guarantee", and says "No refunds will be issued for failing to meet a minimum spend limit", though one of its examples reads as a hard floor: with a $50 campaign budget and a minimum of at least $20, "we can’t spend less than $20 on your ad set." Plan on the target: a missed minimum is not refunded. The unit is the ad set. To use one to protect a new creative, that creative needs its own ad set, and then you are back to a fresh learning phase and a higher ad-set count. (How minimum and maximum limits behave inside a campaign budget is in Advantage+ campaign budget vs ad set budgets.)

Meta's guidance contradicts itself, and both halves are live

The help centre says run fewer ads, and caps how many a Page can run:

  • "Avoid high ad volumes. When you create many ads and ad sets, the delivery system learns less about each ad and ad set."
  • "Decrease ads per ad set, but maintain diverse creative assets per ad set."
  • "When an advertiser runs too many ads at once, each ad delivers less often… too many ads can result in worse performance."
  • Learning limited is caused by "issues such as running too many ads at the same time".

Meta's Best Practices posts say run more ads:

  • "we recommend maintaining at least 20 diversified ads in Advantage+ shopping campaigns" and "dedicating 20-30% of your overall ad budget towards testing new ads" (19 November 2024).
  • "Launch new ads every week."
  • A case study quoting an advertiser going "from 3-4 new creatives a week to almost 50 on average" (22 April 2025).

Both bodies of documentation are live, and neither has been amended to match the other. The help-centre pages carry no date at all, so you cannot even argue that the newer posts supersede them. If you have felt that Meta's advice on creative volume is incoherent, that is because, as published, it is.

One way to reconcile them (the diversification side has its own post): the volume warnings are about a fixed budget spread thin across near-identical ads, and the diversification guidance is about different concepts in a campaign type built to sort them. But Meta does not say that anywhere, and an advertiser reading both pages is entitled to be confused.

Andromeda does not do what it's blamed for

Meta documents Andromeda as a retrieval system, the stage that narrows "a pool of tens of millions of ads to the few thousand we consider showing someone". Meta's claimed result is "an 8% increase in ads quality". A 2026 engineering post adds: "Since 2025, Meta Andromeda has delivered a 2.4% increase in conversions on Facebook."

No Meta source says Andromeda allocates budget between an advertiser's own ads, and nothing in it gives you a lever. If your new creative isn't spending, the documented explanation is the prediction loop above, not Andromeda.

Getting a new Meta creative tested

  1. Stop trying to force spend at the ad level. The control does not exist, in the interface or the object model. Time spent looking for it is wasted.
  2. Use Creative test if you want the new work funded inside a working campaign. Accept the Highest-volume requirement and the absence of a confidence level, and cap it at Meta's suggested 20%.
  3. Fall back to a separate ad set with a minimum spend limit only when the creative is a genuine new concept worth paying a learning phase for. It has three costs: the minimum is a target, not a guarantee; the ad set starts a fresh learning phase; and if it targets the same people as an existing ad set, it causes auction overlap, where Meta's advice is to combine ad sets aimed at the same people.
  4. Don't drip-feed single ads into a winning ad set. Each one is a significant edit that resets that ad set's learning. If you're going to disturb it, do it once with a batch rather than five times in a week.
  5. Judge the batch, not the ad. Since you cannot guarantee any individual ad a slice of budget, the unit you can control is how different your ads are from each other, which is what the diversification guidance is about. The free creative diversity check groups the look-alikes in a batch before you launch it.

We sell a creative tool, so one caveat: "make more ads" is not, by itself, an answer to this problem. If ten new ads go into one ad set, Meta says "each ad delivers less often" and that "it's normal for some ads to have more impressions than others", so most of the ten may barely deliver, and the ad set's learning resets for the trouble. Volume only pays when the ads are different enough to reach different people, and when the structure you put them in gives them somewhere to spend.

Sources

Accessed September 20, 2026, unless a line gives another date.

Meta revises these pages without notice and most carry no date, so re-check before relying on a quote.

Terrads makes ads that differ in concept rather than in crop — which is the half of this problem an ad tool can actually help with.