You launched four new ads on Monday. It is now Friday, the account has spent £900, and the new ads have about £6 of it between them. Nothing was rejected, everything says Active, and the tired winner you were trying to replace is still taking every pound. When your Meta ads are not spending, the platform leaves you to work out which of two very different problems you have: ads the auction judged and turned down, or ads the auction never judged at all.
The distinction matters because the fixes point in opposite directions. If the creative is weak, you need better creative. If the structure is starving it, better creative will starve too, and you can burn a quarter's production budget feeding ads into a machine that tests none of them. Most troubleshooting guides hand you the same checklist (billing, review status, spend caps) and stop exactly where the real question starts.
One scoping note before we start. This page is for ads getting little or no spend. If your ads are spending fine but results have fallen off a cliff, that is a different investigation, and we wrote it up separately: run the Meta ad performance drop diagnostic instead. For everyone staring at a column of zeros, work the four gates below in order. Each gate either names your problem or clears you through to the next.
Gate 0: clear the mechanical blockers
Two minutes, before anything clever. When the answer is boring, it is usually one of these:
- Payment failed. Expired card, bank block, unpaid balance. Everything stops at once.
- A spending limit you forgot. Account spending limits and campaign spending limits both kill delivery silently when hit.
- Ads stuck in review or rejected. Check the Delivery column, not just the toggle. New accounts can wait hours in review.
- Scheduling. An end date that already passed, or dayparting that excludes most of the day.
These all share one signature: the whole account stops, or never starts. If other campaigns are spending normally while your new ads sit at pennies, none of these is your problem. The rest of this page is.
Until the setup gives a new ad a fair test, zero spend is a verdict on the setup, not the ad.
Gate 1: did the ad actually get a fair test?
Before you grade the creative, check whether it ever sat the exam. A fair test has a definition you can hold the account to: spend roughly equal to 1x your CPA target, accumulated over 7 days. Sam Piliero of The Moonlighters puts it as a minimum-budget rule: if your CPA target is $47, a new ad set gets a $47 per day minimum for its first 7 days, so it can prove itself on actual purchases rather than add-to-cart proxies. His wider point is that the old "spend 20% of budget on testing" convention is backwards. The best test spends as little as possible while still reaching a verdict.
The 7 days are not negotiable, and that comes from Meta, not from us. Meta's learning phase documentation says an ad set needs about 50 optimisation events in the 7 days after its last significant edit before delivery stabilises, and warns that performance during learning is not necessarily indicative of what follows. Judge an ad at 48 hours on £12 of spend and you are reading noise with a confident face.
How often do ads fail this gate? In our accounts, constantly. In September we audited a UK ticketing and hospitality account we work on, roughly 120 live ads. In a single week, 43 to 48 of them sat in the state we tag as untested: approved and active, but starved. Twelve UGC variants received between £1 and £18 each for the whole week. Meanwhile three static ads ran at over 100x ROAS on £30 to £35 of spend each, while one retargeting static took £343 to return 3.6x. The sharpest case was a UGC ad with 2.3x the account's median CTR, exactly the early signal you hope for, sitting at £45 of lifetime spend with no verdict. Spend in that account was not tracking creative merit. It was tracking tenure.
Gate decision: if your quiet ad has not had 1x CPA of spend across 7 days, you cannot grade the creative yet. Something upstream is starving it. Go to gate 2.
Gate 2: is the structure starving it?
Start with how Advantage campaign budget (CBO) actually treats new ads, because most operators assume it tests them. It does not, and Meta says so. There is no per-ad budget control anywhere in Ads Manager or the API, and Meta's own troubleshooting pages describe uneven delivery between ads as intended behaviour: "If your ad is not receiving enough results but other ads using the same budget are, that's okay." Terrads documented the full paper trail through Meta's help pages in September 2026, and it is worth reading in full. The delivery system backs whichever ad it predicts will produce the cheapest next result. A proven ad carries thousands of impressions of prediction data; a new ad carries none. Piliero's shorthand for the outcome: the tenured winner hogs $700 while the new ad set gets $10. And the loop compounds, because every impression the winner receives makes Meta's predictions about it more confident.
So the question at this gate is which structural shape is doing the starving. Five show up over and over:
- New ads sharing a CBO with tenured winners. The classic. Fix: set an ad set minimum budget of about 1x CPA target for the first 7 days, or move testing into its own campaign where new ads never compete with incumbents.
- Too many ads per ad set. Fifteen ads in one ad set means most never reach the auction at all. Piliero's structure is packs of 4 to 6 ads grouped by avatar and concept, judged over 7 to 14 days. Past six, most of the pack is just queueing.
- Fragmented ad sets. Ten ad sets at £20 a day each need their own ~50 weekly optimisation events to exit learning, and none of them will ever get there. Consolidate until each ad set can plausibly clear the bar.
- Exclusion stacks strangling broad delivery. Every audience exclusion shrinks delivery for the whole ad set, and the newest ads feel it first. The current replacement is value rules: instead of excluding a segment, tell Meta a purchase from your high-LTV group is worth 50% more, and let delivery tilt rather than choke.
- Learning resets from constant edits. Significant edits (budget, targeting, creative changes) push the ad set back into learning, per the same Meta documentation above. An operator who fiddles daily keeps resetting the 7-day clock, and an ad set in perpetual learning never stabilises into spend.
There is also a volume check hiding in this gate. Piliero's production formula is one new ad per week for every $10k of monthly spend, with a floor of two a week. Launch 20 new ads a week into a £6k-a-month account and starvation stops being a malfunction. It is arithmetic. The budget physically cannot give that many ads 1x CPA each, so most of them were dead on upload. We put our own tiered numbers on this in the creative testing quota by ad spend.
Gate decision: if any of the five shapes describes your account, fix the structure and give the ads their 7 days before you touch the creative. If your structure is genuinely clean, go to gate 3.
Gate 3: is something capping delivery?
A clean structure can still be throttled from the edges. Three checks:
Bid caps and cost caps set too tight. With a cap in place, Meta simply stays out of auctions it predicts will breach it. The symptom is distinctive: ads active, no warnings, spend near zero. If you run caps, loosen or remove them on the testing campaign and watch whether delivery returns within 48 hours. You can reinstate discipline once the ads have a verdict.
Audience too narrow. Ads Manager will usually warn you, but stacked exclusions plus a small custom audience can leave an ad set with nobody affordable to reach. Broaden first, refine later.
CPMs inflated by your landing page. The strangest one, and the most underdiagnosed. Meta's pixel reads the page you send traffic to, and aggressive or sensitive claims on it can gate how widely your ads are shown, which surfaces as slow spend at CPMs several times your account norm. Piliero describes an account whose CPMs fell from around $200 to around $60 after editing website claims alone, with no creative changes. If you sell in a claim-heavy category (supplements, beauty, fintech, health) and spend trickles at absurd CPMs, audit the landing page copy before you brief another video.
Gate decision: caps loosened, audience broadened, landing page cleaned, and still no spend after a reset week? You are one of the rare true setup mysteries; open a support case. But if spend now flows and results are still poor, proceed to gate 4, because you finally have a creative verdict coming.
Gate 4: now you can judge the creative
Only here, with 1x CPA of spend across 7 days and the structure acquitted, does "not spending well" become a statement about the ad. Grade it on the leading metrics rather than CPA alone: hook rate, hold rate and CTR, read against current hook rate benchmarks. An ad that got a fair test and posted a weak hook rate with a weak CTR lost honestly. Kill it without sentiment; the CPA kill rule gives you the exact thresholds.
When the creative really is the problem, the fix is rarely another edit of the same concept. Go back a step: mine your reviews for the language buyers actually use, and build the next batch across different personas and angles instead of five recuts of one script. The hard part is sustaining that cadence of fresh, distinct ads week after week at the volume your spend demands. That is the gap a done-for-you creative studio exists to fill, and it is most of what we do all day. Across Spark client accounts, that loop of structured testing plus steady creative volume runs at an 8.7x blended ROAS.
The decision table: symptom, cause, fix, wait time
| Symptom | Most likely cause | Fix | How long to wait |
|---|---|---|---|
| Nothing in the account is spending | Payment failure, spending limit hit, ads in review or rejected, end date passed | Fix billing, check account and campaign limits, check Delivery column and Account Quality | Same day |
| Account spends, new ads get pennies while an old winner takes everything | CBO backing the proven ad; intended Meta behaviour, not a fault | Ad set minimum budget of ~1x CPA target for 7 days, or a separate testing campaign | 7 days at the minimum budget |
| New ad set spends slowly, flagged Learning limited | Budget too fragmented to reach ~50 optimisation events in 7 days | Consolidate ad sets; raise budget per ad set; cut to 4 to 6 ads per set | 7 days after the last significant edit |
| Delivery is a trickle with audience warnings | Stacked exclusions or a tiny audience | Remove exclusions; use value rules to tilt spend instead of excluding | 24 to 48 hours after the change |
| Low spend and CPMs far above account norm | Bid or cost cap too tight, or landing page claims gating distribution | Loosen or remove caps; audit and soften landing page claims | 48 to 72 hours |
| Fair test given (1x CPA over 7 days), hook rate and CTR still weak | The creative genuinely lost | Kill it; brief the next batch from what the survivors share | No wait. The verdict is in |
Key takeaway
A new ad has not failed until it has spent roughly 1x your CPA target across 7 days without delivering. Before that point, treat zero spend as a structure problem: set ad set minimum budgets, cut ad sets to 4 to 6 ads, strip the exclusion stack, and stop editing mid-learning. Judge the creative only after the account has let it compete.
FAQ: Meta ads not spending
Why are my Facebook ads not spending at all?
If the whole account is at zero, the cause is almost always mechanical: a failed payment method, an account or campaign spending limit that was hit, ads stuck in review or rejected, or a campaign end date. If the account is spending but specific new ads get almost nothing, that is different: Meta's delivery system is allocating budget to the ads it predicts will convert cheapest, which favours proven ads over new ones. Meta documents that uneven split as normal behaviour, so the fix is structural (ad set minimum budgets, smaller ad sets, fewer exclusions), not new creative.
How long should I let a new ad set run before judging it?
Seven days minimum, at a daily budget near your CPA target. Meta's learning phase documentation says an ad set needs about 50 optimisation events in the 7 days after its last significant edit before delivery stabilises, and that performance during learning is not necessarily indicative of future performance. A verdict passed at 48 hours mostly measures delivery luck, not creative quality.
What is the minimum budget for a Meta ad set?
Meta enforces a technical floor of only a few pounds a day, but the practical testing minimum is about 1x your CPA target per day for the first 7 days. If your CPA target is £47, give a new ad set a £47 daily minimum so it can prove itself on actual purchases. If that is unaffordable across many ad sets, run fewer concurrent tests rather than starving all of them.
Why is CBO not spending on my new ads?
Advantage campaign budget allocates spend to the ads it predicts will deliver the cheapest next result, and a tenured winner has far more prediction data than a new ad, so the winner compounds while new ads starve. Meta's own troubleshooting pages describe this uneven delivery as intended. To force a fair test, set an ad set minimum budget of about 1x your CPA target for 7 days, or test new ads in a separate campaign.
If this page solved it, the rest of our resources library covers the next problems you will hit: test budgets, kill rules and reading the creative report.