If you already run Meta ads and your spend is split across three or four ad sets that never seem to stabilize, the problem usually is not creative or audience. It is that no single ad set is clearing the event threshold the algorithm needs. "Minimum budget" is really a question about events per ad set, and the honest answer ties directly to your cost per purchase and your margin.
This is the number most guides skip, because they stop at Meta's published floor and never connect it to whether the spend is profitable. Let's fix that.
The two "minimums" — and why only one matters
There are two completely different numbers hiding behind "Meta ads minimum budget," and conflating them is how operators waste money.
Meta's technical floor
Meta sets a hard platform minimum by optimization goal. It is about $1 a day for awareness and clicks and roughly $5 a day for conversion and app-install goals, according to Cropink's 2026 guide. Spend below this and the ad set will not deliver at all.
This floor is almost never your real constraint. A conversion ad set running at $5 a day will deliver — and sit in the learning phase forever, because it cannot gather enough events to stabilize.
The learning-phase floor (the one that governs performance)
Meta's delivery system enters a learning phase every time an ad set launches or gets a significant edit. It exits once the ad set collects about 50 optimization events within a roughly 7-day window — Meta's own published benchmark, confirmed in AdManage's learning-phase guide. Miss it, and the ad set drops to "Learning Limited": unstable delivery, higher and more volatile cost per result.
So your real minimum budget is whatever buys ~50 of your optimized events per week, per ad set. That depends entirely on your cost per result.
The budget math: (CPA × 50) ÷ 7
The derived formula practitioners use is straightforward: multiply your expected cost per acquisition by 50 events, then divide by 7 days to get a minimum daily budget per ad set. Cropink states it directly as (Expected CPA × 50) ÷ 7.
This is arithmetic, not a Meta rule, so treat it as a sane floor rather than a law. Here is what it looks like with operating numbers.
Say you run a POD store doing 340 orders a month at a $31 average order value, spending $2,800 a month on Meta — about $92 a day. Your cost per purchase is $20. To clear learning on one ad set you need 50 × $20 = $1,000 in a week, which is $1,000 ÷ 7 = roughly $143 a day.
Your entire account spends $92 a day. That is already below what one ad set needs to exit learning at a $20 CPA — and if you split that $92 across three ad sets, each gets ~$31 a day and collects only ~10 purchases a week. All three sit in Learning Limited at once. The fix is not more total budget; it is fewer ad sets, so the events concentrate. More on that below.
The profit line every "minimum budget" article skips
Hitting 50 events means nothing if those events lose money. Before you commit a learning-phase budget, check that your target cost per purchase clears break-even.
Break-even ROAS is pure arithmetic: it equals 1 ÷ your contribution margin, where contribution margin is the share of revenue left after variable costs (product, shipping, payment fees) and before ad spend.
Take the same store. Your Printify base cost is about $12.40, shipping $4.90, and payment fees near $0.90 — call it $18.20 in variable cost on a $31 order. That leaves $12.80 of contribution, or a 41% margin. Break-even ROAS is 1 ÷ 0.41 = about 2.4x, and your maximum sustainable CAC is that $12.80.
Now the learning-phase math bites. A $20 cost per purchase against a $12.80 ceiling means every order you buy to exit learning loses roughly $7. You can "win" the learning phase and still bleed. The question is never just "what's the minimum budget" — it is "can my margin fund the CAC that reaches 50 events?" If it can't, you raise AOV or margin first, or you optimize for a cheaper, more frequent event. Our guide to profitable ad scaling walks the full margin-first sequence.
What resets learning — and wastes the budget you just spent
Every significant edit throws the ad set back into learning and burns the events you paid for. AdManage lists the resets: changing the audience, swapping creative, changing the optimization event or bid strategy, adding new ads, pausing for roughly 7+ days, or a large budget change — practitioners commonly draw the line around a fifth of current spend at once.
That last one is the trap when you try to scale. Double an ad set's budget mid-learning and you reset the clock, pay the learning tax again, and often push into diminishing returns at the same time. Small, spaced budget nudges generally do not reset learning; large jumps can. If you want the mechanics of scaling without resetting, the ABO vs CBO breakdown covers when a campaign-level budget sidesteps per-ad-set resets entirely.
Fewer ad sets, more events
The single highest-leverage move for a budget-constrained store is consolidation. Each ad set needs its own ~50 events, so four ad sets need ~200 events a week between them. Collapse to one and that same spend buys one ad set across the threshold.
This is the opposite of the "test ten audiences" advice built for big accounts. If your total Meta budget is a few thousand a month, you cannot feed a wide testing matrix and clear learning anywhere. Test fewer concepts, longer. For the updated platform thresholds, see the 2026 minimum budget requirements.
The hidden reason you're "stuck": tracking, not budget
The ~50 count is what Meta sees, not what actually happened. If your pixel or Conversions API drops events to iOS privacy, ad blockers, or a broken tag after a site deploy, Meta undercounts conversions and keeps the ad set in learning even when real-world sales were fine.
So before you raise budget to escape Learning Limited, reconcile Meta-reported purchases against your actual store revenue for the same window. If your backend shows the orders but Meta doesn't, the problem is measurement, and no budget increase fixes it.
This is exactly where seeing ad spend and real orders in one place pays off. Victor — the AI employee from PodVector AI — computes the true per-order profit by reading every Meta ad dollar against every Shopify order and the Printify, Printful, or Gelato cost behind it (product costs he knows before the first sale). When an ad set is stuck, Victor reconciles the reported conversions against real orders, pauses the ad set that loses money, and changes the budget on the one clearing learning profitably. He is not a dashboard you have to read — he does the work inside Meta Ads and Shopify directly.
Once an ad set is healthy, the next question is how long before results stabilize. We cover that in how long the Facebook ads learning phase takes.
FAQs
What is the real minimum budget to exit the Meta learning phase in 2025?
Whatever buys about 50 of your optimized events in 7 days on one ad set. Using the common (CPA × 50) ÷ 7 formula cited by Cropink, a $20 cost per purchase needs roughly $143 a day per ad set; a $40 cost per purchase needs closer to $286. Meta's technical floor of about $5 a day for conversions is far below what actually stabilizes delivery.
Is $10 a day enough for Meta ads?
For a conversion objective, almost never on its own. At $10 a day you would need a cost per purchase near $1.40 to hit 50 events in a week — unrealistic for most POD stores. At that budget you can test an awareness or traffic objective, but conversion ad sets at that level typically stay in Learning Limited.
Does raising my budget reset the learning phase?
A large jump can. AdManage notes that increases over roughly 20% at once often trigger a reset, along with audience, creative, or optimization-event changes. Small, spaced nudges usually do not. If you must scale fast, expect to pay the learning tax again.
Why is my ad set stuck in "Learning Limited"?
Usually one of three things: the budget is too low to reach ~50 events, too many ad sets are splitting the same events, or your pixel/CAPI is undercounting conversions that actually happened. Check tracking first, then consolidate ad sets, then consider budget.
Should I use the minimum budget or a higher one?
Use the budget your margin can fund, not the floor. If break-even ROAS is 2.4x and your learning-phase CAC exceeds your contribution per order, hitting 50 events just loses money faster. Raise AOV or margin, or optimize for a cheaper event, before raising spend.
How can I tell if the problem is budget or profit?
Compare marginal spend to marginal revenue and reconcile Meta's reported orders against your store's actual orders. If backend revenue is steady but Meta shows a drop, it is measurement. If every new order costs more than your margin covers, it is a profit problem. Victor handles both reconciliations for you — seeing ad spend, orders, and supplier cost together so he can tell you the true profit on each ad set and act on it.