How Much Should You Spend on Facebook Ads Per Day?

Most founders ask this question backwards. They decide what they can afford to lose, divide it by 30, and put that number in the daily budget field. Meta accepts it, spends it, and returns a result that means nothing, because the amount was never large enough for the delivery system to learn anything from.

The daily budget is not a statement about your appetite for risk. It is an input to a machine with a published requirement, and that requirement is the thing that should set the number.

What is the minimum daily budget Meta actually allows?

Meta’s published minimum daily budget is $1 a day for an ad set optimising toward impressions, clicks, likes or video views, and $5 a day for one optimising toward conversions or app installs, according to Stackmatix’s 2026 breakdown of Meta’s minimum daily budget rules.

There is a third floor almost nobody budgets for: where the optimisation event is genuinely low frequency, the minimum is $40 a day, per Cropink’s 2026 guide to Meta’s minimum budgets.

What the ad set optimises for	Published minimum daily budget
Impressions, clicks, likes, video views	$1
Conversions, app installs	$5
Low-frequency events	$40

Both sources read those tiers off Meta’s own documentation, and Meta’s best practices for minimum budgets states the general case, that an advertiser charged for impressions should allocate at least $1 a day per ad set.

Two things about that ladder are easy to get wrong. The figure depends on your billing currency as well as the optimisation goal, per the same sources, so an account billed outside the US sees its own numbers rather than these ones converted.

And the $40 tier is where app advertisers get caught, because per the table above an app install optimisation is listed at $5 while several guides describe app install ads as the low-frequency case.

The practical answer is to read what the interface refuses to accept in your own account and currency rather than to budget from a published table.

Lifetime budgets follow from the daily floor rather than having one of their own. The minimum lifetime budget is the daily minimum multiplied by the number of scheduled days, per Stackmatix’s breakdown, so a seven-day conversion campaign cannot be funded below $35.

A second rule catches people out. If you use a cost cap, the daily budget has to be at least five times the cap, per the same breakdown, so a $20 cost cap forces a $100 daily minimum before you have formed any opinion about whether $100 is the right amount to spend.

Those figures are platform floors, not recommendations, and the distinction is where most wasted budget starts. Meta will take five dollars a day and deliver against it.

What it will not do is produce enough conversion events at that level for its own optimisation to function, which means the money buys impressions and no information.

The ad set runs, the dashboard fills with numbers, and none of those numbers tell you whether the creative works or the offer works, because the delivery system never had enough signal to stop guessing.

Treat the published minimum as the point below which a campaign cannot legally run, and set your actual floor from the arithmetic in the next section.

How do you find the exact minimum daily budget for your own account and currency?

You can read the exact minimum daily budget for your own account and currency out of Meta’s API instead of inferring it from somebody’s published table.

The Marketing API carries a Minimum Budget endpoint whose documented job is to return “the minimum daily budget value for an ad set in an Auction campaign”, and per Meta’s Minimum Budget reference it answers with four numbers and the currency they are denominated in.

Field Meta returns	The floor it sets, in Meta’s own words
min_daily_budget_imp	an ad set optimised for impressions
min_daily_budget_high_freq	“high frequency actions like clicks and likes”
min_daily_budget_low_freq	“low frequency actions like mobile app installs”
min_daily_budget_video_views	an ad set optimised for video views
currency	“the currency these budgets pertain to”

Two things in that list settle questions this page has had to hedge.

There is no conversions category in it at all, and Meta’s own field description files mobile app installs under low frequency actions, so an app advertiser reads the low-frequency value for their own account, per that reference, rather than choosing between the $5 and $40 tiers in somebody’s table.

And the numbers are per currency by construction rather than by conversion, since the endpoint hands back a currency field alongside them.

The reference types those values as plain integers and does not spell out the unit, which is the standard way to misread them by a factor of a hundred.

Per Meta’s documentation on bidding and budgets, the amount you set for bid and budget is “at ad account currencies minimum denomination level, such as cents for USD”, so read what comes back in the same units: a returned 500 on a dollar account is $5.00 rather than $500.

The endpoint is read only and takes no parameters, and under manual bidding the figure it returns depends on the bid amount in use, so a cost cap or a bid change can move your own floor without anything else changing.

If nobody on the team works with the API, the fallback is still the interface. Type a number into the daily budget field in your own account and currency and read what it refuses, because that refusal is generated from the same four values.

How much does Meta actually charge per day?

Meta charges for what it delivered, not for the number you typed into the daily budget field, and it may exceed that number on any single day. Its daily budget rules set the single-day ceiling at up to 75% above the daily budget and cap the calendar week at seven times it, according to Jon Loomer’s write-up of Meta’s increased budget flexibility.

So, per those rules, a $100 daily budget can bill $175 on a Saturday and still finish the week at or under $700.

A single day above budget is designed behaviour rather than a fault. A week above seven times the daily budget is not.

Daily budget you set	Most Meta may spend on one day	Most Meta may spend in a calendar week
$20	$35	$140
$50	$87.50	$350
$100	$175	$700
$179	$313.25	$1,253
$286	$500.50	$2,002

The 75% figure is the current one, and it replaced a 25% ceiling that guides published this year still quote, which is why advertisers working from the older number read a normal Saturday as a billing fault. Per the same write-up, a $20 daily budget allowed $25 under the old rule and allows $35 under the current one.

Meta’s own developer documentation is one of the places still carrying the older number, which is worth knowing before you quote it at anyone.

The Marketing API’s page on bidding and budgets states that “up to 25% more than your daily budget may be spent” and works it through as a $10 daily budget spending up to $12.50.

Treat the 75% ceiling as the one your billing will follow, per the write-up above, and the 25% line as documentation that has not caught up. Do the planning on the weekly figure, which both versions agree is seven times the daily budget.

None of that is what lands on your card, because Meta does not bill daily. It bills when unpaid spend reaches a billing threshold and again on a fixed monthly date for whatever is outstanding, and new accounts start at the lowest threshold of $25, according to Meta’s help page on billing thresholds.

Published versions of the ladder above that first step contradict each other, so do not plan cash flow around a specific sequence. The shape holds even where the steps do not: a new account is charged small amounts every day or two at first, and a mature account at the same spend is charged larger, rarer amounts.

Two things follow for how you read the number. Read spend on a rolling seven days rather than day by day, because a day above budget is designed behaviour rather than a fault, and a daily-variance panic is how most people end up making the edits that reset the learning phase.

And if the weekly ceiling is more than you can carry, set an account spending limit, which is a hard stop at account level rather than a pacing preference, so unlike the daily budget it does not flex.

If the account is brand new, the setup steps that gate a first campaign are worth clearing before you fund it, because a payment method that fails its first charge costs more delivery than any budget decision you will make in week one.

What does a daily budget actually buy on Meta?

A daily budget buys impressions at the going auction rate, and the useful way to read it is in conversions per week rather than in dollars per day.

At the medians Triple Whale reports for ecommerce across full-year 2025, a $14.19 CPM and a $38.19 cost per acquisition, $100 a day buys roughly 7,000 impressions a day and about 18 conversions a week, and it takes about 19 days to accumulate the 50 events that end the learning phase.

Daily budget	Impressions a day	Conversions a week	Days to reach 50 events
$10	705	1.8	191
$25	1,760	4.6	76
$50	3,520	9.2	38
$100	7,050	18.3	19
$273	19,240	50.0	7
$500	35,240	91.6	4

Every row in that table is arithmetic on two numbers from Triple Whale’s 2026 Facebook ad benchmarks, the $14.19 median CPM and the $38.19 median cost per acquisition it reports for ecommerce brands on full-year 2025 data. Your own CPM and cost per acquisition will differ by category and country, and the shape of the table will not.

The row that matters is $273, because that is where a single ad set reaches 50 events inside seven days at the median cost per acquisition. Everything below it is an ad set that never leaves the learning phase, which is a different purchase from a smaller version of the same one.

A $50 a day ad set is not a third of a $150 a day ad set. Per the table above it is an ad set that spends 38 days collecting what Meta needs in 7, by which point the creative is fatigued and the season has moved.

Two adjustments make the lower rows usable rather than wasted. Optimise for a cheaper event, because the arithmetic runs on the cost of whatever event you chose rather than on a purchase: at a $4 cost per add-to-cart, 50 events in a week costs about $29 a day instead of $273.

Or run one ad set rather than four, since the 50-event threshold is measured per ad set and splitting a small budget guarantees that none of them clears it.

How do you calculate a daily budget from your target CPA?

Calculate the daily budget by multiplying your target cost per acquisition by 50, then dividing by 7.

The 50 comes from Meta’s own guidance that an ad set needs roughly 50 optimisation events in a rolling seven-day window before it exits the learning phase and delivery stabilises, as set out in AdLibrary’s 2026 guide to the Meta learning phase.

The count is measured at ad set level and includes browser pixel events, server-side events sent through the Conversions API, and Meta’s own modelled conversions.

Run the arithmetic and the answers are larger than most brands expect.

Target cost per acquisition	Events needed in 7 days	Minimum daily budget per ad set
$10	50	about $71
$15	50	about $107
$25	50	about $179
$40	50	about $286
$75	50	about $536
$150	50	about $1,071

Cropink’s 2026 Facebook ads budget guide works the same calculation and lands on about $179 a day for a $25 target, and Stackmatix reaches roughly $214 a day for a $30 target using the same rule. The number is not a matter of opinion. It falls out of what one conversion costs you and how many of them Meta says it needs.

If the figure the table gives you is more than you have, the fix is to change the optimisation event, not to shave the budget and hope.

Two limits on the arithmetic are worth stating. It sizes a single ad set optimising toward a single conversion event, so it says nothing useful about a budget already split across a structure. And it is a floor for learning rather than a forecast of results: clearing 50 events buys you a cost per acquisition you can trust, not necessarily one you can afford.

Why doesn’t $10 a day work on Facebook ads?

Ten dollars a day does not work on Facebook ads because it cannot produce enough conversion events for Meta’s delivery system to model anything, no matter how long you leave it running. At a $25 cost per acquisition, $10 a day buys roughly three conversions a week against a requirement of 50.

The ad set does not fail loudly. It sits in the learning phase or tips into the “learning limited” state, and the costs stay volatile because the system is still effectively sampling at random.

What makes this expensive is that the campaign still produces a number. You spend $300 over a month, get a dozen or so purchases at a cost per acquisition well above the target, and conclude the creative does not work or the product is wrong.

Neither conclusion is supported. You ran a test with a sample size too small to distinguish a good ad from a bad one, and then made a strategic decision on the result.

There are two honest ways out at that budget. Optimise for a cheaper event higher in the funnel, such as an add to cart or a lead, so that 50 events a week is actually reachable.

Or accept that the spend is buying reach rather than learning, run it on a traffic or engagement objective, and judge it on click-through rate and cost per click rather than on cost per acquisition. Both are defensible.

Running conversion optimisation at that budget and reading the result as evidence is not.

How should you split the budget between ad sets?

Do not split it. The 50-event requirement is measured per ad set, so dividing $200 a day across four ad sets produces four ad sets that each miss the threshold rather than one that clears it comfortably. Fragmenting the budget is the single most common way brands with adequate spend still fail to get out of the learning phase.

The practical default for a brand running one offer is a single ad set with broad targeting and several creatives inside it, letting Meta’s delivery system decide which creative to serve. Campaign budget optimisation does the same job one level up, concentrating spend on whichever ad set is currently delivering rather than holding a fixed amount hostage in an ad set that is not working.

Audience splits earn their cost only once you can afford for each split to clear 50 events on its own. At a $25 target that means roughly $179 a day per audience, so a two-audience test needs about $358 a day before it is a test rather than two underfunded guesses.

Creative splits are different and cheaper, because several creatives can live inside one ad set and share its event volume. That is why creative testing scales better than audience testing at small budgets, and why creative volume rather than audience segmentation is where most brands should put their effort first. We work through the volume side of that in how many ad creatives to test per month.

What does a realistic first month of spend look like?

A realistic first month is one ad set, broad targeting, a budget sized from the table above, and no structural changes for at least 14 days. For a brand with a $40 target cost per acquisition, that is roughly $286 a day, or about $8,600 for the month, in one place rather than spread across a structure that looks sophisticated and learns nothing.

Expect the media cost itself to be higher than last year.

Triple Whale’s 2026 Facebook ad benchmarks, drawn from roughly 35,000 ad accounts across full-year 2025, put the median ecommerce CPM at $14.19, up roughly 20% year over year from about $11.82, with a median cost per acquisition of $38.19, a conversion rate of 1.57% and a return on ad spend of 1.86x.

Notably, Triple Whale reports that cost per acquisition stayed close to flat despite the CPM increase, rising about 1%, which it attributes to Meta’s Andromeda delivery engine matching ads to better-fit buyers.

The first month buys two things: a stable cost per acquisition you can trust, and a shortlist of creatives that beat the others. It does not usually buy profit. A brand that needs month one to be profitable should lower the optimisation event, shorten the funnel, or wait until it can fund a real test, because the alternative is paying for a month of data and then not believing it.

How do you know the budget is big enough to be learning?

You know the budget is big enough when the ad set has stopped learning, so check its delivery status rather than its results.

Meta labels an ad set as in the learning phase, out of it, or “learning limited”, and that label answers the question directly: an ad set that is still learning after a week is an ad set whose budget is too small for its optimisation event, and every cost figure it reports is provisional.

Work through it in order. Start by counting actual optimisation events over the last seven days and comparing the total with 50.

Then look at whether the daily cost per acquisition is settling into a range or still swinging wildly day to day, because stability is what leaving the learning phase actually buys you.

Last, check whether you reset the clock yourself: significant edits, including budget changes above roughly 20%, bid strategy changes, targeting changes, new creative and a switched optimisation event all push an ad set back into learning, per Cometly’s 2026 write-up on learning phase optimisation.

Self-inflicted resets are the quiet killer. A campaign that has been live for three weeks with a tweak every few days may never have assembled seven uninterrupted days of learning, which means it has been paying learning-phase prices for a month.

Set the budget, leave it alone, and diagnose after the delivery system has finished doing its job. If performance drops later, the causes are narrower than they look and we separate them in why did my Facebook ads suddenly stop working.

So what is the short version?

The short version is that your daily Facebook ads budget is set by your target cost per acquisition, not by your tolerance for loss. Multiply that target by 50, divide by 7, and you have the daily spend per ad set that Meta’s learning phase requires. At a $25 target that is roughly $179 a day and at a $40 target roughly $286 a day, per Cropink’s 2026 budget guide.

Keep it in one ad set. Splitting a marginal budget across several ad sets guarantees that none of them clears the threshold. Leave it untouched for at least 14 days, because edits reset the learning clock and a campaign that has been edited weekly has never actually finished learning.

If the required number is out of reach, change the optimisation event rather than the budget. Optimising toward a cheaper event higher in the funnel is a legitimate answer at small spend.

Optimising toward purchases at that kind of budget and treating the result as a verdict on your creative is not, and it is the most expensive mistake available at the lowest budget.

Whether the return you then get is any good is a separate question, which we cover in what is a good ROAS on Meta and TikTok ads.