You have about €1,500 a month for paid ads. That is €50 a day, and you want to know which platform lets you start smallest. The help centers hand you numbers that look like an answer: Meta takes a few dollars a day, Google Ads accepts tiny budgets, TikTok wants more than $20 per ad group and more than $50 per campaign. On those numbers Meta and Google look cheap and TikTok looks expensive.

For most small accounts that ranking is wrong. What decides whether €50 a day works is your cost per conversion, multiplied by what each platform says it needs to learn.

Short answer: The minimum daily budget that works on Meta, TikTok or Google Ads is your target CPA times each platform's own learning multiplier. Meta's learning phase implies about 7× CPA a day per ad set, TikTok's help center suggests 10× CPA per ad group, and Google suggests at least 3× CPA for Performance Max.

The takeaways

Official floors ($20 per TikTok ad group, $50 per TikTok campaign, a few dollars on Meta) decide whether an ad is allowed to run. They say nothing about whether it can exit learning.
At a €20 target CPA, a learning-viable day costs about €143 per Meta ad set, €200 per TikTok ad group and €60 for a Performance Max campaign.
If no platform's number fits, optimize for an earlier, cheaper event or consolidate to one campaign on one platform before splitting money three ways.

What are the official minimum daily budgets on Meta, TikTok and Google Ads?

The official minimums are low and they differ in kind. TikTok's help center is the most explicit: a campaign daily budget must exceed $50 and an ad group daily budget must exceed $20. Meta sets its floor per ad account. It depends on your currency and on what you are charged for, and the Marketing API even exposes a "minimum budgets" lookup that returns the value for your specific account. Google Ads is the most permissive. It accepts very small campaign budgets, and it treats the number as an average: on a single day a campaign can spend up to twice its daily budget, according to Google's help center.

None of these figures were designed as advice. Each is the smallest amount the auction accepts, which is where most "start TikTok for $20 a day" articles stop.

Why doesn't the platform floor tell you what to spend?

The platform floor measures what the auction accepts, while the algorithm needs conversions to learn who to show your ad to. Those are two different thresholds, and only the second one decides your results.

Take a TikTok ad group at the $20 floor with a €30 target CPA. That is under one conversion a day. The ad group stays in learning, delivery swings day to day, and the CPA you read after a week rests on five or six events. A bad ad and a starved one look identical.

The same thing happens on Meta at €10 a day and on Google with a thin Performance Max budget. The floor kept the campaign legal, and that is all it did.

How do you calculate a learning-viable daily budget per platform?

Multiply your target CPA by the multiplier each platform publishes for itself:

Meta: Meta's documentation puts learning-phase exit at about 50 optimization events within a week per ad set. 50 divided by 7 is roughly 7.1, so budget about 7× CPA per ad set per day.
TikTok: TikTok's help center recommends an ad group daily budget of target CPA × 10.
Google Performance Max: Google's Ads API documentation suggests an average daily budget of at least three times your CPA for the selected conversion actions.

One caveat: the three are measured differently. Meta's is an event count over a week, TikTok's a daily sizing rule, Google's a soft minimum for one campaign that pools Search, YouTube, Display and more. Read each as the platform's own statement of what it needs, and resist ranking platforms by them.

Which platform is cheapest to start on with a small budget?

The cheapest platform to start on depends on your CPA, because each floor scales with it. Here is the math for two accounts with the same €50 a day:

Target CPA	Meta (per ad set)	TikTok (per ad group)	Performance Max
€20 (a purchase)	~€143/day	€200/day	€60/day
€5 (a lead or install)	~€36/day	€50/day	€15/day

At a €20 CPA, €50 a day fits none of them at purchase optimization, and Performance Max comes closest. At a €5 CPA, all three fit, but only with a single ad set or ad group each. Add a second ad set on Meta and the requirement doubles, because the learning threshold is counted per ad set.

That second point is where small accounts lose the most. The platform floor tempts you into five ad sets at €10 each. The learning math says you can afford one.

What if no platform's learning budget fits your daily budget?

When the multiplication comes out above your budget, change one of the inputs instead of hoping the algorithm copes.

Lower the CPA in the formula by optimizing for an earlier event. Add-to-cart, lead or install events are cheaper and more frequent than purchases, so they clear the threshold on less money. You trade precision for a campaign that can learn.

Consolidate. One campaign, one ad set, a handful of creatives, one platform. Splitting €50 a day across Meta, TikTok and Google gives you three campaigns that never leave learning. Once one platform carries its own weight, the question becomes how to split a budget across platforms using real results. And if the plan is to test creatives inside that budget, the testing math per creative follows the same CPA logic.

Where this fits in a budget plan

The multiplication above is yours to do, and it takes a minute. The harder part comes later: deciding which platform earns the next euro once you have data. I built Adscalr's budget planning around that. It splits spend across Meta, TikTok and Google from each platform's real performance score, with a clearly labelled low-confidence fallback when there is no history. It does not compute learning thresholds for you. Its pacing checks run every 5 minutes and flag an ad set that has spent under 70% of its cap after midday, often the first sign a budget is too thin to deliver. The budget-intelligence page shows how those plans get built.

Minimum daily budget on Meta, TikTok and Google