How much budget does an advertising test actually need?

"How much should I spend?" is the first question almost everyone asks me, and it is asked as if there were a price list.

There isn't one. The budget is not the price of the advertising. It is the amount of data you are buying, and data is what a decision is made of.

The budget buys data, not clicks

A campaign gives you an answer only when enough people have gone all the way through: seen the ad, opened the page, written or booked. Three enquiries tell you almost nothing — you cannot tell a working campaign from a lucky week.

That is also how the platforms themselves work. Meta describes a learning phase during which an ad set's delivery is still unstable, and its own documentation puts the end of that phase at roughly fifty optimisation events in a week. Meta on the learning phase.

So the practical question is not "how much can I spend". It is "how much do I need to spend before the numbers mean something".

What a test budget actually buys

The daily budgetas many impressions as it pays for
Clickssome of those impressions reach the page
Enquiriessome of those clicks write or book
Enough datato tell a pattern from a lucky week
A decisioncarry on, change something, or stop

Start from your own numbers, not from a benchmark

Someone else's cost per enquiry tells you nothing about your market, your city or your price. Your own figures do.

What a client is worth to you

Take the price of the service, and then take the more useful number: what a client is worth over a year, if they come back. A €70 treatment that a person repeats six times is not a €70 decision.

What you can pay for an enquiry

Decide, before the launch, the highest you would still be happy to pay for one person who writes to you. Not for a click, not for a follower — for an enquiry.

How many enquiries you can actually handle

This one gets skipped, and it matters. If your week has room for four new clients, a campaign built to bring thirty is not a success; it is a source of people you will disappoint.

The arithmetic, in one line

Enquiries you need to judge the test, multiplied by the most you would pay for one, gives the floor of the test: the least it can cost to tell you anything. At up to €10 an enquiry and twenty enquiries before deciding, that floor is €200.

It is a floor, not a typical price. It covers your own reading of the results, not the platform's learning, which on Meta is counted in dozens of conversions a week. A test that small usually has to run for longer, or optimise for an earlier step such as a started conversation — and once the market sets the real price per enquiry, tests often cost more.

Those are your numbers, not a promise. What the market actually charges shows up in the first weeks.

Why a budget that is too small produces no answer

A campaign that is too small does not fail loudly. It produces a handful of results that look either wonderful or terrible, and both readings are noise.

The algorithm stays in learning. Too few conversions, and delivery never settles into a stable pattern.
One good day looks like a trend. With five enquiries, a single quiet week reads as a collapse.
You end up optimising on clicks. Because there are not enough enquiries to optimise on, and clicks are the cheapest thing to be pleased about.

The three ways a small budget gets wasted

Splitting it between channels. Half to Meta and half to ChatGPT Ads means neither gets enough data to say anything. On a small budget, take the channel closer to how your clients decide, get it working, and add the second one later as a test. That choice is a subject of its own: Meta Ads vs ChatGPT Ads: different channels for different moments.

Too many audiences and creatives at once. Five ad sets on a small budget is five campaigns that are each too small.

Stopping on day four. The first days are the noisiest ones. Stopping there means paying for the learning and leaving before the result.

The most expensive advertising is not the campaign that spends too much. It is the one that spends just enough to teach you nothing.
Darina Solomakhina

What I check before we agree a budget

I ask about the price of the service, what a client is worth over a year, how many new clients you can take, which city or country we are working in, and what you have spent already.

Then I say one of two things: this budget can produce a readable test, or it cannot, and here is what would have to change first — a smaller area, a narrower offer, a higher-value service, or waiting until there is more to spend.

Turning that down is part of the work. A test too small to read costs the same as one you can act on, and answers nothing.

Frequently asked

Is there a minimum budget you work with?

I don't publish a fixed minimum, because the honest figure depends on the country, the audience and the price of your service. What I do before the setup is check whether the budget you have can produce enough data to decide on. If it can't, I say so.

Is your fee part of the advertising budget?

No. My fee covers strategy, setup and management. The advertising budget is paid from your own account, directly to Meta or OpenAI.

How long should a test run?

Long enough to leave the noisy first days behind and collect enough enquiries to compare. In practice that is usually a few weeks rather than a few days, and it depends on the budget and how quickly your clients decide.

Can I start small and increase it later?

Yes, and that is usually the sensible order. What doesn't work is starting so small that the test can't be read, then concluding that advertising doesn't work for your business.

What if the test says the numbers don't add up?

Then it has done its job for the price of a test rather than a year. Sometimes the fix is the offer or the page rather than the budget, which is why I look at those first: before launching ads, what your landing page must make clear.