How much should I spend on my first ad test?

Work backwards from the number of results you need to tell your ads apart, multiplied by what each result costs. At average Facebook prices, a first test of three versions that optimizes for clicks needs about $50 to $200 per version, depending on how different the versions are.¹ A test that optimizes for purchases or signups from day one usually needs several times that, because the ad system wants about 50 of those events a week before it settles.²

Most first tests fail on budget rather than creative. The founder spends $30, sees one ad get four clicks and another get two, and picks a winner that the next $30 would have reversed.

How do I work out a test budget?

Use one line of arithmetic: versions × results needed per version × cost per result. The hard part is the middle term, so start with the cost of each kind of result. Table 1 uses the median Facebook figures from WordStream’s 2025 benchmarks, drawn from more than 1,200 US campaigns.¹

Source: WordStream by LocaliQ, Facebook Ads Benchmarks 2025, April 2024 to June 2025.¹ Your costs will differ by country, audience and season.

The table shows why the first test should usually measure something cheap. A hundred clicks cost about $70, while a hundred leads cost about $2,770. A new business rarely has a funnel that converts well enough for a purchase-optimized campaign to learn on a few hundred dollars.

How many impressions do I need to pick a winner?

It depends on how big a difference you want to detect. Table 2 shows the impressions each version needs when the baseline is the median 1.71% click rate, at the usual standard of 95% confidence and 80% power.

Source: Standard Trade calculation from WordStream’s median click rate and the CPM it implies.¹ Two-sided test, 95% confidence, 80% power.

Halving the difference you want to detect roughly quadruples the cost. So test ideas that are far apart, such as a different first line, a different offer or different proof on screen, rather than a new button color or a reworded headline. If two versions differ by 10%, a first test can’t tell them apart at any budget a new business would spend, and it doesn’t need to.

Text us a reel. Get a growth test back.

Send Standard Grow the reel you saved and what you sell. You get back the ads, the budget and the number that counts as a win. Nothing runs until you say go, and you get the results back by text.

Get my first growth test

Your first test is free. Works over iMessage or WhatsApp.

Why shouldn’t I optimize for purchases on a small budget?

Meta’s ad system learns who to show an ad to from the events you ask it to optimize for. Meta’s guidance is that an ad set leaves its learning phase after about 50 optimization events within seven days.² At the median lead cost of $27.66, that is $1,383 a week for a single ad set, and a purchase usually costs more than a lead.

Below that volume the ad set stays in “learning limited,” and its results swing from day to day. A first test will often learn more from optimizing for link clicks or landing page views, then reading signups or sales in your own analytics. Once one version clearly brings in buyers, move that one to a purchase objective with a budget that can reach about 50 events a week.

What are the platform minimums?

TikTok sets hard floors. A campaign with a daily budget must spend more than $50 a day, and an ad group more than $20 a day.³ TikTok also asks advertisers not to change a budget more than once every two days, and to raise it by at most 40% at a time during learning and 30% after.³ A one-week TikTok test with three ad groups therefore costs more than $420.

Meta’s minimum daily budget varies by currency and by what you optimize for, and it’s a few dollars a day for most US campaigns. The practical floor on Meta comes from Table 2 rather than the platform’s rules.

When should I stop a test?

Set the stop rules before launch, alongside the budget:

• Stop a version when it has spent twice your target cost per result with nothing to show for it.
• Don’t edit an ad set during its learning phase. A big edit, such as a new creative or a large budget change, restarts learning.²
• Call the test when every version reaches the impressions in Table 2 for the difference you care about, not when one pulls ahead on day one.
• Write down the result, including a test where nothing won. A reel’s idea that didn’t transfer to your buyer is a finding you won’t have to pay for twice.

How do I get my first ad test running this week for my online store?

Here is the week we would run for a candle shop on Shopify with $150 to spend, using the numbers above. The shop tests three first lines from one reel it saved, each on its own ad set.

Source: Standard Trade, using the median Facebook prices in Table 1.

At $50 per version, each one buys about 4,200 impressions at the median price, close to the 4,444 that Table 2 says it takes to spot a version with 1.5 times the clicks. Write the win rule on Monday, before anything runs. For example: “the version with the best click rate gets $20 a day next week, if it beats 1.71%.” Expect most sales to come later than the clicks, so a week this small will tell you which hook works better and only rarely which one sells.

Tell us what you need checked: info@standardtrade.ai.

Common questions

1. Is $5 a day enough to test an ad?

   On Facebook at median prices, $5 buys about 420 impressions and 7 clicks a day.¹ Over a week that is enough to tell apart two versions where one gets twice the clicks of the other, and not enough to judge signups or sales.

2. Should I split the budget evenly across versions?

   Yes, for a test. If all versions share one budget, the ad system shifts spend to whichever looks best in the first hours, which is often noise. Give each version its own ad set and budget, or use Meta’s A/B test feature, which splits the audience evenly.

3. How long should a first ad test run?

   Until each version has the impressions it needs, which at $15 to $50 a day per version is usually three to seven days. Running past a week can also smooth out weekday and weekend differences.

4. What if every version loses?

   Then you learned that the idea doesn’t sell to this audience before you spent more on it. Change the offer or the audience next, not the button.