Testing volume is limited by how many ads you can make, not by how much you can spend testing them. Once you see that, the problem changes shape, because production is something you can fix.
The budget objection is usually wrong
Brands tell us they cannot test more because testing costs money. Then we look at the account and find twelve ads live, which means the constraint was never the budget.
A test does not need much spend to tell you something useful. It needs enough to get past the noise, and then it needs the discipline to stop. What it really needs is something to test, and that is where most accounts run dry.
The real bottleneck is that every new ad costs a photoshoot, a designer, a round of revisions and two weeks. At that cost per ad, nobody tests much, and the ads that do get made are too expensive to risk on an unproven angle.
Three ways to get more tests out of the same work
Hold the scene, change the line. One photograph can carry five headlines. If the picture is identical and only the words differ, you have five tests of the argument rather than one test of everything at once. This is the cheapest volume available to anyone and most brands never use it.
Stop treating every ad as a production. Not every test needs a shoot. A flat typographic card with three lines on it is a legitimate ad, costs almost nothing, and sometimes wins.
Use what you already own. Your product pages, your reviews, your packaging, your existing photography. The angles worth testing are usually already written somewhere on your own site, in your customers' words, unused.
Volume only counts if the ideas differ
More ads is not more testing. We count live ads in the Meta Ad Library one at a time, and the pattern repeats: a merino apparel brand running about 580 ads across two pages, where seven of the statics were versions of one sentence. A skincare brand with 320 ads live, where seven of its twelve images carried the same customer quote.
Those are big accounts running one idea. Forty ads that are forty versions of one concept is one test.
So the target is not a number of files. It is a number of separate arguments, with enough executions inside each to read the result.
What good looks like
Six to ten angles a month, four to five executions inside each, everything held constant except the thing being tested. That lands around forty ads, which is roughly what it takes to find something genuinely better rather than confirm what you already run.
The constraint on that is production, which is a solvable problem, rather than budget, which usually was not the problem in the first place.
We set out the arithmetic behind forty separately.
If you want to see what this looks like on your account, we build the first batch free. We research the market, your customers and your competitors, then make 40+ static ads and test them. You get them in 24 hours.