Testing budget split: size it from what you can read

Decide the impressions one new creative needs before you would act on its result. That is a sample size question, not a budget one.

Multiply by your cost per thousand impressions. That gives the spend floor per creative.

Multiply that by how many creatives you intend to test this month. The result is your testing budget.

Whatever remains goes to the creatives already carrying results, which is the scaling half.

If the testing figure comes out above what you can afford, reduce the number of creatives rather than the spend per creative.

Testing ten creatives at half the readable spend teaches you nothing. It is the most common way a testing budget gets wasted.

The percentage falls out at the end. Quote it if somebody asks, but never start from it.

Two inputs, three multiplications, one number. The whole exercise takes ten minutes with a calculator.

A creative that did not get enough impressions has not lost. It has not been measured.

Killing it produces a false negative. Keeping it produces a decision made on noise.

Both feel like activity and both leave you where you started, minus the money.

The sample size approximation for comparing two proportions is 16 times p times one minus p, divided by the difference squared.

Use your own baseline rate as p and the gap you want to detect as the difference. That gives the impressions you actually need.

Fewer creatives, properly funded, beats more creatives read badly. Every single month.

That is the one place where cutting your creative count is the right call, and it is a media decision rather than a production one.

Published benchmarks put the winner share at roughly 5 to 8 percent, from Motion's analysis of 550,000+ Meta ads.

So thirteen to twenty properly funded tests sit behind each winner you expect to find.

If your testing budget funds four reads a month, expect a winner somewhere between month three and month five.

If it funds fifteen, expect one most months. That is the difference between a process and a hope.

That timeline, rather than a percentage, is what should be agreed with whoever owns the budget.

It also sets expectations honestly, which matters more in month two than any split does.

A finished ad here costs about 113 credits. That is 100 for the batch of a source take up to three minutes, and roughly 13 to export forty seconds.

Fifteen tested creatives is about 1,695 credits. That sits inside Basic at $39.99 for 2,500 credits a month.

So the production line is a rounding error against a media budget that can fund fifteen reads.

That is worth stating plainly. Teams often cut testing volume to save production cost and end up saving almost nothing.

Rendered presenters are the exception at 320 credits a minute, where fifteen attempts is over 5,600 credits.

Keep those on a separate line so the decision to use them stays visible against the testing plan.

On the filmed route the whole testing half converts into attempts, which is the shape you want when you are hunting for a winner.

That is why the argument in a planning meeting should be about impressions per read, and almost never about production.

You have no winner. Everything is testing until something works, and arguing about a percentage is premature.

One creative is carrying the whole account. The risk is concentration, and the answer is to test more regardless of the split.

The offer is the constraint. A testing budget cannot find a winner among variations of a proposition nobody wants.

Your consideration cycle is longer than your reporting window, so the readings arrive after the decisions.

The placement you need is horizontal, since every export here is 9:16.

Check those five before optimising a ratio. Four of them make the ratio irrelevant.

In all five, the same move helps: put as much as you can afford into finding the first thing that works.

That is affordable here in a way it usually is not. An attempt costs about 113 credits, and the fifteenth costs what the first one did.

Then come back to the split with a winner in hand, write the two inputs next to the answer, and it takes ten minutes instead of a meeting.

A split with its workings attached survives a change of budget. A percentage without them gets defended long after it stopped being right.