How to test playable ads: from one lucky winner to a steady pipeline

Most studios find one good playable and run it until it dies. A practical way to test concepts and variations, read the right numbers, and always have the next winner ready.

Here is a pattern we see all the time. A studio ships a playable. It does well, sometimes very well. They run it for months. Then the numbers slide, slowly at first and then fast, and there is nothing ready to replace it. The next playable takes six weeks, and it does not beat the old one.

The problem is not that the second playable was bad. It is that the studio was relying on luck instead of a testing system. This post is about building one, even if you are a small team.

Concepts and variations are different things

Most testing problems start here. People test "new playables" without saying what is actually new about them.

A concept is a different idea for the ad. A different moment of the game, a different hook, a different fantasy. A rescue scene and a speed challenge for the same puzzle game are two concepts.
A variation changes one thing inside a concept. A shorter level, a harder ending, a different end card, a seasonal theme, the first tap in a new place.

Concepts are how you find big wins. Variations are how you keep a winner alive and squeeze more out of it. You need both, and you need to know which one you are running, or you cannot learn anything from the result.

A useful rule: when a variation wins, you learn something small and certain. When a concept wins, you learn something big and you should go and build variations of it straight away.

How many to test at once

More than you think, fewer than the big publishers do.

Large studios run playables at a scale that is hard to copy, with whole teams producing them. A small studio does not need that. What it needs is a rhythm. A pattern that works for many of the studios we work with:

Start with two or three concepts. Different enough that a winner tells you something.
Give each one a fair budget. Enough impressions that the result is not noise. Cutting a concept after a day of spend is how good ideas die early.
Take the winner and build three or four variations. Change one thing in each.
Keep one new concept in every round. This is your insurance for when the winner fades.

That last point is the one people skip, and it is the one that saves them three months later.

Which numbers to trust

Clicks are the easiest number to look at and the least useful. A playable that is confusing can get lots of taps and very few installs. A playable that is too easy can get installs from people who never open the game again.

What we look at, in order:

IPM (installs per thousand impressions). The best single number for comparing playables on the same network. It tells you how good the ad is at turning a view into an install.
What happens after the install. Day 1 and day 7 retention, and early revenue or ad income if you have it. A playable that brings in players who stay is worth more than one with a higher IPM that brings in tourists.
CPI, carefully. Cost per install depends on your bids and the auction as much as on the creative. Use it to compare creatives inside the same campaign, not across campaigns.
Completion rate, if the network shows it. If most players drop off halfway, the playable is too long or the middle is confusing.

Compare playables on the same network, in the same campaign, over the same period. A playable that wins on AppLovin can lose on Meta, where it plays after a video and reaches a different audience. That is not a contradiction. It just means you are running two tests.

When a winner starts to fade

Every creative wears out. The same people see it again and again, and it stops working. How fast depends on your spend and your audience, but there are signs:

IPM drops week over week while spend stays the same.
CPI rises without any change on your side.
The network starts delivering it less.

Do not wait for the winner to die before you test the next one. When the first sign shows up, you should already have variations in the market. The cheapest refreshes are often small: a new theme, a different opening level, a new end card. Players react to "this looks new" more than you would expect.

Watch the category, not just your own ads

Your playables do not run in a vacuum. The other games in your category are buying the same players, and what they run changes what works for you.

If three competitors start running the same kind of hook, it will feel fresh for a few weeks and then everyone's version gets tired at once. If nobody in your category is showing a certain moment of gameplay, that can be your next concept. Looking at what is running around you each week is one of the cheapest testing tools there is.

A simple testing plan you can start this month

Week 1: Launch two or three concepts on your main network. Equal budgets.
Week 2: Read IPM and early retention. Pick a winner. Do not overthink a close result; run both a little longer.
Week 3: Launch three or four variations of the winner, plus one new concept.
Week 4 and on: Every two weeks, retire the weakest playable, add one variation and one concept. Write down what you changed and what happened. That written record is the most valuable thing this process produces.

How we help

The plan above has one obvious problem: it needs a steady supply of playables, and for most studios that is exactly what they do not have. That is the part we built MakePlayable for.

We build concepts and variations from your own game art, usually a new version in a day, and we deliver each one as a separate file for every network you buy on. We also watch your category every week, so the next concept we suggest is based on what is running around you right now, not on a guess. When you share your results with us, we keep them next to every version we have built, so each new playable starts from what actually worked for your game.

If you have a playable that is fading and nothing ready to replace it, send us your store link. We build one for you first, free, and you can put it in a test before we ever talk price.

Common questions

How long should I run a playable test?

Long enough to reach a stable IPM on a fair budget. For most studios that is several days to a week per round. Short tests on small budgets mostly measure luck.

Should I A/B test playables against video?

Yes, but run both. They often reach different people and do different jobs. Many studios find playables bring in better players, while video is cheaper to make in volume.

How often should I refresh a playable ad?

When IPM starts to slide at steady spend. For active campaigns, plan a new variation every two to four weeks and a new concept every month or so.

What should I change first in a variation?

The first few seconds. The opening decides whether the player engages at all, so it usually moves the numbers most. The end card is a close second.

New to playables? Start with our guide to playable ads in 2026, and keep the playable ad specs by network handy when you upload.