Most new creatives fail. That is normal: even experienced teams find that only a small share of new concepts beat their current best. When one does win, it becomes the most valuable thing the team has. It also begins to fade from the day it launches, as the audience sees it repeatedly and the network's delivery saturates.
Iteration is how teams get the most from a winner: making variants that extend its life, find better versions and reveal what made it work. This guide sets out a systematic approach, and explains when to stop iterating and look for the next concept.
Concepts and iterations
A concept is a creative idea: a mechanic, a story, a promise. An iteration is a variant of a concept with one deliberate change. Healthy creative pipelines balance both. Too few new concepts and the team runs out of ideas when the winner fades. Too few iterations and every winner is retired before its best version is found. Many teams split their production between the two, adjusting the balance as results come in.
Step 1: understand why it wins
Before making variants, form a hypothesis about what makes the winner work. Look at more than the headline numbers:
- Interaction rate: is the hook unusually good at getting the first touch?
- Completion rate: do more viewers reach the end than with other creatives?
- Click and install rate: is the ending or the CTA unusually persuasive?
- Early retention and revenue: is it bringing better players, or just more of them?
- Placement and country: does it win everywhere, or only on one network or in one region?
The answer points to which element to iterate. A creative with a great hook and a weak ending should keep its hook and try new endings. Playable ad metrics explains what each number measures.
Step 2: change one thing at a time
Each iteration should change one element, so its result tells you something. A variant that changes the hook, the theme and the end card at once may win or lose, but you will not know why. The elements worth varying, roughly in order of their usual impact on playables:
- The hook: the opening scene and the first few seconds.
- The first interaction: the gesture, the difficulty and the first reward.
- The ending: a win, a near-miss, a choice or a reveal.
- The theme or setting: the same mechanic in a new world.
- The characters: who appears and how they react.
- The copy: the headline and the message angle.
- The CTA: wording, colour, size and timing.
- The pacing: overall length and how quickly the CTA appears.
Start with the elements your hypothesis points to. If you have no hypothesis, start at the top of the list, because the opening seconds decide whether anyone sees the rest.
Step 3: make variants in batches
Launch iterations in small batches, each against the original as a control. Three to five variants per batch is manageable for most teams: enough to give the network options, few enough that each gets meaningful delivery. Name them consistently so the results can be grouped; creative naming conventions shows how.
Keep the original running beside the variants. Comparing a variant this month against the original last month mixes the effect of the change with seasonal shifts in cost and audience.
Step 4: read the results carefully
Networks' delivery systems do not split traffic evenly. They push budget towards whichever creative they predict will perform, often within days. That means a variant can lose simply because the network never gave it a fair chance. Where you need a clean comparison, use the network's own split-testing feature or a dedicated test campaign with an even split. The A/B testing guide covers how.
- Wait for enough data. A variant with a handful of installs has not won or lost.
- Judge on the metric the change was meant to move: a new hook on interaction rate, a new ending on install rate.
- Check downstream quality: a variant that wins installs but loses retention has not won.
Step 5: compound the winners
When a variant wins, it becomes the new control. The next batch iterates on it. Over several rounds, the creative can improve significantly, each round adding one proven change. Record each winning change in the creative library; patterns across concepts, such as near-miss endings winning repeatedly, become principles for new concepts.
Iterations that extend a creative's life
Some iterations are not about finding a better version but about refreshing a fading one. When a winner shows signs of fatigue, such as a rising cost per install at constant spend, falling click rate or rising frequency, a light refresh can restore much of its performance:
- A new colour palette or background.
- A seasonal theme for a holiday or event.
- A new character or skin for the main piece.
- A different first board or level with the same mechanic.
- A localised version for a new market.
Creative fatigue explains how to spot fatigue and plan a refresh cadence.
When to stop iterating
Iteration has diminishing returns. Signs that a concept has given what it can:
- Several batches in a row produce no variant that beats the control.
- Refreshes restore performance for shorter periods each time.
- The winning changes get smaller and the remaining ideas feel cosmetic.
At that point, shift production towards new concepts, using what the iterations taught you. A concept that has been iterated thoroughly leaves behind a list of proven ingredients, the hooks, endings and themes that worked, and the next concept should start from them.
Making iteration cheap
Iteration only works if variants are fast to make. Playables built from scratch for each variant make the process slow and expensive. Building on a reusable structure, where the hook, the theme, the copy and the ending can each be swapped without rebuilding the rest, makes a batch of variants a day's work rather than a week's.
Playable Ads Maker lets you duplicate a project, change one element, such as the art, the copy, the colours or the end card, and export each variant for several networks. AI ad variations for A/B testing covers making single-variable variants with AI Lab.
Create playable end cards in minutes—no code required.
Open Playable Ads Maker