Playable ad benchmark: methodology and results
The size figures on our homepage come from one sample game exported four ways. This page documents exactly how they were produced so you can reproduce them.
Before launch: replace every number on this page and the homepage with your own measured results, and publish the sample project so readers can rerun the test.
The sample game
A three-level match-3: 40 sprites in one 1024×1024 atlas, two audio clips (a 3-second match sound and a 6-second win jingle), tweened UI, an end-card with a store CTA. No physics, no particles. It was chosen because it is representative of the median casual playable, not because it favours any tool.
The four exports
| Export | Method | Final size | Time to first interactive frame (Slow 3G) |
|---|---|---|---|
| Hand-built HTML5 | Rewritten in vanilla JS by a contractor, assets inlined | 1.4 MB | 2.1 s |
| Stock Unity WebGL | Default WebGL build, Brotli, then inlined into one file | 5+ MB (fails limit) | 9+ s |
| Luna / Playworks | Default export settings, AppLovin target | ~2 MB | 3.4 s |
| PlayableForge | Default export settings, AppLovin target | ~0.8 MB | 1.3 s |
How to measure
- Export each build targeting the same network (AppLovin, single HTML) so packaging is identical.
- Record the on-disk size of the final file.
- Open it in Chrome on Android with DevTools throttling set to Slow 3G and record the time until the first frame that accepts input.
- Repeat three times and take the median.
What is not measured
Conversion rate, install rate and auction win rate depend on creative and audience and are outside this test. The claim here is narrower: smaller files load faster, and faster-loading playables get more completed plays.
Skip the manual work. PlayableForge exports a validated build for this network (and 20 others) from your Unity project in one click. Get early access.