Can autocomplete predict App Store popularity? We measured it
· Jovain Chisholm
Not well enough to show as a number. On popular terms, autocomplete depth matched Apple's popularity with a rank correlation of 0.21, no better than word count. On niche terms the best signal reached 0.31. Across all 300 terms it reached 0.57, but that only tells popular terms from niche ones.
The idea
When you type into the App Store search box, it suggests completions. A common idea is that the fewer letters a term needs before it shows up, the more people search for it. If that held, autocomplete could fill the gap where Apple gives no popularity score.
How we tested it
We took Apple Ads popularity as the truth and measured how well each signal ranked terms in the same order. We used Spearman rank correlation: 1 means the same order, 0 means no relation. We set 0.7 as the bar for showing a number to users.
Popular terms: 0.21
On 150 terms Apple scores, autocomplete prefix depth reached 0.21. Term length alone scored the same. Autocomplete added nothing over counting letters.
Niche terms: 0.31 at best
On 150 niche terms, 95 of which appeared in App Store suggestions, the results were:
| Signal | Rank correlation |
|---|---|
| Blend, 70% App Store autocomplete | 0.31 |
| App Store autocomplete alone | 0.27 |
| Google autocomplete | 0.24 |
| Ratings of the top 10 apps | -0.14 |
Ratings of the top 10 apps, a common stand-in for volume, ran slightly backwards on niche terms. On popular terms it reached 0.16.
All 300 terms: 0.57, and why that misleads
Put both groups together and prefix depth reaches 0.57. That looks close to useful. It is not. The score only splits popular terms from niche ones, which you already know from whether Apple scores the term. Inside each group, it cannot tell you which term is bigger.
What we do instead
None of the signals came near 0.7, so Honest ASO shows no invented popularity. You get Apple's number or a ceiling, and difficulty from the live top 10. Try it on your own keywords with the free keyword checker.