Keyword research is the part of shipping an app that most developers do once, at launch, and then never revisit. The listing goes live, the keyword field gets filled in from memory, and the whole thing is left alone until downloads flatten out. These are the mistakes that come up most often, and what to do instead.
1. Chasing head terms you cannot win
“Fitness” and “budget” feel like the obvious targets because they have enormous search volume. They are also defended by apps with years of install velocity and review counts you are not going to match this quarter. Ranking 180th for a huge term sends you nothing.
Popularity is only half the picture. A term is worth chasing when demand is meaningful and the apps currently occupying the top ten are beatable. A less popular phrase you can realistically reach the first page for will out-earn a famous one you cannot.
2. Ignoring the title, which is the strongest field you have
The app name carries more ranking weight than anything else you control, and it is the field developers are most reluctant to touch because it feels like branding. You do not have to rename the product. A short, honest descriptor after the name captures real terms without damaging the brand.
3. Repeating words across fields
Words already in your title do not need to reappear in the keyword field. The keyword field is finite, and every character spent duplicating something you already rank on is a character not spent on a term you do not cover yet. The same applies to your subtitle.
4. Treating one storefront as all of them
The phrasing that works in the US is frequently not the phrasing people use in the UK, Germany or Japan, and competition differs wildly by region. A keyword set copied unchanged across every storefront is usually optimised for exactly one of them and mediocre in the rest.
5. Checking rank instead of tracking it
This is the one that quietly does the most damage. Searching for your own keyword and noting where you land tells you one number, on one day, from one device, and personalised results make even that unreliable.
What you actually need is the shape of the line: whether a position is trending up or down, and whether it moved when you changed your metadata. That requires the same measurement taken the same way every day. A spreadsheet updated whenever you remember cannot give you that, which is the specific problem SonarTree was built to solve, but the principle holds regardless of what you use to do it.
Where to start
- List the terms you think you rank for, then check what you actually rank for. The gap is usually surprising.
- Score each candidate on demand against how beatable the current top ten looks, not on demand alone.
- Change one thing at a time, and give it a couple of weeks before judging it.