Guide · Keyword research
Choosing keyword candidates before you have any data
Most keyword advice assumes you already have performance data: spend on one side, revenue on the other, a verdict in the middle. That is the easy half. The hard half is the first one, choosing which terms to put in your metadata at all, when nothing has happened yet and every candidate looks equally plausible on a list.
Start wider than feels sensible
Brainstorm 15 to 20 seeds: what your app does, what problem it removes, what someone would call it who does not know your brand exists. Then expand to 50 to 100 candidates before cutting anything. The width matters because the terms that pay are usually long-tail, and a shortlist of five will be five head terms you have no chance of ranking for.
Sources worth mining: App Store autocomplete, the terms competitors actually rank for, and, if you already run ads, the search terms that triggered your ads. That last one is the only list on this page made of words real buyers typed rather than words you imagined them typing.
The four factors, and where each number comes from
Score every candidate on four axes. None of them is decisive alone, which is exactly why people who use one of them get poor results.
1. Relevance
Does the term describe what your app actually does? This is the only factor with no number attached and the only one that can sink you outright. A mismatch does not just convert badly, it brings in people who install, discover you are not what they searched for, and leave, which costs you on every other axis at once.
2. Search popularity
Apple's own signal, on a 1 to 5 scale where 5 is the most searched. It comes from the Apple Ads API, which means it is Apple's estimate rather than a third-party guess, and it is the number Veldo shows on each keyword row. Anything at 1 is rarely worth the metadata space unless the phrase is unusually specific to what you sell.
3. Competition
Read the current top ten, not a difficulty score in isolation. If those positions belong to apps with large budgets and tens of thousands of ratings, the term is expensive in both metadata space and ad money, and your realistic ceiling is somewhere nobody scrolls to. A slightly narrower phrase with a weaker top ten is almost always the better trade.
4. Conversion intent
Does someone typing this have a payment method out? "Free photo filter" and "professional photo editor" can have similar volume and produce completely different businesses. Intent is the factor with the least data behind it and the largest effect on revenue, which is why it stays a judgement call until you have shipped and measured.
Popularity is not the same as opportunity
The most common mistake is ranking candidates by popularity alone. A popularity 5 term where you would sit 80th is worth less than a popularity 3 term where you can hold 4th, because the first is invisible and the second is not. Opportunity is where popularity, a beatable top ten, and genuine relevance overlap, and that overlap is usually two or three words long and describes a specific job rather than a category.
Placing what you picked
You have 30 characters of app name, 30 of subtitle, and a hidden 100 character keyword field. Highest-value terms go in the name, secondary ones go in the subtitle while it still reads like a sentence a human wrote, and the keyword field takes what is left. Do not repeat a word across fields; the index only needs it once and repetition wastes space you cannot get back. Our freekeyword field counter checks the limits and shows which of your tracked keywords the draft actually covers.
Then wait, and measure the right thing
Give any metadata change three to four weeks before judging it. The listing has to be re-indexed and then observed, and a reading taken after three days is noise you will act on and regret. When the data does arrive, the question stops being "which candidates look good" and becomes "which of these actually paid", which isa different job with a different method.
One thing to carry from selection into evaluation: keep the ones that hold their users, not the ones that produce the most installs. A term delivering 500 downloads a day that all churn inside a week is a term you are paying for twice, once in acquisition and again in support.Retention is where that shows up, several weeks after the install count made you feel good.
Common questions
How many keywords should I start with?
Brainstorm 15 to 20 seeds, then expand to 50 to 100 candidates before you cut. The point of the wide list is that the profitable terms are usually long-tail, and a list of five is all head terms you will never win.
What is a good search popularity score?
Apple's search popularity runs 1 to 5, where 5 is the most searched, and it is the number Veldo shows on each keyword. Treat 1 as barely worth the metadata space unless the term is unusually specific to what you sell, and remember that a 5 you rank 80th for is worth less than a 3 you rank 4th for.
Should I target high-volume keywords?
Only where you can realistically place. A popular term whose top ten is held by apps with six-figure budgets and tens of thousands of ratings will take your metadata space and give nothing back. The winnable middle is usually two or three words long and describes a specific job.
How long before I know whether a keyword worked?
Three to four weeks after the metadata change ships. The App Store has to re-index the listing and then watch how people respond, so anything you read in the first few days is noise.