Review miningMetadata optimisation - PART I
What is in this resource:
How to build your ASO strategy from end to end
Learn to pick target keywords, write metadata, follow rank and base every decision on data.
From data to action
Simplify the keyword data and stay in line with App Store rules.
Tools that lift rank
Carry the keywords drawn from competitor reviews into your metadata and take your app out of the cusp band into the top 10.
What a rank is and how it is measured
A rank on its own says nothing. Its meaning comes from what you put beside it: popularity, history, a control group, and the noise the measurement makes by itself.
Rank is where your app appears when a given phrase is searched in a given country. The definition is simple, but the details of the measurement change what the answer means.
The measurement is one search request and the first 200 results are read. If your app is not in those 200 the rank is unknown. We write that as "outside 200"; we do not invent 201 or 999, because an invented number becomes sortable and ruins comparisons.
The measurement runs once a day. Measuring more often does not produce better data: the same phrase can move a few places within a day and that movement is the measurement's noise, not the market's. Once a day, stacked up, makes a trend visible - and a trend is clean of noise.
The measurement runs per phrase, not per row. Every app tracking the same phrase in the same country is read out of one search. That looks like an efficiency detail but it changes the result: your rank and your rivals' are read at exactly the same moment from the same result list, which makes them comparable.
And the result is stored. Without rank history no analysis is possible at all; one day's number is a photograph, thirty stacked days are a story.
What a rank does not say on its own
Being 7th for a phrase looks good. It may be, it may not. Four things have to sit beside the number before it can be decided on.
The first is popularity. Being first for a phrase nobody searches brings nothing. Visibility is rank multiplied by popularity; being first for a lightly searched phrase reaches fewer people than being fifteenth for a heavily searched one.
The second is country. The same phrase can be 7th in one storefront and 84th in another. Merging countries and taking an average makes you decide wrongly in both.
The third is history. 7th is a loss if last month it was 3rd, and a gain if last month it was 24th. The same number carries two opposite meanings and only the history says which.
The fourth is the control group. In a week when every phrase fell three places, yours falling three places is not a signal, it is the market's movement. Without a control group no movement can be interpreted.
Read together, those four turn a rank into a decision instrument. On its own it is only a number.
- Without popularity, a rank does not say how many people it reaches.
- Without a country, it does not say what to do in which market.
- Without history, it does not say whether it is a gain or a loss.
- Without a control group, it does not say whether the movement came from you.
Telling noise from signal
Ranks move a little every day and most of that movement is meaningless. Separating what is not is the most practical skill in rank analysis.
The first rule: do not interpret a single day. Two places in a day happens to almost every phrase and the cause is usually small differences in measurement conditions. Drawing a conclusion from it is like tossing a coin and building a strategy.
The second rule: three days running in the same direction is a signal. If the direction is consistent something is behind it - your move, a rival's move, or the market shifting.
The third rule: read the size of a movement against the rank itself. One place inside the top 10 is a large event; ten places at 150 is almost nothing. Density is higher at the top, so the same distance is harder to cover.
The fourth rule: movement seen across many phrases at once reads differently from movement on one. If they all move together the cause is your metadata or your score; if one phrase diverges the cause is specific to it.
The fifth rule: mark the days a measurement was missed. A gap makes the difference between two measurements larger than it was and it looks sharper than it is.
| What you see | How to read it |
|---|---|
| Two places in a day | Noise. Do not interpret. |
| 1-2 places in one direction for three days | Signal. Look for the cause. |
| One place inside the top 10 | A large move; density is high there. |
| Ten places at 150 | A small move; distance is cheap down there. |
| Movement across every phrase at once | Metadata or score - not specific to a phrase. |
Measuring the effect of a move
This is what rank analysis is really for: knowing what a change you made actually did. The order below is the only way to answer that reliably.
- Freeze the starting pointBefore the change, record the rank of every phrase on your tracking list - not just the one you are about to touch. The ones you do not touch will be your control group.
- Change one thingOne field per release. Change the app name and the subtitle together and you cannot separate which one carried, so you will not know what to repeat next time.
- Wait for the index to settleApple's index does not update immediately; arrival usually takes a few days and longer in some countries. A second change inside that window mixes the two moves together.
- Subtract the control group's movementIf the phrases you did not touch fell three places on average, your phrase falling three places is not the move's effect. Subtract that average when computing the difference.
- Write down what is left, and record itWhat remains is the measurable effect of the move. If you won, apply the same kind of move to the next phrase; if you lost, undo it and check a week later whether the old rank returns.
Reading accumulated history
A few months of accumulated rank history gives you what individual measurements cannot. Four ways of reading it are useful.
The first is move markers. Mark your metadata changes on top of the history and it becomes visible which move carried which phrase how far. That can be the most valuable document your team has.
The second is seasonality. A phrase's popularity is not flat across the year, and when popularity rises so does the number of apps entering it. Comparing against the same period last year gives a truer answer than comparing against last month.
The third is overlap with rival movement. Which rival rose in the weeks you fell? If there is overlap you can plan the next move by reading that rival's metadata.
The fourth is silent losses. Phrases falling from the top 10 into the band usually go unnoticed, because nobody looks at something already won. Searching the history for those drops finds losses that can still be undone.
There is one condition for history to hold: the measurement has to keep running the same way. If the method changes, two periods stop being comparable.
Things that move a rank without metadata
Rank does not depend on metadata alone. Your ranks change without you touching anything, and knowing the causes stops you from fiddling with metadata for nothing.
Rating score and count. Between two apps competing for the same phrase, the stronger one moves ahead, and the heaviest component of strength is ratings. A score drop after a release weakens your rank across every phrase at once.
App size and freshness. Apps not updated for a long time gradually slip. That is less a direct penalty than a relative decline caused by rivals updating.
Category choice. Which category an app sits in affects which phrases it is associated with. Changing category is a large move and it affects every rank.
Removal and re-publication. An app going off sale in a country loses every rank there. When that happens to one of your rivals, your ranks improve on their own.
Changes on Apple's side. How often the index refreshes and how heavily each field counts change from time to time. Those changes are not announced; they are only noticed by seeing many apps move similarly on the same day.
What those five have in common is that none of them can be answered with metadata. When your rank falls, looking at them first saves you an unnecessary move and the measurement pollution it causes.
- A score drop weakens every phrase at once, not just one.
- An app not updated for a long time declines relatively.
- Changing category affects every rank; it is not a small move.
- A rival being pulled is not your gain and will not repeat.
- Sudden movement across the board is usually Apple's side; waiting is the right answer.
The one output an analysis should produce
A rank analysis can generate any number of charts. For it to count as useful it needs exactly one output: which phrase do you touch this week?
An analysis that cannot answer that question stays a report no matter how detailed it is. Reports get read, admired, and forgotten.
Producing the answer takes three filters. First: phrases between 11 and 30. Second: the ones whose popularity could be measured. Third: the ones whose strength threshold is reachable. The phrase at the top of what remains is this week's work.
A second output is acceptable too: touching nothing this week. If the control group is moving as a block, if the previous move's index has not settled, or if a rating drop is under way, the right decision is to wait. Waiting is a decision and it can come out of an analysis.
The one output that is not acceptable is "we need to improve our rankings". That sentence is not the result of an analysis, it is the state before one. The analysis exists to reduce that sentence to one phrase and one field.
Frequently asked
- Why do my ranks move every day?
- Partly rivals moving, partly the measurement's own noise. Two places in a day should not be interpreted; three days in the same direction is a signal.
- What is my rank if I am not in the first 200?
- Unknown. We write "outside 200" and show a short dash. Writing 201 would make that number sortable and ruin the comparisons built on it.
- Would measuring several times a day be better?
- No. Intraday movement is the measurement's noise rather than the market's; measuring more often raises the noise, not the information.
- When will I see the effect of my move?
- Usually within a few days, later in some countries. Do not draw a conclusion before a week has passed; saying "it did not work" after two days means the measurement was not made.
- Every one of my phrases dropped at once. What happened?
- If you did not touch metadata, look at the rating curve first - the score may have fallen after a release. If not, check whether many apps moved the same day; it may be a change on Apple's side, and waiting is the right answer.
Explore our resources
The turning point in ASO: a guide to raising visibility and staying in the race
Ready to rethink ASO from the start? Our new guide explains step by step the road from rank to download. We show how to find the keywords in the cusp band, which phrases to draw out of competitor reviews and how to build the 100-character field. Download the ASO guide now and pass the app ahead of you in search!
See resource
Metadata optimisation - Part II
When an app falls outside the top 10 on the App Store, it loses most of the visibility that comes from search. The real gain is hidden in the 11-30 cusp band: the words only one step from the first page. This guide explains step by step how to find that band and carry it up.
What is inside?
- A clear frame for how the cusp band is measured.
- The method for drawing keywords out of the user reviews of competitor apps.
- An analysis of the mistakes most often made when building the 100-character keyword field.
See resource
The complete guide to optimising your app title
Most apps lose downloads not because they do not appear in App Store search, but because they appear on the wrong words. Our new report, Finding Keywords with Review Mining: Methods and Application Strategies, explains how keywords are drawn out of the source nobody looks at.
Discover the ways to pick meaningful phrases out of the user reviews of competitor apps, prioritise them by country and place them in the 100-character keyword field. Learn which signals to look at while carrying cusp band keywords into the top 10.
See resource