The resource centre

The difference between a guide and an article

Guides are not texts you read and put down. Each one describes a job from start to finish and leaves you something you can use: a tracking list, a rewritten metadata, a week's plan.

An article explains a subject. A guide gets a job done. The difference is measured by what you hold at the end: finish an article and you have understood something; finish a guide and you have done something.

That distinction changes the structure completely. A guide has to be read in order, because step three uses step two's output. Every step ends in something checkable: a file, a list, a number. A guide contains no step you cannot perform; if a step needs another tool, it says so at the start of the step.

Every guide here follows the same skeleton. First it says what you will measure and which starting point you have to record. Then it breaks the job into steps. At each step it says what to do, roughly how long it takes, and what the output should look like. At the end it says how to check whether the move worked.

Reading half a guide and applying it produces a worse outcome than not reading it. Most ASO changes can be undone, but they pollute the measurement window: when you undo a change, Apple's index takes time to return too, and nothing you measure in that period is clean.

What order to read them in

The guides were written to stand alone, but they have an order. If your app is new to the App Store, or you are looking at ASO systematically for the first time, this order wastes the least time.

Keyword selectionEverything starts with a list. Without knowing which phrases you will track there is nothing to measure. This guide covers widening the pool from five channels and then cutting it down to something measurable. Output: a tracking list.
The cusp bandOnce the list is measured, the phrases you rank 11-30 for appear. This guide covers ordering that band and choosing which phrase to start with. Output: a priority order.
Metadata optimisationThe first item on that priority order has to be turned into metadata. This guide covers planning the app name, subtitle and keyword field together, and clearing the repeats and spaces. Output: rewritten metadata.
Rival trackingAfter your own move, keeping the measurement clean means knowing what your rivals did. This guide covers choosing which apps to track and separating their movement from yours. Output: a rival list and a weekly check.
Return on timeThis is the guide to leave for last, because it only means anything once a few weeks of measurement have accumulated. It covers how to compute what your time bought and which moves are worth repeating. Output: a decision about which work to stop.

The four mistakes made while applying one

Because the steps are simple they look skippable. Skipped, the guide stops working and the reason is hard to find. Here are the four most common departures.

The first: not recording the starting point. If you do not record your current ranks before changing metadata, you cannot know what changed afterwards. A feeling that "it used to be better" is not a measurement.

The second: changing more than one thing at once. Change the app name, the subtitle and the keyword field in the same release and when the rank rises you will not know which one carried. You will not know what to repeat next time either, so the learning stops.

The third: expecting a result too early. Apple's index does not update immediately. A metadata change usually takes a few days to reach the ranking, longer in some countries. Looking two days later and saying "it did not work" means the measurement was never made.

The fourth: forgetting the control group. The phrases you did not touch are your control group. If they all went down together, your phrase going down is not the move failing - it is the market moving.

Record the ranks before the change; that is the only thing you will compare against.
Change one field per release; it is the only way to learn which move carried.
Wait at least a week; two days of data is noise.
Measure the phrases you did not touch too; without them the result cannot be interpreted.

The measurement discipline underneath

A guide's steps only mean something if the measurement beneath them is reliable. Every guide here was written on the same measurement rules, and knowing them explains why the steps are shaped as they are.

Rank is measured once a day. Measuring more often does not produce better data, because the same phrase can move a few places within a day and that movement belongs to the measurement, not the market. Once a day, stacked up, shows the trend - and a trend is clean of noise.

Measurement runs per keyword, not per row. Every app tracking the same phrase in the same country is read out of one search. That makes rival tracking practically free: a rival's rank for a phrase you track comes out of the search you were making anyway.

Phrases are normalised before they are stored, and normalisation depends on the country. In Turkish and Azerbaijani a capital I becomes a dotless ı when lower-cased; everywhere else a dotted i. Fold the same phrase one way in one place and another way elsewhere and you get two phrases, both wrong.

An unmeasurable value is left blank. A phrase whose popularity could not be measured carries a short dash, not a zero. When a guide says "pick the three most popular phrases", the blank ones are not in the running - and they are not counted as zero and pushed to the bottom, because nobody measured them to be zero.

The week-long loop

Guides look like one-off texts, but their real value shows when they are repeated. The loop below is the shared skeleton of every guide here.

Day 1 - measure and freezeRecord the rank of every phrase on your tracking list. That is this week's comparison point. Record your rivals' ranks the same day, because their movement will affect your measurement.
Day 2 - pull the band and chooseOrder the 11-30 phrases by popularity. Look at the top three and pick one. Picking all three is tempting and it breaks the measurement.
Day 3 - one moveMake one metadata change for the phrase you chose and ship it. Note the change somewhere: which field, its old form, its new form, the date. Next week that note will be your only evidence.
Days 4-6 - do not touch itDo not touch the metadata for three days. The index has to settle. The work available in that window is widening the pool for next week's candidate.
Day 7 - compare and decideMeasure the same phrases again. Subtract the control group's movement and write what is left down to the move. If you won, apply the same kind of move to the next phrase; if you lost, undo it and note why.

What the guides do not promise

For a guide to work it has to be realistic. Here are the promises deliberately not made - promises made freely elsewhere.

No guide says you will reach a particular rank. Apple's ranking algorithm is not published and what your rivals will do is not in your control. A guide describes a move's cost and its odds, not its outcome.

No guide says "this many installs in this many days". Install count depends on the product itself, the screenshots, the price and everything else in the market that week. We can measure rank; we cannot measure the whole path from rank to install.

No guide makes using a tool compulsory. The methods described can be applied by hand; it is only repeating them daily across 175 countries that is impractical. The text is complete for somebody who wants to read it and do it manually.

And no guide sells a one-off move. ASO is not a project, it is maintenance work. Work at it for a week and stop, and the places you won are given back within a few months, because your rivals do not stop.

Who should read a guide, and who should not

These texts were written with one-person teams and small studios in mind. They consist of steps one person can apply in a week, and no step needs a separate budget.

If you work in a large team, we suggest one person reads a guide end to end before it is shared. Because the steps depend on one another, splitting them between people breaks the result: if one person writes the metadata and another does the measurement, the delay between them shifts the measurement window.

If your app is not live yet, reading a guide is still useful but applying it is not possible. Rank measurement needs a live app. The best work before launch is running the pool-widening part of the keyword selection guide and writing the first metadata against it.

If your app is live in many countries, we suggest starting a guide with one. Run the loop for a month in the storefront where you get the most downloads, learn how it behaves, then add the second. Starting in ten countries at once ends with measurement in none of them.

Situation	Where to start
Newly live, no data	Keyword selection guide, the pool-widening part
Ranks exist but are not tracked	The cusp band guide
Metadata unchanged for a long time	The metadata optimisation guide
Ranks moving inexplicably	The rival tracking guide
Time going in, return unclear	The return guide

The language they are written in, and why

These texts are written in the second person: do this, measure that, wait. That is not a stylistic preference, it is a control mechanism. Write a sentence as "it should be done" and who does it stays unclear, and work that stays unclear does not get done.

The second choice is that every step ends in a measurable output. If there is no list, number or file in your hands at the end of a step, the step is not complete. That lets you know where you are while reading; a half-finished step is noticed immediately.

The third is giving durations. Beside every step is roughly how long it takes. Those are upper bounds rather than averages: finishing sooner is fine, but taking longer means something has been misread.

The fourth is naming the tool where it is needed and not where it is not. If a step can be done by hand, how to do it by hand is written. If it is impractical - daily measurement across 175 countries - that is said plainly. A guide exists to get the job done, not to sell a product.

The fifth is not giving numbers. No guide says "this move gains this many places on average". Building that sentence requires long measurement across many apps and that data has not accumulated. What is given instead is the mechanism: why the move might work, under what conditions it will not, and how to check.

The last choice is not hiding the exceptions. Where a rule has an exception it is written directly under the rule. That makes the texts longer, but the alternative is worse: a rule written without its exception misleads the reader who hits the exception, and that reader stops trusting the text.

Frequently asked

How long does a guide take to finish?: Half an hour to read, a week to apply. Most of the steps are short but you have to wait between them; the time Apple's index takes to settle cannot be shortened.
Do I have to read them in order?: Each one can be read alone, but the order exists for a reason: each guide uses the previous one's output as its input. You cannot apply the cusp band guide without a tracking list.
Can I use the same guide for more than one app?: Yes, but not at the same time. Making a move in two apps in the same week does not pollute either measurement; the real problem is your attention dividing and your notes getting mixed.
Can I do the step without your tool?: Every step can be done by hand. Measuring rank by hand is possible for one country and ten phrases, not for 175 countries and hundreds of phrases. The guides say so plainly at every step.
I applied a guide and my rank fell. What now?: Look at the control group first: if the phrases you did not touch also fell, it is not the move's fault. If only yours fell, undo the change, wait a week, and see whether the old rank returns. If it does, the move was wrong; if it does not, the cause is elsewhere.