THE VERDICT · A REAL MEASUREMENT ISSUE, WITH LIMITS
One matching change can show up in several countries.
In my sample, 239 country observations became 153 groups when I matched identical title transitions across countries. But the difference was much larger on Google Play than on the App Store, and a few apps contributed much of the result.
My combined figures passed the initial target, but the stronger uncertainty check did not. I cannot claim that at least a quarter of title change records are repetitions everywhere.
See what I found ↓MY QUESTION
Is that one idea or nine?
Imagine your competitor report lists nine title changes for an app. It might look as if the team ran nine different tests. But what if each country recorded the same old title changing to the same new title?
Those are still nine useful country observations. What I cannot assume is that they represent nine separate decisions, ideas or experiments.
I compared consecutive observed titles for the same app in each country.
I looked for the same app in the same store with the same old title and new title in different countries.
I grouped matching changes first observed within 72 hours, without merging repeated changes in the same country.
A group is a matching observed transition. It does not prove that a team made one coordinated decision.
MY THEORY VS THE RESULT
The counts changed, but the stores looked different.
My theory
Matching repeated changes across countries would materially reduce the count of distinct title transitions.
What I found
There was a clear measurement issue in this sample. The combined result passed my initial target, but the uncertainty check and concentration in a few apps prevent a broad claim.
country observations → matching groups
Groups match identical observed title transitions. They do not prove the number of developer actions.
See the original result chart
Each bar uses the same scale, from zero to 165. A group keeps one matching title transition rather than counting each country again.
Across both stores, the count went from 239 country observations to 153 groups. That is a 36% reduction within this sample, not a verified count of developer decisions.
Five Google Play app identities contributed 47 of the 86 additional matching country observations. Removing those five lowered the combined reduction to about 21%. This result is sensitive to which apps are included.
REAL APPS IN MY SAMPLE
What did repeated observations look like?
These examples come from archived observations in 2026. I am not claiming to know the reason for any edit, or suggesting these apps are my clients.
GOOGLE PLAY · NINE COUNTRIES
Grok
The source recorded “Grok AI” becoming “Grok” in nine countries, with new titles first observed over about 23 hours. I counted nine country observations but one matching group.
GOOGLE PLAY · NINE COUNTRIES
Grok changed back later
The source later recorded “Grok” becoming “Grok AI” in nine countries. This was a separate group because the direction and timing differed.
APP STORE · THREE COUNTRIES
Amaze GO!
The archive recorded “Amaze GO!” becoming “Amaze GO!®” in three countries. The added symbol is a real title difference, but it does not tell me whether anyone was testing keywords.
The two Grok transitions produced 18 country observations and two matching groups under my rules. Matching text does not prove one shared rollout.
DID MY IDEA PASS?
The answer is mixed.
I set a target of at least 25% repeated observations. The observed combined reduction was 36% overall and 31% in the later period. But when the supplied analysis resampled apps, the later range included values below 25%. Removing the largest contributors also brought the reduction below the target.
So I can say repeated country records mattered in this dataset, particularly on Google Play. I cannot use it to promise that competitor reports generally shrink by a fixed percentage.
MY ASO TAKEAWAY
Count countries and ideas separately.
My ASO tip: When a competitor monitor reports many title changes, check whether the old and new wording is identical across markets.
My ASO strategy: Keep two views: one showing where changes were observed, and another grouping matching title transitions. Record your grouping rules and keep different localized wording separate.
My ASO warning: A title change is not automatically an ASO test. Branding, legal updates, corrections or other reasons may explain it, and this study did not measure ranking or conversion effects.
WHAT I COULD NOT PROVE
Here is where the evidence stops.
- Developer intent: I could not see whether changes came from one rollout or separate decisions.
- Every market: The archive is not a census of all app listings or title changes.
- Every language: Country labels do not prove separate language treatments, and matching text can reflect a shared language version.
- Results: I did not test whether a change improved keyword rankings, installs or conversion.
- Timing: The records bound when a title changed; they are not developer edit logs.
SOURCES & METHODS
Want to check my work?
I used AppOrbit’s archived metadata. I reran the supplied verifier to reconstruct the title transitions and group memberships. It checks calculations against the archive, not the developers’ decisions or the original collection.
The source evidence and full technical report are kept locally while third party redistribution rights are reviewed.
Read the original report and technical details
ASO RESEARCH · EXPERIMENT 6 · 27 OCTOBER 2026
Nine countries can repeat the same app-title change.
In our Google Play sample, 165 country-level title-change observations became 85 groups after matching repeated changes across countries. Apple showed much less repetition.
The measurement issue is real in this sample. Counting each country separately can make repeated title changes look like more distinct activity. The size of the effect depends heavily on the store and the apps included.The ASO question
Imagine an ASO competitor report shows nine title changes for one app. That sounds like a busy testing schedule.
But what if those nine records all show the same change, observed in nine countries?
Those are nine legitimate country observations. They are not evidence of nine different title ideas, nine separate decisions or nine A/B tests.
Our prediction was that at least 25% of the country-level change observations would repeat a matching transition seen in another country. This is an audit of how ASO activity is counted, not a test of whether title edits improve rankings.
What we counted
We looked for changes between an app’s previously observed title and its next observed title in the same country. To keep timing reasonably tight, the old and new records had to be no more than 48 hours apart.
We then grouped records only when they had:
- The same app ID in the same store.
- The same old title and the same new title.
- Different countries.
- First observations of the new title within 72 hours of the group’s first observation.
We preserved punctuation, capitalization and language. A different localized title was not treated as the same text change. Repeated changes within one country were not merged.
The result
| Store | Country-level observations | Groups after matching | Additional matching country observations | Reduction in count |
|---|---|---|---|---|
| Apple App Store | 74 | 68 | 6 | 8.1% |
| Google Play | 165 | 85 | 80 | 48.5% |
| Combined sample | 239 | 153 | 86 | 36.0% |
Google Play: the original count of 165 was almost twice the grouped count of 85. Eighty observations repeated a matching transition already represented in a group.
Apple: the count changed from 74 to 68. Only six additional country observations were grouped away.
Combined: 239 observations became 153 groups, a reduction of 86, or 36%.
“Group” means a matching observed title transition across countries within the defined window. It does not mean we verified one underlying release, decision or experiment.
A real example: Grok
On Google Play, the source recorded “Grok AI” changing to “Grok” in nine countries: Brazil, Germany, the United Kingdom, Indonesia, India, Japan, South Korea, the United States and Vietnam.
The first observations of the new title were spread across about 22.7 hours. Counting by country produced nine observations. Matching the identical old-to-new text produced one group.
Later, the source recorded the reverse change, “Grok” to “Grok AI”, again across nine countries, within about 19.6 hours. That became a separate group because the direction and dates were different.
For these two transitions, the country-level count was 18. The grouped count was 2. We did not collapse all of the app’s history into one change.
Examples from both stores
These examples are the largest country groups within each store and date period, with ties resolved by the earliest observation. They illustrate repetition; they were not selected to calculate the totals.
| Store | Old title | New title | Countries | Observation spread | First observed |
|---|---|---|---|---|---|
| Apple App Store | Songless: Guess the Song! | Songless.fm ~ Guess the Song! | 3 | 6.3 hours | 2026-10-06 |
| Apple App Store | Amaze GO! | Amaze GO!® | 3 | 8.6 hours | 2026-10-20 |
| Google Play | Grok AI | Grok | 9 | 22.7 hours | 2026-09-22 |
| Google Play | Grok | Grok AI | 9 | 19.6 hours | 2026-10-21 |
The Amaze GO example adds a trademark symbol. It is a title change, but not evidence of a new keyword strategy. The dataset does not tell us the purpose of any change.
Did the pattern repeat on later dates?
We applied the same matching rules separately to the earlier and later periods:
| Period | Country-level observations | Groups after matching | Reduction |
|---|---|---|---|
| Through October 7 | 104 | 62 | 40.4% |
| October 8 to October 26 | 135 | 93 | 31.1% |
The combined later-date reduction was 31%, above the proposed 25% threshold. However, the stores again differed substantially:
| Later-date sample | Country-level observations | Groups after matching | Additional matching observations | Reduction |
|---|---|---|---|---|
| Apple App Store | 49 | 46 | 3 | 6.1% |
| Google Play | 86 | 47 | 39 | 45.3% |
| Combined sample | 135 | 93 | 42 | 31.1% |
On later dates, Google Play fell from 86 observations to 47 groups, a 45% reduction. Apple fell from 49 to 46, a 6% reduction.
The early and later group counts add to 155, whereas grouping the whole period gives 153. Two additional groups combine across the period boundary when the whole sample is processed together. Period results were calculated independently so they do not borrow matches from outside their dates.
Was the 72-hour window too generous?
No change appeared when we tightened it to 24 hours. Every multi-country group in the main analysis fit within 24 hours already. The largest observation spread was about 23.0 hours.
Requiring the countries’ old-title/new-title intervals to share a possible editing time also left the count unchanged. That makes the observations compatible with a common editing time; it does not establish when, how or why the update happened.
Five apps account for much of the result
This is the biggest qualification to the headline. Five Google Play app IDs contributed 47 of the 86 additional matching observations, about 55% of the combined total:
| App or observed title family | Additional matching country observations |
|---|---|
| Grok | 16 |
| Game Booster / Game Launcher | 12 |
| Duolingo | 8 |
| Amaze GO | 6 |
| Hole.io | 5 |
After removing those five app IDs, the combined reduction fell to 21% overall and 19% on later dates. Both are below our 25% target.
So the finding is not that every competitor report will shrink by a third. The finding is that repeated country observations can materially change activity counts, especially when a few apps update the same title across several markets.
What this means for ASO research
Country-level counts answer: “In how many country records did we observe a change?”
Grouped counts answer: “How many matching old-to-new title transitions did we observe within this time window?”
Both are useful. The problem is interpreting the first count as evidence of distinct ASO ideas or experiments.
If you are studying how often a competitor changes its title, checking for repeated transitions across countries gives a clearer picture. If you are checking where a change appeared, the separate country observations are exactly what you need.
Publishable finding: “Matching repeated app-title transitions across countries reduced 239 observed changes to 153 groups in this sample. The reduction was much larger on Google Play than on Apple and was concentrated in a small number of apps. Country-level observations should not automatically be interpreted as distinct ASO experiments.”Did the theory pass?
The observed combined percentages passed the 25% target: 36% overall and 31% on later dates.
The stronger evidence requirement did not pass. When we resampled apps, the later combined reduction ranged from about 18% to 43% in the middle 95% of results. That range includes values below 25%. Removing the biggest contributors also moved the result below the target.
The evidence supports a measurement problem in this sample, particularly on Google Play. It does not support a universal claim that at least a quarter of all ASO title-change records are repetitions.
Open the detailed methods, checks and limits
Data and scope
We examined 244,883 archived metadata observations across 19,360 store-country-app combinations. The archive spans September 19 to October 26, 2026 and was collected alongside app-chart tracking. It is not a complete census of app-store edits, and this title analysis is not restricted to the free-app chart cohorts used in experiments 4 and 5.
There were 307 observed normalized title transitions. Sixty-eight had more than 48 hours between the last old-title observation and the first new-title observation, so the primary sample contained 239 transitions involving 118 distinct store-app IDs. Apple and Google Play IDs remain separate, even if they represent the same brand.
The first record for an app-country combination was not counted as a change. Blank titles or conflicting titles at an identical timestamp would interrupt continuity; the audit found no such interruptions in these name records. We compared consecutive observations, not the number of metadata records, because many records change for reasons unrelated to the title.
Matching rules
We normalized Unicode representation and collapsed whitespace while retaining case, punctuation and language. The raw-string check produced the same counts. Identical old/new pairs were grouped within each store-app ID, ordered by the first-new-title timestamp and then country.
An observation joined the earliest eligible group only if its new-title timestamp fell within 72 hours of that group’s first timestamp and its country was not already in the group. We did not chain a long sequence of individually close observations into a group spanning more than 72 hours. The same transition recurring in one country remained separate.
The common-edit-time check also required the latest old-title observation across a group to precede the earliest new-title observation. All primary groups met that condition. Observed timestamps bound possible changes; they are not developer edit logs.
Different percentages answer different questions
The main reduction is (239 minus 153) divided by 239, or 36.0%. It counts the additional observations after keeping one representative per group.
The original count is 56.2% larger than the grouped count because (239 minus 153) divided by 153 uses a different denominator. These statements are consistent; 36% reduction and 56% excess are not interchangeable labels.
There were 31 multi-country groups containing 117 observations. That includes each group’s first representative, so it is not the same as the 86 additional observations. We did not label all 117 as extra repetitions.
Alternative checks
| Check | Overall observations | Overall reduction | Later observations | Later reduction |
|---|---|---|---|---|
| Group within 24 hours instead of 72 | 239 | 36.0% | 135 | 31.1% |
| Require old/new records within 24 hours | 115 | 33.0% | 71 | 26.8% |
| Keep changes with longer observation gaps | 307 | 32.2% | 182 | 27.5% |
| Compare exact raw titles | 239 | 36.0% | 135 | 31.1% |
| Require a shared possible editing time | 239 | 36.0% | 135 | 31.1% |
| Remove the five biggest repeat contributors | 183 | 21.3% | 108 | 18.5% |
The broader-gap sample includes changes whose timing is less precisely observed. The tighter-gap sample reduces uncertainty about observation timing but removes many records. The top-five exclusion uses the five store-app IDs contributing most additional country observations in the full primary sample, as specified in the analysis rules.
Uncertainty
We resampled store-app IDs 5,000 times, retaining all country events and group contributions for each selected app. The middle 95% of later-period reductions was:
| Later-date sample | Lower end | Upper end |
|---|---|---|
| Apple App Store | 0.0% | 13.5% |
| Google Play | 27.9% | 57.0% |
| Combined sample | 18.1% | 42.7% |
These ranges reflect sensitivity to which observed apps are included. They do not account for every source of uncertainty, such as missing markets, shared timing effects or collection errors. The pooled sample weights each country transition equally rather than giving each store or app equal weight.
The analysis rules were written locally before the experiment’s outcomes were calculated. They were not externally preregistered. The dataset had already been used for other studies. The 25% target was a chosen prediction, not an industry standard.
Verification
A separate script reconstructed all 307 transitions from the archived metadata, verified the 239 primary transitions and independently rebuilt the group memberships. It used a global chronological grouping pass rather than the analysis script’s per-transition pass.
It checked all 72-hour, 24-hour, shared-edit-time and later-period group memberships, then recalculated every store, period and sensitivity total. All checks passed. This verifies calculations against the source, not the developers’ actions or the original collection process.
Country labels are not separate language experiments
App listings can be localized by language, and a language version can be visible in multiple countries. Google’s listing API represents a localized listing with language and title fields. Apple’s localization documentation describes how language settings, supported localizations and defaults affect what users see.
Therefore, matching country records may reflect a shared language version. These archived country-tagged observations are not proof of separately edited country listings, and they do not establish what every user saw. The retained analysis extract does not identify a separate language treatment for every observation.
What this study cannot tell us
- Whether a matching group came from one coordinated rollout or several decisions.
- Whether a title change was an A/B test, a branding update, a typo correction or another kind of edit.
- Whether the title change improved keyword rankings, downloads or conversion.
- How many localized versions changed outside the observed countries or between observations.
- Whether any particular ASO monitoring product counts or labels changes this way. We audited the data, not a vendor’s reporting behavior.
Localized title changes with different text may belong to one broader initiative but remain separate here. Unobserved countries can hide repetitions. Conversely, matching text changes can still come from separate decisions. The grouped count is an explicit measurement choice, not a verified count of strategic actions.
Sources and reproducibility
Source data: AppOrbit, pinned commit 9621042dde5ad42193bc84c8b54e5bf9b2d4e9eb. The evidence download includes all metadata rows in the retained seven-field extract, transition records, group memberships, code, written rules, results and verification.
Platform context comes from the official Apple and Google documentation linked above. It explains why country and language records should be distinguished; it does not validate the particular historical changes in this dataset.
AppOrbit’s database rights are licensed under ODbL 1.0. Its license does not grant rights to third-party app-store content. The original license and data-format document are included. This is experiment 6 in an exploratory ASO series; earlier unsupported findings remain part of the record.
The ZIP is built into this HTML file. Open the saved report in a browser to use the download button.
Keep exploring
All ASO experiments · See my listing edit study · Track listing changes · Research local keywords