ASO Research · Experiment 10
Most “new” chart arrivals were returning apps.
About three in four Apple arrivals and two in three Google Play arrivals had already appeared in the same country’s top 100 during the previous two weeks.
Prepared 27 October 2026 · Arrival dates: 3 to 26 October 2026, UTC · Earlier observations provide history · 1,465 arrivals with complete 14-day history · 811 store-specific app IDs
The findingReal examplesRepeat arrivalsChecks and coverageMeaning for ASOMethod and evidence
The theory is supported in this dataset.An app entering the top 100 was often a familiar app coming back after a short absence. The majority-returning pattern appeared in both stores and both time periods, and survived checks using different dates, timing rules and treatment of repeat arrivals.
1. What we found
We identified apps present in today’s complete top 100 chart but absent from the previous daily chart. Then we checked the preceding 14 daily charts for the same app ID in the same store and country.
| Store | Arrivals checked | Confirmed returns | Share returning | Not seen in prior 14 days |
|---|
| Apple App Store | 694 | 529 | 76.22% | 165 |
| Google Play | 771 | 509 | 66.02% | 262 |
529 of 694 Apple arrivals were returns. On Google Play, 509 of 771 were returns. These were observed prior appearances, not guesses based on an app’s name or reputation.
“Not seen in the prior 14 days” does not mean newly launched. An app could have existed for years or appeared in the chart before our lookback period. We did not have the release evidence needed to classify launches.
The headline counts arrival events. The same app can arrive more than once or appear in more than one country. The primary sample covers 190 chart-date comparisons across seven countries in each store. The stores have different app populations.
2.
Amazon Prime Video generated five arrivals in 🇩🇪 Germany
In Apple’s German chart,
Amazon Prime Video entered the top 100 on five dates in the study sample. Every one was a confirmed return.
| Arrival date | Arrival position | Most recent earlier appearance | Earlier position | Daily charts absent in between |
|---|
| 2026-10-04 | #87 | 2026-09-21 | #86 | 12 |
| 2026-10-11 | #85 | 2026-10-05 | #94 | 5 |
| 2026-10-14 | #44 | 2026-10-12 | #82 | 1 |
| 2026-10-17 | #98 | 2026-10-15 | #85 | 1 |
| 2026-10-24 | #91 | 2026-10-22 | #85 | 1 |
That is five arrival events involving one app in one country. On 14 October it returned at #44, so returns were not confined to apps just scraping into #100.
One Google Play app generated seven arrivals in 🇮🇳 India
FRND: Talk to Friends Online appeared as an arrival on 5, 8, 14, 16, 19, 22 and 24 October. It had been observed in the Indian chart before every one of those arrivals.
| Arrival date | Arrival position | Daily charts absent since last appearance |
|---|
| 2026-10-05 | #83 | 2 |
| 2026-10-08 | #86 | 2 |
| 2026-10-14 | #96 | 4 |
| 2026-10-16 | #97 | 1 |
| 2026-10-19 | #93 | 2 |
| 2026-10-22 | #88 | 1 |
| 2026-10-24 | #83 | 1 |
Its absences lasted one to four daily observations. The repeated events describe one app moving in and out of the recorded top 100, rather than seven different competitors.
These examples were selected after analysis to illustrate repeated arrivals. Full timestamps, stable app IDs, prior appearance records and source snapshot IDs are included in the evidence. We also checked that each example was absent from its preceding complete chart.
3. Counting arrivals overstates the number of distinct competitors
The return rate and the number of repeated arrival events answer different questions. An app can return from an appearance before our study period yet generate only one arrival inside it. We therefore counted distinct app-country combinations separately.
| Store | Arrival events | Distinct app-country combinations | Events beyond one per combination | Reduction after counting each combination once |
|---|
| Apple App Store | 694 | 448 | 246 | 35.45% |
| Google Play | 771 | 506 | 265 | 34.37% |
Apple’s 694 arrivals involved 448 distinct app-country combinations. Google Play’s 771 arrivals involved 506. Counting each combination once reduced the totals by about 35% in each store.
An app in 🇩🇪 Germany and the same app in the US remain two separate app-country combinations here. This reduction comes from repeated arrivals within a country, not from merging countries together.
Two numbers that should not be confused:76.2% of Apple arrivals were confirmed returns within 14 days. Separately, 35.4% of its recorded arrival events were additional events beyond one per app-country combination during the study sample.
4. Many apps returned after a short absence near the chart boundary
| Store | Median absence among returns | Absent for 3 daily charts or fewer | Returns arriving at#81 to#100 | Median return position |
|---|
| Apple App Store | 3 daily charts | 291 / 529 (55.01%) | 68.62% | #89 |
| Google Play | 2 daily charts | 336 / 509 (66.01%) | 75.05% | #90 |
Returning Apple apps were typically missing from three daily charts; Google Play apps from two. Most returns landed in positions 81 to 100.
That helps explain why an arrival list can look busy even when the competitors are familiar. Apps near the cutoff can disappear from a top 100 view and come back shortly afterwards.
But the finding was not limited to the last 20 positions. Among arrivals at #80 or better, the full-sample returning shares were 71.86% on Apple and 59.91% on Google Play.
“Absent” means missing from the selected daily top 100 snapshots. It does not mean removed from the app store or continuously absent between observations. We cannot see intraday departures or returns.
5. The result repeated, including in later dates
| Store | Arrival dates | Returns / arrivals | Returning share | 95% interval, apps resampled | 95% interval, date blocks resampled |
|---|
| Apple App Store | 3 to 14 October | 390 / 511 | 76.32% | 72.16% to 80.15% | 73.57% to 80.40% |
| Apple App Store | 15 to 26 October | 139 / 183 | 75.96% | 69.23% to 82.26% | 69.44% to 81.60% |
| Google Play | 3 to 14 October | 295 / 447 | 66.00% | 60.89% to 70.75% | 57.97% to 73.38% |
| Google Play | 15 to 26 October | 214 / 324 | 66.05% | 60.40% to 71.34% | 57.42% to 70.96% |
The later-period figures were 75.96% for Apple and 66.05% for Google Play. Both uncertainty methods remained above 50% in both stores.
App resampling keeps an app’s countries and dates together. Date-block resampling keeps groups of neighbouring dates together. These address different forms of repeated observations; they do not validate the collector or prove the same rates hold across the whole app market.
Repeatedly returning apps did not explain the entire majority
| Later-period check | Apple returns / arrivals | Google Play returns / arrivals |
|---|
| Charts recorded 22 to 26 hours apart | 139 / 183 (75.96%) | 206 / 309 (66.67%) |
| Arrivals ranked #80 or better | 31 / 49 (63.27%) | 56 / 86 (65.12%) |
| First arrival per app-country in the period | 95 / 139 (68.35%) | 139 / 249 (55.82%) |
| First arrival per app in the period | 91 / 133 (68.42%) | 129 / 231 (55.84%) |
| Every history interval 18 to 30 hours | 139 / 183 (75.96%) | 214 / 324 (66.05%) |
| Use each country’s local calendar date | 364 / 503 (72.37%) | 477 / 724 (65.88%) |
| Shorten history to 7 days | 296 / 467 (63.38%) | 274 / 505 (54.26%) |
Keeping only the first eligible arrival for each app-country combination in the later period lowered the shares to 68.35% and 55.82%. Frequent comebacks strengthened the headline result, but the majority remained in this check. These sensitivity percentages are descriptive; we did not calculate a separate confidence interval for every check.
Looking back further found additional returns
We compared 14-day and 21-day history on exactly the same arrivals with complete 21-day coverage. Keeping the event sample fixed avoids confusing a longer history with different countries or dates.
| Store | Same arrivals in both calculations | Returning within 14 days | Returning within 21 days | Additional confirmed returns |
|---|
| Apple App Store | 326 | 75.46% | 79.14% | 12 |
| Google Play | 420 | 61.19% | 65.00% | 16 |
Some apps classified as not seen in the previous 14 days became confirmed returns with another week of history. Neither classification establishes the age or launch date of an app.
6. Missing history mattered, so we checked it explicitly
The strict 14-day requirement excluded many arrivals. Missing charts cannot be treated as evidence that an app was absent.
| Store | All arrivals with valid daily endpoints | Complete 14-day history | Confirmed return despite incomplete history | Unresolved because history was incomplete |
|---|
| Apple App Store | 1,555 | 694 | 601 | 260 |
| Google Play | 1,303 | 771 | 339 | 193 |
The primary 1,465-arrival sample came from 190 complete-history chart windows. Another 1,393 arrivals had incomplete history and were excluded from that main calculation.
For 940 of those excluded arrivals, an earlier appearance was still directly observed. Missing other dates did not erase that evidence of a return. Another 453 could not be classified reliably.
Even if all 453 unresolved cases were treated as non-returns, confirmed returns still made up the following shares of all 2,858 arrivals with valid daily endpoints:
| Store | Confirmed returns / all endpoint-eligible arrivals | Minimum confirmed returning share |
|---|
| Apple App Store | 1,130 / 1,555 | 72.67% |
| Google Play | 848 / 1,303 | 65.08% |
This broader lower-bound check supports the majority finding without assuming what happened on missing dates. It is separate from the complete-history sample used to measure absence length.
The country coverage is narrower than the headline dataset
Complete 14-day UTC history was available in 🇧🇷 Brazil, 🇩🇪 Germany, the UK, 🇮🇳 India, 🇯🇵 Japan, 🇰🇷 South Korea and the US. In the later primary sample, Apple coverage narrowed to 🇩🇪 Germany, 🇮🇳 India and the US; Google Play covered 🇧🇷 Brazil, 🇩🇪 Germany, the UK, 🇮🇳 India and the US.
We repeated the analysis using each country’s local dates. That admitted 2,831 complete-history arrivals across 10 Apple markets and 9 Google Play markets over the full period. Returning shares were 72.92% and 64.73%. In the later local-date sample, the shares were 72.37% and 65.88%, across 7 and 9 markets respectively.
Country results for the primary 14-day sample
Counts below cover the full arrival period. The small Japanese and South Korean samples need particular care. “No complete-history sample” means excluded coverage, not zero returns.
| Country | Apple returns / arrivals | Google Play returns / arrivals |
|---|
| 🇧🇷 Brazil | 78 / 109 (71.56%) | 89 / 148 (60.14%) |
| 🇨🇳 China | No complete-history sample | No complete-history sample |
| 🇩🇪 Germany | 182 / 230 (79.13%) | 97 / 149 (65.10%) |
| 🇬🇧 United Kingdom | 93 / 118 (78.81%) | 57 / 74 (77.03%) |
| 🇮🇩 Indonesia | No complete-history sample | No complete-history sample |
| 🇮🇳 India | 78 / 118 (66.10%) | 156 / 227 (68.72%) |
| 🇯🇵 Japan | 16 / 20 (80.00%) | 23 / 32 (71.88%) |
| 🇰🇷 South Korea | 17 / 26 (65.38%) | 21 / 28 (75.00%) |
| 🇺🇸 United States | 65 / 73 (89.04%) | 66 / 113 (58.41%) |
| 🇻🇳 Vietnam | No complete-history sample | No complete-history sample |
7. Why this matters for ASO competitor research
A busy arrival list does not automatically mean a flood of new competitors. In this sample, much of that activity came from apps returning to a chart they had recently occupied.
| What a chart record establishes | What it does not establish |
|---|
| “This app entered today’s top 100 after being absent yesterday.” | That the app was recently launched or is a new business. |
| “This app appeared here earlier in the past 14 days.” | Why it left, why it returned, or whether its ASO changed. |
| “We recorded seven arrival events for FRND in 🇮🇳 India.” | That seven distinct competitors entered the market. |
| “We have no appearance in a complete 14-day history.” | That the app had never appeared before or did not compete elsewhere. |
For ASO reporting, the useful distinction is between an arrival event, a returning app and a distinct competitor. The chart data can establish the first two and identify distinct app IDs. It cannot establish who competes for the same keywords or users.
This study concerns top-free app charts. It does not measure keyword search ranks, downloads, revenue, advertising or the effectiveness of listing changes. No ASO vendor’s arrival labels were audited, so the report does not claim that a particular tool misclassifies apps.
8. How the study was done
- Validate the charts. We used 714 valid, complete top-free all-apps snapshots. Each had 100 unique app IDs and every rank from 1 to 100. Paid charts and games charts were excluded.
- Choose daily observations. Take the latest eligible snapshot for each store, country and UTC date. This produced 671 daily charts over the source period. No missing chart or app rank was filled in.
- Allow a 14-day history period. Arrival dates were 3 to 26 October. Earlier September observations supplied history. The first comparison period ended 14 October; the later period began 15 October.
- Identify an arrival. The app must appear in today’s complete top 100 and be absent from yesterday’s complete top 100. The two snapshots must be 18 to 30 hours apart. In the study dates, 414 current daily charts yielded 373 eligible endpoint windows after excluding 29 without a preceding daily chart and 12 with unsuitable timing.
- Classify the history. Require all 14 preceding daily charts for the main sample. An app seen anywhere in that history is a return. Otherwise it is not observed in that lookback. Incomplete histories remain separately identified.
- Keep the units clear. The headline is event-weighted because the question concerns arrivals. We separately count distinct app-country combinations, give apps equal weight and retain only first arrivals as sensitivity checks.
Prediction, uncertainty and independent calculation checks
The written protocol predicted a returning share above 50% in each store and each period, with at least 30 primary arrivals per store-period. All four groups passed. The stronger requirement was for both later-period app-cluster and three-calendar-day-block 95% lower bounds to exceed 50% in both stores. That also passed.
Each interval uses 5,000 resamples. The two methods address app repetition and shared dates separately. They are not a joint correction for every possible dependency or a guarantee against source error.
The protocol was written before calculating Experiment 10 outcomes. It was not registered externally, and the source had been used for previous experiments. The earlier/later split is an internal repeat rather than an untouched external holdout.
A separate verification program rebuilt daily app-rank matrices directly from the raw chart-entry file. It reconstructed all 2,858 eligible arrivals and their history classifications, all 1,465 primary events, the 7-day and 21-day variants and the local-date variant. It checked 60 summary cells, the broader coverage bounds and all six primary app-cluster confidence intervals. All checks passed.
This verifies selection and arithmetic against the same source. It is not a second independent collection of app-store charts.
Protocol SHA256: 19e1371a47d866eaadf886d7f74241ed5237f306567b792a2347fd1aeddb3017
The main limits
- Short observation period. The source spans 38 UTC dates. The study measures recent returns, not lifetime first appearances.
- Selected charts and countries. The source is a third-party collection rather than a representative sample of every app or market. Complete-history filtering changes coverage.
- Observed daily returns. We cannot establish continuous absence, detect every intraday crossing or infer why an app disappeared from a chart.
- App identity is not company identity. Two different app IDs can belong to the same publisher. We do not count distinct businesses.
- No causal ASO conclusion. A return can coincide with many changes. The records do not show which changes, if any, caused it.
9. Download the evidence
The evidence ZIP contains the source chart records, snapshot index, metadata used for app names, all arrival classifications, protocol, calculations, verification output and file checksums.
Evidence ZIP: 9.7 MB, embedded in this HTML. Python and NumPy are required to rerun the study. No images or external scripts are needed to read the report.
Source: AppOrbit, pinned data commit. The included source format guide explains chart completeness, date fields and metadata collection. All results were calculated from previously preserved source files; no live storefront census was performed for this experiment.
Publication finding: Most top 100 arrivals in this sample were apps returning to a chart they had occupied within the previous 14 days. Arrival counts therefore describe chart activity more than the number of newly encountered competitors. Tracking stable app IDs and prior appearances changes the interpretation.
Prepared for Gabriel Machuret · ASO Experiment 10 · AppOrbit database rights under ODbL 1.0. Third-party store content retains separate rights. Preserve attribution and consult the included license before redistributing source data.