Iran's Two App Stores Disagree About Every App

Tehran Index read Cafe Bazaar and Myket on the same day and matched them by Android package identifier. Forty four apps appear in both. The two stores publish a different install figure for all forty four, the ratio between them runs from 0.67 times to 17.1 times, and the same app scores 0.27 of a star higher on the smaller store. Divar's listing reads 100 million against a national Android base of 61 million. The field everybody quotes barely moves; the field nobody quotes moves every day.

Iran’s Android market runs through its own app stores, and the two that matter both publish a number next to every app. On 24 August 2026 Tehran Index captured both on the same day, matched by Android package identifier: 46 apps on Cafe Bazaar, 48 on Myket, 44 present in both.

The two stores disagreed about all 44.

Not roughly, and not in one direction you could correct for. Cafe Bazaar’s figure is the higher one for 40 of the 44 and the lower one for 4. The ratio between them runs from 0.67 times to 17.14 times, with a median of 1.87. Twenty one of the 44 sit at two times or more; twelve sit below 1.5 times. There is no multiplier that turns one store’s number into the other’s, which means there is no way to take a single store’s install count and scale it into a national figure. That is what almost every English-language reference to an Iranian app’s size does.

They are not counting the same thing

The disagreement is not an error in either store. It is two different quantities wearing the same word.

Myket labels its field تعداد دانلود, download count. That is a flow: how many times the file has been pulled, accumulating forever, never falling. Cafe Bazaar’s English listing labels the same position “Install” and renders it as a floor with a plus sign. And Divar’s own annual report for the year to 20 March 2022 describes its Cafe Bazaar number as active installs: “23.6M+ active Cafe Bazaar installs at year-end”, in the company’s words. That is a stock: devices that currently hold the app, which falls when people delete it.

A stock and a flow are not comparable, and the fact that the stock is usually the larger of the two tells you the gap is dominated by something else entirely, which is how many Iranians use each store. Third-party coverage puts Cafe Bazaar at about 92 percent of domestic Android app installs for 2023-24, and separately in a range of roughly 85 to 97 percent depending on the category measured. We carry that with its attribution and not as a company disclosure, because Cafe Bazaar itself stopped publishing its installed base after its sale in January 2025.

The ceiling test, which one of these numbers fails

Divar reads 100,000,000 on its Cafe Bazaar listing. Iran’s population is about 92 million, on the United Nations estimate for mid-2025. Bazaar’s own annual market report for 2024-25 puts the country’s smartphone base at 68 million and its Android users at 61 million.

So Divar’s listing figure sits 64 percent above every Android user in the country. Whatever it counts, it cannot be devices that currently hold the app, and the phrase “active installs” cannot be doing the work the words suggest. Divar itself published 23.6 million active Cafe Bazaar installs at the end of its 2021-22 year. A move from 23.6 million to 100 million in four years, inside a base of 61 million Android users, does not describe a device count. It describes something cumulative.

This is not a Divar problem. It is what the field is. We already carry the same failure in miniature on one company’s own storefront: OmpFinex states 3.9 million users in the text it wrote for its listing, on a page whose install figure reads 480,000. Both numbers are the company’s. They are not reconcilable from anything public, so we record neither as a user count.

The same app is rated better on the smaller store

The second disagreement is quieter and more interesting. Across the 44 paired apps, the Myket rating is higher for 28, the Cafe Bazaar rating is higher for 8, and the two agree for 8. The mean gap is 0.273 of a star in Myket’s favour.

A quarter of a star sounds small until you look at where it lands. Digikala Jet reads 2.9 on one store and 4.0 on the other. SnappBox reads 2.4 and 3.4. Tapsi Shop reads 3.2 and 4.2. A product manager reading only Cafe Bazaar sees three apps in trouble. A product manager reading only Myket sees three ordinary apps.

The honest caveat is that Myket’s rating rests on a smaller rater base for 35 of the 44, so a single store’s score is partly a function of who bothers to rate on it. But that cuts the same way as everything else here: the star rating on an Iranian app is a property of the store, not of the app.

The number nobody quotes is the only one that moves

Both stores publish a rater count next to the rating, and it is the only field in either listing that behaves like a time series.

Over four days on Myket, 47 of 48 install figures were unchanged and every one of the 48 ratings was unchanged, while 46 of the 48 rater counts moved. Over seven days on Cafe Bazaar, 39 of 46 install figures held, the seven that moved all moved up, and 44 of 46 rater counts rose.

So the field everyone quotes is a coarse bucket that updates when it crosses a threshold, and the field nobody quotes changes almost every day on almost every app. Rater velocity is the usable growth signal in Iran’s app market. It is not a user count either, and we do not present it as one, but it is a directional series with a denominator you can state, which the install figure is not.

What to do with it

Three rules follow, and none of them need anything beyond what the stores print.

Never quote a single store’s install figure as a company’s reach. The gap between the two stores is app-specific, not a constant: on the widest pair it is 17.14 times and on the narrowest it is 0.67 times, a spread of more than twenty five times across the sample. No adjustment carries through a model. Quote the store, the date and the field, or quote nothing.

Never convert an install figure into users. It is a bucket, it is cumulative on at least one store, and on the largest app in this sample it exceeds the national Android base. Where a company publishes its own figure with a period attached, as Divar does in its annual report, use that and cite the report.

Track raters, not installs. That is the field with movement in it, and building the series is the only way to get one, because neither store publishes a history. Ours runs weekly from 29 June 2026 on Cafe Bazaar and from 20 August 2026 on Myket, and it is visible on the App Index.

This is the measurement layer under a lot of published claims about Iranian consumer technology. It also sits directly under the disclosure problem we mapped in The Shape of Iran’s Digital Economy: where a company will not publish an operating figure, the store listing becomes the substitute, and this piece is the argument that the substitute does not hold weight.

What we do not know

We do not know Cafe Bazaar’s definition in its own words. The field carries no label in the listing’s structured data, the page is client-rendered, and the English mirror that does label it “Install” is running a version of the app last updated on 7 November 2024, so we used it for the field name and for nothing else. The reading that the number is an active-install count comes from Divar describing it that way in its own report, not from the store.

We do not know either store’s bucket boundaries. Cafe Bazaar reports in finer steps than Myket, which never showed us a value between two round ones, so part of every gap in the table above is rounding rather than reality and we cannot say how much.

We do not know that any app is absent from either store. Our Myket capture carries no Divar row and a check of four candidate package identifiers on 20 August 2026 returned nothing, but Divar’s own download page links only Cafe Bazaar and Google Play, and we have not confirmed a delisting. Six of the 50 apps in the two captures appear on one side only, and at least one of them, Filimo, is confirmed present on both stores and simply missing from that day’s Cafe Bazaar run. Those six are a property of our capture set, not a finding about the stores.

We do not know how either rating is computed, whether both stores weight recent reviews, or whether a rating survives a version change. And 22 of the 44 paired apps link to a company record in our registry of 371, so this is a scoreboard of the apps we track and not a census of Iran’s app economy.

The stores publish these numbers every day and nobody keeps them. That is the whole reason a two-store disagreement this large has gone unrecorded until now.

Cite as: Tehran Index, “Iran’s Two App Stores Disagree About Every App”, 28 August 2026. Install figures, star ratings and rater counts read from Cafe Bazaar and Myket listings and archived in the Tehran Index App Index; captures of 17, 20 and 24 August 2026. Active-install wording from Divar’s annual report for the year to 20 March 2022. Smartphone and Android base from Bazaar’s annual market report for 2024-25. Population from the United Nations estimate for mid-2025. All ratios, means and counts are Tehran Index arithmetic. tehranindex.com