Home/Research

Markets · App marketplaces

Do app marketplaces bring customers by themselves? What we learned publishing on four with no audience

They do not. We submitted apps to Shopify, Google Workspace, Atlassian and monday.com without an audience of our own. Listings went live, people installed some of them, and paying customers stayed close to zero. Each marketplace failed in a different place, and the platforms' own rules explain part of it.

Published 2026-10-11

The short answer

An app marketplace does not bring paying customers by itself. A listing is a shelf: it lets a buyer who is already searching find and pay for an app, but it does not create that buyer. We published apps on Shopify, Google Workspace and Atlassian, and submitted one to monday.com, with no audience, list or community of our own. On 11 October 2026 we had 39 apps live out of 61 built and, across all four marketplaces, paying customers had stayed close to zero.

The failure was not the same everywhere. On Shopify our apps were partly invisible because of our own settings. On Atlassian the review gate held most of them for weeks. On Google Workspace people did install, but almost nobody wanted the paid tier. On monday.com the submission was rejected. This article sets out what we measured on each, what the platforms' published rules say, and what we would do differently.

What we put on the shelves

As of 11 October 2026: 20 apps live on the Shopify App Store, 15 add-ons live on the Google Workspace Marketplace, 4 apps public on the Atlassian Marketplace out of 25 submitted, and 0 on monday.com, where our one app was submitted on 16 August and rejected on 28 September. We had no mailing list, no community and no advertising budget behind any of them. The question we were testing was whether marketplace search alone would bring paying users.

Shopify: invisible by our own switch, then visited by testers

In early September we found that a large part of our approved, published Shopify apps were set to unlisted. Shopify's documentation is explicit that only a listed app appears in category pages, App Store search and third-party search results; an unlisted app can be installed only from its listing URL. The public directory showed only a fraction of our apps; after we switched the setting, all of them appeared. The same audit found apps with no public pricing plan configured, so their cards read "Free" and a merchant had nothing to click to pay.

Fixing visibility brought installs, not customers. The outside installs we saw followed one pattern: install, cycle through the paid plans in test mode within minutes, uninstall with the reason "Testing multiple apps". Few installs were real, fewer stayed, and when we checked on 20 September the only active subscription on an installed store was a test charge. On Shopify, the paying side stayed empty.

For context, our own scan of the Shopify App Store found that the median app has 0 reviews and 52% of listed apps show no sign of a single customer. Being one of them is the base rate, not bad luck.

Atlassian: a review gate, then the name

We submitted 25 Forge apps. By the end of August, most of them had been rejected because review timed out, and a few for a real security defect, which we fixed. Most of the delay came from a rule we had not read in advance: Atlassian's Forge documentation says that once an app is submitted for listing, or carries a licence in its manifest, it can no longer be shared through an installation link. The reviewer could not install the apps from the listing either, because that is only possible after approval. The documented workaround is an unlicensed copy of each app, which we built for all of them.

The first app was approved on 8 September. By 11 October the public apps had no installs beyond our own and no reviews. They were findable only where the app name contained the search phrase: near the top for a query that matched the name, far down the results for one where the phrase was only in the tagline. We describe that measurement in a separate research note. Paying customers had not appeared there either.

Google Workspace: installs that are not people, and people who do not pay

Google Workspace was the only marketplace that sent real traffic. It also taught us to distrust the public number. Google's documentation separates domain installs, seat installs (users who have the app installed and enabled) and individual end user installs (users who installed it directly). When an administrator installs an add-on for a whole organisation, every user in it becomes a seat, so a single domain install can add an entire company to the public figure while the number of people who chose the app themselves stays much smaller. On our own add-ons, most of the public figure turned out to be administrator installs, not people.

The traffic came from inside the marketplace. Almost all sessions on our busiest add-on were direct, meaning the marketplace and the editor's add-on picker, and web search sent next to nothing. The listing itself converted well: most visitors pressed Install, though a noticeable share of them abandoned at Google's consent screen.

Payment is where it stopped. Very few people across the whole fleet ever reached a pricing page, and among the many who had installed an add-on themselves, only a handful touched the licence screen. Paying customers stayed close to zero. The payment path works end to end; what is missing is people who want what the paid tier offers.

monday.com: the policy decided before the market did

monday.com's developer documentation says the review team no longer accepts apps that copy the core functionality of an existing marketplace app, and rejects apps whose main purpose is integrating with a third-party product that already has a marketplace integration. It names Google tools, Microsoft tools, WhatsApp, Shopify, Zendesk, Calendly, Zoom and Telegram, and says apps built mainly with no-code platforms or AI-generated code are not eligible. Since July 2024 new apps must also use monday's own billing.

Our scan of all 1,281 monday.com apps on 16 August showed why discovery there is hard for a newcomer. Apps created in 2025: 288, median 62 installs, 3 above 2,000. Apps created in 2026: 235, median 5 installs, 0 above 2,000. Install counts are a stock, so younger apps have had less time; even so, the curve is falling while new supply keeps arriving.

Where the report was wrong

We read an AI deep-research report that ranked marketplaces for a small studio with no audience. It put monday.com first and suggested building a competitor to a popular Google Sheets integration there, on the strength of its public install count and weak rating. monday.com's own policy names Google tools among the integrations it will reject. The report did not mention that policy.

The same report treated public install counts as evidence of demand. On Google Workspace the public figure counts seats, so one administrator can add an entire company; on our own add-ons, the public number was mostly administrator installs rather than people. The report's closing warning was right, though: the riskiest assumption for a newcomer is whether marketplace search shows a new listing to strangers at all.

Where we were wrong

We treated publishing as distribution. Building another app felt like progress because it was measurable; whether anyone would find it was not measured until weeks later. A large part of our Shopify fleet sat unlisted, and nothing in our process looked.

We submitted 25 Atlassian apps at once, and all of them hit the same gate together. One pilot would have found the install-link rule in a day.

We read a public install number as people, and ranked our own add-ons by it for weeks. And twice, when revenue was zero, we went looking for a broken payment path. The path worked. The number that answered the question, how many people even looked at the paid tier, took seconds to read once we thought to read it.

What we would do instead

Read the platform's submission and duplication policy before choosing an idea. Pilot one app through review before submitting a fleet. Check, from outside, that each listing appears in the marketplace's own search for the phrase a buyer would type. Measure the funnel in the order money flows: listing visits, installs by individuals, people who see the price, people who pay. And find the first buyers by hand, where they already describe the problem, so the listing has someone to amplify. The companion study on how small B2B companies got their first customers shows that the founders who succeeded mostly did exactly that.

Method and limitations

The observations about our apps come from our own measurements between August and October 2026: marketplace consoles, partner event logs, public search queries and our own databases. We describe our own results qualitatively and publish no counts of our installs, users, visitors or customers; the catalogue counts are the ones shown publicly. The platform rules were checked at their official documentation on 11 October 2026. We started from an AI deep-research report on marketplaces and kept only claims we could confirm at the source.

This is one small studio's experience. Our apps were mostly utilities in crowded categories, and our listings were written to pass review before they were written to be found, so a stronger product or better listings could do better. The observation window is short: the Atlassian apps had been public for about a month. We report no money figures.

Sources

1. Shopify: Visibility in the Shopify App Store
2. Shopify Partners blog: search improvements in the Shopify App Store
3. Google Workspace Marketplace: use analytics
4. Atlassian Forge: distribute your apps
5. monday.com developer docs: planning your app (duplication policy)
6. monday.com changelog: new monetization requirements for marketplace approval

Questions and answers

Will publishing an app on the Shopify App Store bring customers on its own?

Not in our experience. We had 20 Shopify apps live; the outside installs were mostly testers cycling through plans, few stayed installed, and the only active subscription we found was a test charge. Our scan found the median Shopify app has zero reviews, so silence is the base rate for a new listing without outside traffic.

What does the user count on the Google Workspace Marketplace mean?

It mostly counts seats. Google's documentation separates domain installs, seat installs and individual end user installs. When an administrator installs an add-on for a whole organisation, every employee becomes a seat. On our own add-ons the public figure was mostly administrator installs, far above the number of people who installed the add-on themselves.

Why can an Atlassian Forge app not be installed by the reviewer?

Atlassian's Forge documentation says that once an app is submitted for Marketplace listing, or has a licence in its manifest, it cannot be shared through an installation link, and installing from the listing is only possible after approval. The documented route is an unlicensed copy of the app with sharing enabled, sent to the reviewer.

What kinds of apps does monday.com reject from its marketplace?

monday.com's developer documentation says it no longer accepts apps that copy the core functionality of an existing marketplace app, or whose main purpose is integrating with a product that already has a marketplace integration, naming Google, Microsoft, WhatsApp, Shopify, Zendesk, Calendly, Zoom and Telegram. Apps built mainly with no-code platforms or AI-generated code are not eligible.

If marketplaces do not bring customers, how should a small software team find the first ones?

Find people who are already describing the problem in public, answer them usefully, and contact them individually before relying on the listing. Pilot one app through review, confirm the listing appears in the marketplace's own search, and measure how many people see the price, not only how many install. A listing amplifies demand; it rarely creates it.

All research