I scanned every app on the Shopify App Store before building one
Twenty-five thousand listings, one request a second, eleven hours. What the whole store looks like from above, why a regulation made one app and buried forty-three others, and how I chose what to build.
I wanted to build a Shopify app, and I did not trust my own guess about what to build. Everyone who has spent time inside Shopify stores has a list of things merchants complain about, and every one of those complaints already has six apps for it. So before writing any code I read the whole store.
Not by hand. The App Store publishes a sitemap of every listing, 26,004 URLs when I pulled it, and each listing page is server rendered: rating, review count, launch date, developer, pricing, categories. Shopify blocks search from crawlers (its robots.txt disallows any URL with a query in it) but the listings themselves are open. A script fetched them one at a time, slept when the server answered 429, and resumed where it stopped. One request a second was the rate the server would tolerate. A full pass took about eleven hours and came back with 25,848 apps across 122 categories.
One caveat that decides how you read everything below. Shopify does not publish install counts. The only public number that moves with adoption is the review count, so I used reviews per month since launch as the demand signal. That is a proxy, and a biased one: free apps collect reviews far more easily than paid ones, so any ranking by review velocity will put free apps at the top whether or not anyone is paying for anything. I come back to that.
What the store looks like from above
Three numbers I did not expect.
Sixty percent of the store has never been reviewed. 15,472 of the 25,848 apps have zero reviews. Another 6,314 have between one and nine. Only 1,203 have reached a hundred, and 165 have reached a thousand. The listings you see when you search are the thin top of a very long tail.
More than half the store is new. 13,713 apps launched in the fourteen months before the scan. That is the AI-built app wave arriving, and it changes what "competition" means: most of your competitors did not exist last year and most of them will have no reviews next year either.
Breaking through is rare in a way that is worth stating precisely. Of those 13,713 new apps, 3,048 have any reviews at all, 62 have reached forty, and 14 have reached a hundred and fifty. One in a thousand. If the plan is one app and it has to work, the plan is wrong. The plan has to be a loop that can afford the first one failing.
The regulation that made one app and buried forty-three
The clearest pattern in the data is a European one. Directive 2023/2673 required a withdrawal button on distance contracts from 19 June 2026: a visible control on the storefront that lets a customer cancel. Eighty-four apps in the store mention it.
The one that won launched on 9 January 2026, five months before the deadline. It has 2,287 reviews, close to three hundred a month, and it is the fastest-growing app of any kind in the fourteen months I can see. The next two launched in March and May and have 534 and 319. Then the deadline arrived and so did everyone else: 47 withdrawal-button apps launched within a month either side of 19 June, and they average 1.2 reviews a month. An app that has offered a withdrawal button since 2021 has 81 reviews in five years.
So the lesson is not "be early". The app from 2021 was early and got nothing, because nothing required the button yet. The lesson is that the prize goes to whoever is already in the store when the requirement lands, in the months when merchants are searching for it and before the copies arrive. The window was roughly February to May. After June the category was a queue.
The counter-example is the same regulation family a year earlier. The EU's General Product Safety Regulation created data fields merchants were supposed to fill in. Forty-six GPSR apps exist. Forty-three of them have zero reviews, and the two with any traction have 29 and 15. Same legal urgency on paper, no prize at all, because nothing on the storefront visibly changed and nobody was enforcing it. What separated the two was not the deadline. It was a required element a merchant could see, and a country (Germany, with its cease-and-desist letter culture) where ignoring it had a price.
Re-cutting by who pays
This is where the review proxy has to be corrected. Take the AI-search-visibility category, the "get found in ChatGPT" apps, which the scan surfaced early as a wave. The three apps with the most reviews there are all free or list no price. Every paid one sits between fifteen and thirty-one reviews. By velocity the category looks alive; by money it is unproven.
So I applied one filter to the whole store: launched in the last three years, a starting price of $9.99 or more, no free tier, and at least forty reviews anyway. That leaves 27 apps out of 25,848. Those are the apps that merchants paid for from day one and still reviewed, and they are the only honest evidence of a paying market I could find in public data. They cluster in three places: marketplace syncs (TikTok Shop, eBay, Walmart, Etsy), cart upsells and bundles, and order editing. One studio has three of the 27, all syncs, which says more than any single app does: a studio that ships repeatedly into one category is the pattern that works.
What I chose
Two things, both built on Shopify Functions, the newer way to change checkout behaviour, because the incumbents in those categories predate it and each does one slice. One does quantity breaks and bundle discounts; four of the 27 paid-only apps are bundle apps, so that market pays. The other hides, renames and reorders payment and delivery options and blocks a checkout that breaks an order limit. That category is a different bet: its recent entrants grew fast (one launched in April 2025 has 521 reviews, another 215) but every one of them has a free tier, so it is a proven install market rather than a proven paid one, and a free tier is the price of entry. Neither app touches money or customer data, which keeps support light. Neither is listed yet. I have said elsewhere on this site that I do not announce things before they are real, and that holds here.
What I did not choose, and why, is the more useful half. I did not build for the next regulatory deadline, because the ones I can see either land on incumbents (the UK subscription rules, which the subscription platforms will absorb) or are too far out and keep moving. And I did not build in the categories that look busiest, because busiest in this data means most installed for free.
If you want to do this yourself
Pull the sitemap. Fetch listings one a second and expect to be throttled; the scan is a day, not an hour. Treat reviews as installs, not revenue, and re-cut by price before believing any category. Look for a required, visible change to the storefront with someone enforcing it, and be in the store months before it lands rather than the week it does. And plan for the first app to fail, because the numbers say it will, and build the loop so the second one costs a week rather than a quarter.