Why I Built Installgraph

For years I have been frustrated by how poorly general tech-stack data fits the Shopify ecosystem. So I built Installgraph, a separate product that scans every larger Shopify store at least once a week and reports which apps are installed where.

Over the last couple of months I have been building sales automations for Littledata. One of them looks for brands that have recently switched off a Littledata competitor. A store that has just removed its tracking tool is a store with an open question, and that is a good week to get in touch.

The automation was easy to build. The data to feed it was not there.

Why an uninstall is harder to detect than an install

A new install is simple to spot. A script appears on the storefront that was not there last week.

An uninstall takes more care. A script missing from one page load might be a temporary glitch. And code that disappears is not always an app being removed. Often it is a developer cleaning dead HTML out of the theme, years after the app stopped working. To call something an uninstall you have to be sure the code was active before and is gone now.

We first tried to get this from BuiltWith. We found apps still marked as installed six or twelve months after the store had removed them. For broad market research that lag is tolerable. For a sales trigger it is useless, because the moment has passed long before the record changes.

I know many other people selling Shopify apps have the same problem. They need to know what is installed where as of this week. They also want market data on which apps are winning among established stores, not among the long tail of stores that never make a sale.

What Installgraph scans

Installgraph scans every larger Shopify store at least every seven days. By larger I mean those doing over $1 million a year in revenue on Shopify. That is around 145,000 stores globally, and they hold the vast majority of the purchasing power in the Shopify ecosystem.

Each scan reads which Shopify apps the store runs, plus common scripts and tags that are not loaded through a public Shopify app. Every install and removal is recorded with a date.

The results are public as aggregates. You can browse install counts by App Store category, such as email marketing or product reviews, or look at a single app such as Klaviyo to see how many stores run it, in which countries and verticals, and at what order value.

An API priced for trying it out

The store-level data sits behind an API.

• Free to start — a company email address gets up to 50 free credits. One credit returns the full technology stack for one storefront domain, or one store in the list of stores running an app. Counting how many stores run an app is always free.
• Pay by the credit — beyond that, credit packs start at 10 cents per credit and fall to 2.5 cents in bulk.
• Webhooks — subscribe to an app and get an alert as soon as a scan sees a brand install or remove it.

That last feature is the one I wanted for our own outreach.

Data sources have to be accurate for AI agents to be useful

Building our own marketing automation software over the last six months, I have seen how easy Claude Code makes it for a company to integrate a new data source. The Installgraph docs include a guide for agents for that reason.

But the output is only as good as the input. An agent working from a stale install record will write a confident, well-personalized email about a tool the brand dropped last year.

Sales teams already use many sources of company and people data. I expect Installgraph to become one more part of that stack for anyone selling into Shopify, called directly or as a step in an enrichment waterfall in a tool like Clay.

Who it is for

Installgraph is run by a separate company and is not officially affiliated with Littledata. I founded both. It sells to other Shopify app companies, developers and investors rather than to brands.

It should still be useful to brands. If you are choosing a reviews or subscriptions app, you can see which are the most popular choices in the category among stores like yours, by industry vertical, geography and average order value.

A year ago this needed a full-time team

For me this is a clear example of how AI coding is changing what a small company can take on. Detecting thousands of app signatures, and keeping them current, was not feasible without a full-time team a year ago. New apps launch every week. I now think it is possible to maintain and improve detection at that pace with a fraction of the people.

Most apps are not installed by large merchants

One of the first findings from the project surprised me. The Shopify App Store lists more than 30,000 apps. Around 20,500 of them load something on the storefront that a scan can see. We have found only 15% of those on even one larger store, and only 9% on ten or more.

Some of that gap is detection we have not finished. Most of it is not. An App Store listing does not in itself drive installs among the merchants who spend the money. If you are building or backing a Shopify app, that is worth knowing before you plan around the listing.