Why desktop software vendors still don't know which features get used
If you sell desktop software, how much did you make last quarter? Odds are pretty high you know that number cold. Here's some more: How did your customers use your software? What did they use the least? What usage increased? How many errors were thrown? If you can't get numbers for these, you're missing out. We take it for granted that the billing system needs to be wired up - it is 2026 after all. But what about the product? That's the gap this article is about.
Disclosure: I build Beacon, one of the answers to the gap below.
Why the gap exists
Web companies fixed this around 2009. Mixpanel launched, Amplitude and Heap followed a few years later, and within a couple of years it was normal for any web startup to ship every click to an analytics service. Adding a JavaScript snippet was a five-line change.
Desktop and server ISVs got skipped. The SDKs assumed a browser, a fresh HTTP context on every page load, and constant network access. None of that exists when your customer is using your installer-based app at 30,000 feet, or behind a corporate firewall, or on a laptop that hasn't touched the internet in a week.
So most ISVs either skipped analytics entirely, or rolled their own: an events table, a POST endpoint, and a quiet drift into "nobody queries this anymore." In 2026 a software company can have tens of thousands of paying customers and still not know which of its features get used.
What you actually want to know
It's a small list. None of these are answerable from billing alone.
Which features are people opening? Not in your gut. In the data, broken down by version. You will be surprised. Everyone is surprised. That thing the loudest 5% of customers ask about is usually different from what the other 95% use every day.
Where do trial users drop off? A funnel of "App installed → License activated → Data source connected → First report exported" tells you which step is killing the trial. And when you see "23 of 38 trial users dropped between step 2 and step 3," you want to know who those 23 were, what version they were on, what time they bailed.
Which customers are quiet? If you sell B2B, this is the one that hides churn. An annual license can show "healthy" in the billing system while one user inside a 50-seat customer carries the whole engagement number. The renewal happens because not renewing is harder. Two years later, they don't renew. Beacon tracks customer accounts and licenses as first-class context, so you can ask "which accounts haven't logged in this week?" instead of "which users haven't logged in?" Different question. Different answer. (The IDs are opaque values your app picks. Beacon never sees names, emails, or anything that identifies a real person.)
What does the field actually run? Two years after you shipped v4.0, how many customers are still on v3.2? You probably don't know. Until you do, you can't decide whether to keep backporting fixes or sunset the branch. Same for OS versions, runtime versions, screen sizes below 1080p. These are product decisions you're making by guessing. The data is in your install base; it's just not where you can see it.
Why "we'll just build it ourselves" usually stalls
It starts with an events table, an endpoint, and a weekend estimate. Six months in, the table has fifty million rows, queries time out, nobody built a UI, only the SQL-fluent engineers can ask questions. The offline queue was never finished, so customers behind corporate firewalls or on flaky networks just don't appear in the data. Unit and integration tests you ask? Nope, got skipped.
Building it properly (ingestion, aggregation, UI, offline persistence, retention, privacy controls, rate limits, alerts, and robust testing) is roughly a year of focused work for a small team. Most ISVs don't have that team to spare. They have one engineer doing five other things, and analytics is the first thing to get paused.
What to do this week
Pick the five most important user actions in your product. Instrument them. Wait two weeks. Look.
That's the exercise. The technology choice matters less than the discipline of looking at the data once it arrives. The most common reaction is "I can't believe nobody uses feature X."
If you're wondering how you'd actually instrument this, try Beacon. 21-day free trial, no credit card.