Early in my career, a support incident taught me a lesson about mental models. Here’s what happened: I was contracted to create a small promotional app for executive assistants who used Windows PCs. Many didn’t have CD drives, so the app was designed to fit on a floppy disk.

To install the app, users would slide the disk into their computer and double-click on a file called something like INSTALL.EXE. Then they’d follow the onscreen prompts. The disk included printed instructions that spelled out the process.

Shortly after we released the app, I got a message from the client. A user was having trouble installing the app. Would I mind taking a look? So I drove to the user’s office and asked her to show me what she was doing. What I saw blew me away.

She put the floppy in and looked for INSTALL.EXE — but not using Windows’s File Explorer, as I’d expected. Instead, she opened Microsoft Word, where she looked for the Open command. She found the installer, but it wasn’t a DOC file, so it wouldn’t open.

At first, I was relieved. I’d feared the user might have discovered a bug after we’d already copied and distributed scores of disks. Instead, the issue was what we derisively call “user error.”

But was it? As the designer, I’d assumed users had an accurate mental model of the overall system. A computer’s OS is a platform for running apps. Our multimedia app ran on that platform, alongside other apps like MS Word. Everybody knows this.

Except they don’t. For this user, MS Word was the platform. Word was likely the app she used the most. She seldom installed other apps. She was used to opening files in Word, so it was natural for her to interpret the instructions in this way.

This “unsophisticated” understanding wasn’t the user’s fault. Computers are complex. (And not as widespread then.) Rather, our instructions were vague. Instead of writing, “Open INSTALL.EXE,” we should’ve spelled out the steps for finding and opening INSTALL.EXE using File Explorer.

That day I learned users have different mental models of how systems work. Some models are better than others. As designers, our models are likely more sophisticated than those of most users.

Designers need to understand the system. But we must also understand how users think about the system and how it works. To do so, we must meet users where they are — literally and figuratively. Without further work, our deep understanding of the system can be a liability.

A version of this post first appeared in my newsletter.