I built an early version of Caretrics that let people ask questions about their data. I thought that freedom would be useful.

But people didn’t necessarily know what to ask.

I had built around what the technology could answer. I needed to spend more time thinking about where the person using it was starting.

A clinic owner opens the software between everything else they have to do. They might know something feels off. They might just want to know how the week went. Either way, “What would you like to know?” gives them another job.

Showing them something worth investigating gives them a starting point.

Consider an illustrative example. Suppose your clinic billed less than usual last month. Your first question might be whether you need more patients.

But looking closer, you discover that bookings were steady. More appointments were cancelled, and those spaces stayed empty.

Now you have a different question: why couldn’t we fill them?

One insight. A better next question.

Illustrative clinic example

  1. 01Notice

    Revenue is down.

    Do we need more patients?

  2. 02Look closer

    Bookings were steady. More appointments were cancelled.

    Where did the bookings fall away?

  3. 03Investigate

    Why did those spaces stay empty?

    Timing? Short notice? Difficulty reaching the waiting list?

Each answer should help you ask a better question.

Perhaps cancellations cluster around particular appointment times. Perhaps your team learns about them too late. Perhaps there’s a waiting list, but contacting people takes more time than the front desk has.

Those possibilities call for different decisions. Spending more on marketing might bring in enquiries while leaving the same gaps in your schedule.

That’s why the second and third questions matter. Each answer should help you understand what to investigate next. A pattern gives you somewhere to look; you still need to check what’s happening with the people doing the work.

This changed how I think about building Caretrics. Start with an insight someone can understand. Make the evidence easy to examine. Let them dig deeper as their questions become more specific.

Some people will want a chart. Others will ask a follow-up question. The interface should support both without making them learn how the software thinks.

AI can produce an extraordinary amount of information. The harder design problem is deciding what will help someone move forward.

“Ask anything” is a promise about what the software can do.

“Here’s something worth looking at” gives the person a place to begin.