I had this tool for almost a year before I finished it. That is the first thing I learned, and it has little to do with software. I kept measuring and never cut the piece. But once I started using personal AI assistants, I had an online active nag to help me follow through. Here is my first pass.

Here is what it does. You drop in a QuickBooks sales report. It looks at three things about each customer: when they last ordered, how often, and how much they spend. Each answer gets a score from 1 to 5, compared only with your other customers. Then every customer lands in a group: who to keep close, who to call, who to let rest.

Three things I learned along the way.

1. Simple beats clever. I started with ten groups. On a short customer list, most sit empty, so the home page shows seven. Longer lists can split out a few more. If a group needs a paragraph to explain, it is probably one group too many.

2. Your customer list stays home. Nobody should have to hand over a customer list to learn who their best customers are. The tool runs in your browser. Your data is never shared.

3. The groups are signals, not verdicts. A short list makes the groups rougher. A good customer who had a slow quarter can look like a slipping one. The sort tells you where to look. You know the person, and you decide.

Two plain day-count rules sit on top of the scores (45 and 180 days by default).

Using the tool reminded me of customers that I enjoyed but lost track of—one I have