Ask a retail founder to show you their analytics and you will usually see the same thing: a dashboard with forty tiles, opened twice — once the day it launched and once for this demonstration. We have built our share of those. This post is what we learned from watching which numbers store owners actually act on, and why the dashboards that work are the ones that fit on a phone screen without scrolling.
The wallpaper problem
Dashboards fail for the opposite reason software usually fails: not too little, but too much. When a screen answers forty questions, it answers none urgently, and checking it stops feeling like a decision and starts feeling like homework. Homework loses to running the store, every single day.
The fix is not better charts. It is a brutal editorial stance: a metric earns a place on the main screen only if there is a specific action the owner takes when it moves. No action, no tile.
The five numbers that survive the test
Across the POS and retail systems we have shipped, five metrics pass that bar almost everywhere.
Today's sales against the same day last week
Not against yesterday — retail breathes weekly. Tuesday against last Tuesday is signal; Tuesday against Monday is noise. When it drops hard, the owner walks the floor or checks the till, today.
Gross margin this week, not just revenue
Discount-heavy days can set revenue records while quietly losing money. Owners who see margin next to revenue stop running promotions that work brilliantly on the top line and terribly everywhere else.
The stock-out list
Every item that sold out with demand still active is invisible lost revenue — nothing rings, nothing alerts, customers just leave. This is the tile owners act on most: it converts directly into a reorder before lunchtime.
Dead stock over ninety days
The inverse problem: cash sleeping on shelves. A short ranked list — worst ten offenders — triggers markdowns and freed-up capital. A full inventory report triggers nothing, because nobody reads it.
Average transaction value, per staff shift
Not for surveillance — for coaching. When one shift consistently baskets 20% higher, something teachable is happening on that floor, and the number is what makes anyone go look.
A dashboard's job is not to describe the business. It is to start five specific conversations a week. Description is what the export button is for.
Everything else goes behind a tap
The forty charts do not disappear — seasonality curves, supplier lead times, category drill-downs all live one level down, for the sit-down-with-coffee sessions where they belong. The main screen stays five numbers, phone-sized, glanceable between customers. That restraint is not a smaller deliverable; it is the deliverable.
If your current system reports everything and changes nothing, talk to us — the first thing we will do is delete tiles, and it will feel great.