Restaurant analytics you'll actually open.

Today, this week, the same week last year. Split by order type, by payment method, by hour and by site. You can answer the question standing in the car park at four o'clock instead of waiting for someone to send you a spreadsheet in the middle of next month, by which point the Tuesday you were worried about is four Tuesdays ago.

Popular and profitable is easy. The interesting ones are the dishes that sell all night on a thin margin and the ones with a lovely margin that nobody orders because they're buried at the bottom of page two. Eclyde works out contribution per dish and sorts them so you can see which is which, then you move the menu around and watch what happens.

Labour against takings, hour by hour, not as a month-end figure you can only regret. The pattern that comes out of it is almost always the same: one shift a week that's overstaffed for the covers and one rush where the pass is drowning with two people on. Both are fixable next week if you can see them this week.

New against returning, average spend and where it's drifting, lifetime value, and who's showing the signs of going quiet. Customers group themselves by behaviour rather than by a segment you have to define, so the list of people worth chasing is already written when you open it.

Every night: takings, order counts by type, payment breakdown, the cash up, discounts given, hours worked. In your inbox automatically. Read it on the couch in ten minutes rather than pulling reports at eleven the next morning.

Takings and labour figures are right from your first shift, because they're arithmetic on what happened. Menu engineering and demand patterns are not. Those need a few weeks of trading behind them, and a bank holiday or a snowed-off Tuesday in the middle of that will skew things for a while. Treat the early recommendations as a rough steer, not gospel.