Restaurant analytics, explained.

Fewer than any dashboard will offer you. A report with forty charts on it gets looked at twice and then never again, which is worse than a page with six numbers on it that you read every week.

Here's the short list that covers most decisions an independent has to make. Everything else is a follow-up question once one of these moves.

By comparing like with like, and by never trusting a monthly total on its own.

Months are the worst unit in this trade. They contain different numbers of Fridays, one may have a bank holiday and the other a heatwave, and a five-Saturday month will always look like growth. Compare this Tuesday with the last four Tuesdays instead, and you'll see something real.

Then split by channel, because total revenue hides the thing that matters most. A month where marketplace orders grew and direct orders fell can look flat on the top line while your margin has quietly dropped, since you're keeping 70 to 75 cents in the euro on one and all of it on the other. Watch the mix as carefully as the total.

Finally, look at the hours. Most restaurants have a rush that's narrower than the rota assumes and a long tail that costs more to stay open for than it brings in. That's a rota decision and sometimes an opening hours decision, and you can only see it by the hour.

Sort every item by two things: how many you sell, and the cash it leaves after food cost.

Cash margin, not food cost percentage. A dish at 35% food cost that leaves €9 on the plate beats a dish at 25% that leaves €4, and percentage thinking will have you pushing the wrong one. Cost the portions as they actually go out, and include packaging, sauce pots and trim waste, because delivery boxes and dips are a real line.

That gives you four groups and four different jobs. Popular and profitable: protect them, put them where the eye lands first, never run out. Popular and thin: fix the recipe, the portion or the price, since volume is multiplying the problem. Unprofitable and rarely ordered: cut them, and take back the prep and the stock lines. Profitable but ignored: usually a description or position problem, and often the cheapest win on the whole menu.

Three things, and most independents can't answer any of them.

What share of this month's orders came from someone who had ordered before. How many first-time customers ever place a second order. And how much of your revenue comes from your top fifth of customers, which in most restaurants is a much larger share than owners expect.

Those numbers change what you'd spend money on. If first-to-second conversion is poor, the problem is the first experience or the absence of any follow-up, and no amount of advertising will fix it. If repeat rate is healthy but new customers are scarce, the problem is visibility, and that's a local search and reviews job. Same total revenue, entirely different response.

You need names attached to orders for any of this, which is one of the practical arguments for taking orders on your own channel. Customer intelligence is where this sits in Eclyde.

As a percentage of sales, by hour, while the week is still happening.

Reading last week's labour cost after payroll has run is a history lesson. Seeing on Thursday that the week is running hot means you can adjust Sunday's rota. That's the whole difference between a report and a useful number.

Two views do the job. Labour as a percentage of sales, week to date against your normal range. And sales per labour hour by hour of the day, which tells you honestly whether the extra body at four in the afternoon earns their shift. Clock-in and clock-out at the till rather than a paper sheet, because remembered hours are always generous, and the gap is usually bigger than any saving you'd find elsewhere.

The operational ones nobody puts on a dashboard, which are often where the money is going.

Median ticket time by hour tells you when service breaks down, and it's usually a tighter window than the team believes. Remakes and refunds by reason tell you whether errors start at the order screen, at the pass or on the road. Voids and discounts by staff member, over a month rather than a night, will occasionally tell you something you'd rather not know. Missed calls in your busiest three hours are orders that went to the restaurant across the road. And the share of online baskets that never become orders tells you whether your own checkout is quietly costing you sales.

None of these show up in a sales report. All of them are worth more than another revenue chart.

Little and often beats a grand review that never happens.

A daily glance takes two minutes: yesterday's takings by channel, order count, anything obviously odd. A weekly half hour on a quiet morning is where the real work happens: the seven numbers at the top of this page, compared with the same period, and one question written down for anything that moved. A monthly hour for menu margin and supplier prices. A quarterly session for the bigger decisions, opening hours, staffing shape, whether that Monday lunch is worth the door being open.

Put the weekly half hour in the diary on the same morning each week. The habit matters more than the tooling.

Almost everything, if you don't know what else was going on that week. This is the section that saves you from confident bad decisions.

Small samples are the first trap. Four covers on a Tuesday is not a trend, it's weather. Averages are the second: an average order value that includes a big office catering drop tells you nothing about your normal customer, so look at the median or strip the outliers. Promotions are the third, since a discount week will lift order count and drop average order value at the same time, and comparing that week with anything is meaningless.

Then the quiet ones. Comparing to last month instead of the same weekday. A bank holiday in one period and not the other. Counting marketplace orders and direct orders as equivalent when one of them keeps 70 to 75 cents in the euro. And the biggest of all: assuming the data tells you why. It doesn't. It tells you where to look, and then you go and stand on the floor at that hour and see for yourself.

Turn it into one action a week. Not five, one, and finish it before the next review.

That's the whole discipline. Ticket times fall apart between eight and nine on a Friday, so this week you move one start time. First-order-to-second conversion is poor, so this week you write the thank you message. Three dishes are losing money, so this week you re-cost them and change one price. Next week you check whether the last change did anything, and you pick the next one.

Owners who do this consistently pull ahead of ones who buy a better dashboard. If you want the platform side of it, Eclyde analytics puts these numbers in one place, but the half hour on a Tuesday morning is doing most of the work.