Restaurant analytics: the 9 numbers that actually move profit

Most restaurant dashboards show 50 numbers. About 9 of them actually matter. The rest are vanity charts and noise, and watching all 50 is the surest way to never act on any of them.

This is the working set for a small-to-mid Indian F&B operator. Look at it weekly, on a single screen, with your cook or co-founder. Anything not on this list, ignore until you've solved one of these.

1. Average ticket size (₹ per order)

The single most useful number. Total revenue ÷ number of orders, weekly.

If your average ticket is ₹250 in February and ₹220 in March, something changed. Either prices are softer, or diners are skipping a course, or your menu engineering has slipped. Find out which.

How to move it up:

• Add one ₹80-150 add-on per category (a side, a drink, a dessert). Modal pops on checkout.
• Bundle two cheaper items at a small discount instead of selling them separately.
• Move the dessert section to the top of the menu, not the bottom (read our QR menu setup post for ordering).

What to ignore: trying to push everyone to the most expensive item. That kills volume. Lift the average across the curve, not just the top.

2. Items per order

Total items sold ÷ orders. Should be between 2.5 and 4.5 for sit-down, between 1.5 and 3.0 for takeaway.

If items per order is below 2, your menu is failing to suggest combinations. Common cause: dessert is hidden, drinks aren't discoverable, the category list is too long for diners to scroll.

Quick fix: 6-category menu, drinks as the second category (after the headliner), desserts as the last but with photos that actually look good.

3. Repeat scan rate

Of all unique diners who scanned this month, what % scanned again within 30 days? On QRated this is in Dashboard → Customers.

A 25-30% repeat rate within 30 days is healthy for a casual outlet. Below 15%, you have an experience problem: food, service, or both. Above 40%, you've found product-market fit; double down.

Repeat is also where marketing pays off. A 5% lift in repeat rate is usually worth more than a 20% lift in new diners, because acquisition is expensive and retention is free.

4. Time-to-table (kitchen perspective)

Average minutes from "order received" to "marked served" on KDS. By dish, by station, by cook, by hour.

This number is invisible without a KDS. Whiteboard kitchens have no way to measure it. It's one of the strongest arguments for switching: read our post on KDS vs paper vs whiteboard.

What to do with it:

• Identify dishes that consistently take 3× their median time. Either the recipe has a bottleneck, or the cook is unfamiliar.
• Spot a station that's always slow. Reroute one dish to a different cook for a week. See if it bottlenecks elsewhere or smooths out.
• Watch peak-hour multipliers. If a dish takes 7 minutes at 1pm but 18 minutes at 1:45pm, you're hitting a station capacity ceiling.

5. Drop-off by item

Of diners who viewed an item (clicked into it, looked at the variants), what % added it to the cart?

A drop-off above 80% on a popular-positioned item means something is wrong with the item itself: usually the price feels too high for the photo, or the description is unclear, or the modifiers feel forced.

Drop-off is the closest thing to a focus group you'll ever get for free. Look at it weekly. Pick the item with the highest views + highest drop-off. Fix one thing, usually the photo.

6. Service-time variance (not average)

The average is comforting and hides the truth. Look at the standard deviation of service time, not the mean.

If your kitchen serves at 12 ± 3 minutes most days but 12 ± 10 minutes on Saturdays, you have a peak-hour breakdown. The diners who get served in 22 minutes never come back, and they're the ones who matter.

Fix: stagger orders better, add a runner during peak, simplify the peak-hour menu (mark slow-prep items as "off-menu" between 12:30-2pm).

7. Help-button signal frequency

How often do diners tap "Call waiter / Refill water / Bring the bill" per service? On QRated, this is in Dashboard → Help.

This number is a thermometer for staff coverage. If you're getting 15 "Call waiter" signals per lunch service, you don't have enough waiters on the floor. If you're getting 8 "Bring the bill" signals after 9pm, your closing routine is slow.

The good news: every single signal is a measurable thing the diner is asking for. Not a vague Yelp complaint, but an explicit "I need water". Cure the cause, not the alert.

8. Modifier mismatch rate (or: how often we get an order wrong)

How often does a diner reject or send back a dish? Hardest number to measure because it's not in any analytics tool. Track manually for 2 weeks: every time a dish goes back, write down which item, why, what was different from order.

Then compute: rejections / total dishes served. Healthy is below 0.5%. Above 2%, you have an ordering or modifier problem.

If your modifier mismatch rate is above 1%, investing in a KDS pays for itself in a month.

9. Comp / refund rate

Total value of comped or refunded items ÷ total revenue, weekly. Should be below 1% in healthy operations.

When this climbs above 2%, look at why. Usually it's one of:

• Service errors (wrong dish, slow service, cold food).
• Quality issues (off ingredient, under-cooked).
• Diner-driven (they didn't like it, returned it).

Comp rate is your kitchen's letter grade. Don't celebrate when it's low (it's just hygiene), worry when it's high.

A weekly review template

Print this. Fill it in every Sunday with your senior cook or co-founder. Five minutes.

Three rules:

1. Don't act on noise. If a number moved 3% week-over-week, it's noise. Wait for trend over 3 weeks.
2. Pick one thing to fix per week. Restaurants don't get fixed by 9 simultaneous changes. They get fixed by one focused change at a time.
3. Write down what changed. Last week's review, the change you made, this week's number. Without the journal, three months from now you won't remember why anything moved.

Numbers we don't recommend tracking weekly

These are popular but rarely actionable for a small operator:

• Customer Lifetime Value (LTV). Useful at scale, mostly noise at under 500 customers.
• Net Promoter Score. Doesn't move on weekly cadence. Run quarterly if at all.
• Hourly heatmaps with no context. Pretty chart, hard to act on. If you suspect a peak issue, look at #6 (variance) instead.
• Average rating across all items. A blended average hides the items that matter. Look at the bottom 5 only.
• Social media follower count. Vanity. Track if your job description includes social marketing. Otherwise ignore.

What QRated shows you today

• Scan count, scan-to-order rate, repeat scan rate (Dashboard → Overview).
• Per-item view + add-to-cart breakdown (Dashboard → Menu → item detail).
• KDS time-to-ready & time-to-served per dish (Dashboard → Kitchen → Analytics tab).
• Help-button signal log (Dashboard → Help).
• Order CSV export with all per-item revenue, modifier choices, GST, and refund flags (Dashboard → Orders → Download).

What we don't ship out of the box yet:

• Cohort repeat rate (we show 30-day rolling, not weekly cohorts).
• Service-time standard deviation (you compute from CSV).
• Comp rate (we let you mark refunds, you'll have to compute the ratio).

For a small operator, the CSV export is your friend. Open in Sheets, build the 9-number weekly review on top, you'll be ahead of 90% of operators.

The most important number we didn't list

How happy is your senior cook today? You can't measure it on a dashboard, but if it's low, every metric above will deteriorate within two weeks. Talk to them. Often.