Most restaurant dashboards fail the Monday test. Forty charts open, you scroll through all of them, and you still haven't made a single decision. The problem is rarely too little data. Usually it's the opposite: too many numbers, none of them tied to a threshold that tells you when to act. After watching hundreds of delivery brands run on our platform, we've settled on twelve worth checking every week: four about demand, four about operations, four about money and retention. Each comes with a reason to exist and a target band. Treat the bands below as starting points and recalibrate them against your own three-month baseline, since a healthy number for a pizza chain can be a warning sign for a sushi place.
Demand: reading the top of the funnel
1. Orders per day, by channel
This is your pulse, but the total on its own will lie to you. Split it by channel: website, app, and each marketplace on its own line. A flat overall number can mask your direct orders shrinking while a marketplace charging 30% commission grows to cover the gap. Target: direct channels holding steady or rising week over week. When marketplaces outgrow direct for a full month, acquisition is running ahead of retention.
2. Conversion rate, by channel
The share of visitors who place a paid order. This one tells you whether your money should go into fixing the funnel or into buying more traffic. Across 3M+ orders on our platform, branded apps run up to 35%, ordering sites land at 10–15%, and marketplace listings sit near 3%; there's full context in our conversion benchmarks breakdown. If you're below half of the relevant band, fix conversion before you spend another cent on ads.
3. New vs. returning customer split
Two ways to fail hide in this one ratio. Tip past roughly 70% returning and you've stopped acquiring new customers, so the base only shrinks from here. Push past 70% new and you're paying full acquisition price on every order, with nothing compounding underneath. Most healthy delivery brands land somewhere around 30–45% new.
4. Average order value
Courier cost and payment processing barely move when the ticket goes up, so almost every extra unit of average order value drops straight to margin. Target: hold it steady or add 3–5% a quarter through bundles and smart recommendations. Raising prices works too, right up until conversion drops and hands the gain back.
Operations: keeping the promise
5. Promised vs. actual delivery time (p90)
An average will smooth over exactly the orders that lose you customers: the late ones. The 90th percentile won't let it. Target: keep p90 inside the time you promised, and under 45 minutes for hot food. Every 10 minutes past that promise costs you a measurable slice of reorders, and we've put the numbers on that in the data on delivery time and retention.
6. Kitchen prep time
This is the half of delivery time you actually control. Watch its spread more than its average: when prep swings from 9 minutes to 25 across a day, either couriers stand around waiting or the food goes cold before it leaves. Target: 80% of orders inside a 6-minute band around your norm.
7. Order accuracy
Every wrong or missing item costs you twice: the refund now, and the customer who doesn't come back. Target: complaints under 1% of orders. Once you're above 2%, the culprit is almost always someone re-keying orders from one system into another. Putting the kitchen on screens instead of paper, through orders management, takes that transcription step out.
8. Courier utilization
Orders delivered per courier-hour. This one ratio is the difference between $3 and $7 of delivery cost on every order. Target: 1.5–2.5, depending on how dense your city is. Below 1.5, look at dispatching and zone design. Above 3, you're probably running late, so check KPI 5 before you celebrate the efficiency.
Money and retention: turning it into profit
9. Delivery cost per order
Your true courier cost, counting wages, mileage, and idle time, divided by the orders that actually got delivered. Target: under 12–15% of AOV. This is the number that settles your zone fees and the perennial own-fleet-versus-third-party argument.
10. Effective channel cost
Every deduction a channel takes, divided by the gross sales it brings in: commission plus the ads, promos, and charged-back refunds that ride along with it. Count all of that and marketplaces routinely land at 30–40%. Target: keep your blended cost of sale under 15%. It drops a little every time an order moves onto a channel you own.
11. 30-day repeat rate
The share of this month's customers who come back to order within 30 days. It's the earliest honest read on whether your food and your delivery actually hold up. Operators tell us healthy brands sit at 25–40%. Under 20%, your growth problem lives in retention, not acquisition.
12. Refund rate
Refunds, remakes, and comped orders as a share of revenue. Target: under 1.5%. This is a symptom more than a root cause; it stacks up from accuracy, speed, and packaging, so when it climbs, one of KPIs 5–7 usually tells you which. There's a cause-by-cause fix list in our refunds breakdown.
We'll show you the dashboard on demo data — or on yours, if you bring an export.
The weekly review and the live safety net
These twelve make a good Monday-morning ritual: 15 minutes, each number read against your baseline, one thing to act on that week. A few of them, though, can't wait until Monday. Late orders, stuck orders, a sudden courier gap all need a response the same night. That's why we split the platform in two: a live monitoring panel for what's happening right now, and reports and analytics for the weekly trends. The dashboard tells you a zone slipped 8 minutes this week. The monitor tells you order #1142 is about to run late tonight.
One warning from experience: don't add a thirteenth number until one of these twelve has stayed green for three straight months. Dashboards usually die of bloat rather than neglect. What makes them worth keeping is simple: every number has an owner, a threshold, and a known next move the moment it turns red.
Do this next
Open a blank note and write out these twelve KPIs with two columns next to them: "current value" and "don't know." Everything that lands in the second column is risk you can't see. And the count of "don't know" rows is, honestly, the most useful KPI you'll calculate all week.
Live monitoring plus weekly analytics, per channel and per zone — launched in about two weeks.