Rider management and delivery times: the before and after of counting every minute

Step 1: measure the three time blocks separately for fourteen days

Break the delay into three numbers before you change anything, because a restaurant that believes it has a courier problem almost always has a misattributed kitchen problem. At the Peruvian spot in Chapinero billing 41 million pesos a month on Rappi, with a 3.9 operational score and 18 % of orders flagged as late, pulling tickets with acceptance time, ready time and pickup time split those 15 minutes of delay like this: 9 from the kitchen, 4 from the rider waiting at the door, 2 on the road. What you must have at the end of this stage is a sheet with fourteen days of those three fields plus the Friday peak average, the only day that really matters. Verify it by comparing your real dispatch average against the prep time the platform currently shows; a gap above 4 minutes explains your rating. Raise your declared prep time to your 80th percentile of real dispatch, not to your average, and do it on a Monday so you get a clean week of measurement.

Step 2: declare an honest prep time, even if the number stings

Platforms dispatch the rider using that figure as reference: declare 12 minutes while you dispatch in 21 and the courier shows up at minute 14, stands still for 7 minutes, and those 7 land in the courier wait metric, precisely the one the assignment algorithm punishes. The algorithm does not rank the promise, it ranks the deviation between promise and delivery. A kitchen promising 22 and hitting 22 climbs above one promising 12 and delivering at 21. This is done when the number in your dashboard equals your measured 80th percentile, and you verify it seven days later by checking whether average rider wait dropped from 7 minutes to under 3. Put a screen or a board where the cook can see the rider ETA, and set the rule that hot plating starts only when the courier is less than 5 minutes away. Here sits the tension almost no owner resolves: cutting times pushes you to cook early, food cools down and your food rating drops; stretching them protects flavour but drops you off the proximity ranking.

Step 3: install a sync light on the hot line

The bridge is not speed, it is SYNCHRONIZATION. Split preparation into two blocks — what survives twenty minutes without degrading and what dies in five — and fire the second block off the light. At that Peruvian kitchen in Chapinero, this single split removed 6 of the 9 minutes the kitchen was contributing. Verify by measuring exit temperature on three random orders per shift for a week. Rebuild the assembly station so your ten best-selling delivery dishes get put together without crossing the kitchen, because the minute lost walking never shows up in a platform report but does show up in your rating. Pull the item ranking for the last ninety days, count how many steps and how many hands each one takes, and keep pre-cut, pre-portioned and within arm's reach everything that makes up 70 % of volume.

Step 4: redesign mise en place around your delivery top 10

Delivery is not a side channel you staff with dining-room leftovers: food delivery apps moved USD 110 billion in 2024, up 15.5 % year over year (Business of Apps, Food Delivery App Report 2025), and high-performing kitchens run margins of 10 to 30 % against the 3 to 5 % of a traditional restaurant (OysterLink 2025). Your deliverable is a signed station layout and a measured assembly time per item. Create a rider handoff point separate from the guest entrance, with a table, a visible order number and one person responsible per shift. It sounds minor and it is not: between hunting for parking, walking into the dining room, queuing behind two customers and waiting for someone to locate the bag, 2 to 4 minutes disappear per order, which at 80 daily orders adds up to more than three hours of accumulated wait per week that the algorithm bills to you.

Step 5: negotiate the pickup point and shave seconds at the door

Mark the spot with physical signage, state the location in your merchant notes inside the platform, and hand over with double number verification. It is done when the rider walks in, reads, grabs and leaves without asking anyone anything. Time ten consecutive pickups: if door average falls from 3 minutes to under 60 seconds, the point is properly built. Costliest of all is lowering the prep time the moment your numbers improve, chasing ranking, and landing back at the starting point within ten days. Second comes accepting every order during peak without pausing your slowest item: better to shut down two dishes for thirty minutes than to blow the promise on fifty tickets. Third is cooking by arrival order instead of by committed pickup time, a habit cured by printing the committed hour on the ticket. Fourth is watching monthly averages, which hide exactly the Friday wrecking your score. Diego F.

Four mistakes that blow up this guide by week three

Parra hammers this point with the owners he advises through Masterestaurant: managing riders and delivery times is a positioning lever rather than a logistics matter, because whoever delivers in 25 minutes shows up at the top while whoever delivers in 45 vanishes from the scroll even with better food. Run inaction all the way out and decide with that figure in front of you. A kitchen holding 18 % late orders loses weight in dispatch, receives fewer rider offers, stretches the door wait and feeds the same loop: less visibility, less volume, identical fixed payroll. On 41 million pesos a month, a conservative 20 % drop in delivery volume means over 8 million pesos monthly walking out while no cost goes down, since rent and payroll never hear about your ranking. At the opposite end, the full route in this guide took total delivery from 44 to 29 minutes in four weeks without hiring anyone.

What happens if you do nothing: the ninety-day scenario?

Fifteen minutes nobody bought: they came out of the kitchen itself. That is the whole difference between arguing with couriers and watching the clock on your own line.

Call the implementation finished only once these six markers hold for two straight weeks, not for one good day. Declared prep time equal to your measured 80th percentile, gap under 2 minutes. Rider door wait below 3 minutes on the Friday average. Orders flagged as late below 8 %. Operational score above 4.4. Assembly time measured and documented for your top 10 items. And a named person per shift owning the pickup point. Print those six numbers, tape them to the kitchen door and review them every Monday with your line chief for a month; after that the first and the third are enough. Should any of them drift out of range two weeks running, go back to step 1 and measure the three time blocks again, because the diagnosis has expired.

Five differences that move the needle

The first difference is accounting, not effort. A restaurant without time management owns ONE number; one with it owns three, and those three point at different departments. Once the 15-minute delay splits into 9 of kitchen, 4 of wait and 2 of trip, the conversation stops being a coffee-break rant about lazy couriers and becomes a mise en place decision. Declaring an honest prep time feels like losing positions and does the opposite. Delivery platforms do not rank by the promised number, they rank by how well you keep it — the internal metric is deviation, not the absolute value. A store that promises 22 and delivers in 22 beats one promising 12 and shipping in 21. The sync signal resolves the cooling paradox. Nobody wants to hand over a burger that spent 11 minutes under the lamp, and nobody wants a courier staring at the ceiling either. Firing the pass with the rider under 5 minutes out drops both clocks at once: food leaves at 68 °C and the rider waits under two minutes.

Five differences that move the needle — in practice

Auto-accept needs a brake, and here I was wrong for years recommending it unconditionally. Accepting everything during a saturated shift does not raise sales, it raises cancellations: one order lost to delay weighs more on the score than three delivered well. A load cutoff at 8 tickets protects ranking better than any campaign. Then comes the front almost nobody works: reviews. A reply that blames the courier tells the next reader you do not control your own operation. A reply that names the adjustment —we changed the prep time for the 8 p.m. daypart— turns the complaint into proof that somebody is steering the ship.