Control Room · Ops

How to recover a delivery app rating when one location drops

A location's rating is not recovered by asking for 5 stars. It is recovered by closing the operational cause in 48 hours — times, packing, wrong orders, or 86s — with a per-store threshold and one shift owner. The chain average is makeup: the customer scores a store.

By Control Room · 2026-08-30 · 4 min read

48hwindow to stop the drop

1location. Never average the chain

4causes that actually move stars

What moves rating

Time / cold food1st

Wrong / incomplete order2nd

Packing / spills3rd

86 / missing item4th

Illustrative. Pull the 20 low reviews from the red location and tag them: no cause, no playbook.

When one location's rating drops on Rappi, Uber Eats, PedidosYa, or DoorDash, the usual reflex is to ask for reviews or blame “the app.” Both arrive late. The customer already ate cold food, got an incomplete bag, or waited too long. Recovering a delivery app rating is an operations problem, not a reputation campaign.

What a rating drop is (and what it isn't)

A rating drop is when one store's score — not the brand's — falls in a short window, not a single 1-star review. The number that matters is that location, on that app, on a 7-day and 30-day window, not the chain's historic average.

Rating is a lagging indicator: it reflects what already happened in the kitchen, at packing, and at handoff. If you look at it once a month, you arrive when the store is already losing visibility. If you look at it every morning next to prep times and cancellations, you can still stop the bleed.

Never average the chain

The most expensive multi-location mistake is comforting yourself with the average. A 4.6 chain can hide a 4.1 store that is dragging that point's ranking — and sometimes neighbors that share a zone.

• Watch rating by location and by app, not one KPI.
• Compare 7 days vs 30 days: if the short window is worse, the drop is now.
• Cross rating with volume: 12 bad reviews on an 80-order store is not the same as 12 on a 15-order store.

If you cannot say in 10 seconds which store is red and since when, you do not have control. You have an average.

The 4 causes that actually move stars

Before the playbook, tag. Pull the last 20 low reviews for that store and classify them. Without a cause, the team “improves service” in the abstract and the number does not move.

• Time / cold food. Long prep, stacked tickets, rider waiting, food that travels badly. Most frequent, most measurable.
• Wrong or incomplete order. Extra item, missing item, the wrong sauce. That is a pass error, not “the app.”
• Packing. Spills, open bags, no utensils, crushed dessert. The customer never sees your kitchen: they see the bag.
• 86 / missing item. The order was accepted when it could not be fulfilled. The 86 has to be a data point, not a shift rumor.

Rider complaints exist, but ops moves first on what you control inside. If 70% of the lows are time or packing, the war room is not against the fleet: it is against the pass.

The 48-hour playbook

The goal of the first 48 hours is not “get back to 4.7.” It is stop new reviews of the same type. If you keep producing the same error, asking for stars is noise.

Hour 0–2: isolate the store

• Confirm the red is that location, not a misread average.
• Rank causes from the last 7 days of reviews.
• Name one shift owner (not a committee). One. With a phone.

Hour 2–24: one lever, not ten

Pick cause #1 and move only that:

• If time: cut the peak menu, preventive 86 on the slow items, one packing runner.
• If wrong order: printed ticket + a 5-second pass check (items, sauces, utensils).
• If packing: change the failing package and do not dispatch in the peak without a closed-bag check.
• If 86: pause the SKU in the app in the moment — do not “tell the kitchen.”

Hour 24–48: verify with orders, not hope

Published rating is slow. What does move in 24–48 hours is the raw material of rating: prep time, % incomplete orders, stock cancellations. If those three do not improve, rating will not improve next week.

Threshold: when to open a war room

Set the threshold before the crisis. Examples that work in ops (tune them to your network):

• A store's 7-day rating drops 0.2 points vs its own 30-day.
• Or it crosses a hard floor (the number where that app starts hiding you).
• And there are more than N low reviews in 72 hours, so one outlier does not page everyone.

The war room runs until cause #1 stops showing up in new orders, not until the historic average ticks up. The historic average is slow on purpose.

Mistakes that stretch the drop

• Asking for 5 stars while you still dispatch late. The customer rates what just happened.
• Averaging locations and declaring “the brand is fine.”
• Opening 8 fronts the same day: menu, packing, staffing, ads. Nothing closes.
• Watching rating weekly and times never. Rating is the last to find out.
• Defaulting to “the rider” without looking at door wait minutes.

Frequently asked questions