← All posts

Case study / 5‑minute read

A missing item rates worse than an hour-late order, because the guest finds out at home

Late food is discovered while you can still fix it. A missing side is discovered in a kitchen eight miles away, with the bag open and nothing to be done — which is why one forgotten item costs more rating than a disastrous wait.

The operator came to this backwards. They had assumed their off-premise problem was speed, because speed is what everybody assumes, and they had spent a quarter on it — a second expo screen, a reorganised shelf, a revised quote time. Collection times improved by four minutes.

The rating did not move.

When they finally sorted the off-premise reviews by what the guest was actually complaining about, the picture was not close:

Complaint Reviews in 90 days Mean rating
Something missing from the order 23 1.4
Wrong item sent 8 1.9
Food cold on arrival 17 2.6
Long wait at collection 31 3.1

The long waits were the most frequent complaint by a distance and the least damaging per review. The missing items were a third as common and were pulling nearly two full stars lower.

Thirty-one waits at 3.1 stars and twenty-three missing items at 1.4 stars contribute almost exactly the same amount of damage. One of them had received a quarter of operational attention. The other had received none, because it had never been counted separately.

Why a missing item is the most expensive error in off-premise

It is worth being precise about this, because the intensity seems disproportionate until you look at the mechanics.

It is discovered at the point of no remedy. Every other failure is discovered while you can still act. Food is late — you are there, someone can apologise, something can be comped. A missing side is discovered when the bag is open on a kitchen worktop at home, eight miles away, with the family already sitting down.

The remedy costs the guest more than the item. To put it right they must ring you during your busiest hour, explain, be put on hold, and then either accept a credit for a future visit they did not ask for or drive back. For a £4.50 side, every available remedy is worse than the loss. So they do not ring. They write a review instead, which is the only action available that costs them nothing.

It breaks the meal rather than degrading it. A cold main is a worse version of dinner. No rice with a curry for four is not dinner. One missing component can make the rest of a £60 order unservable, which is why the reviews are so often about a cheap item.

It reads as carelessness, not pressure. A guest can see that a kitchen gets slammed on a Friday. Nobody believes that being busy causes a bag to be sealed without the chips in it. Late is forgivable because it is legible; missing is not, because it looks like nobody checked.

The reviews make all four points, usually in two sentences:

No rice. A curry for four and no rice. Rang twice, nobody picked up. £58 and we ate toast.

Missing both kids' meals. Couldn't go back, they were already in pyjamas. Absolute waste of an evening.

Notice what neither mentions: how long it took, or the quality of what did arrive.

Where the items were actually going missing

Thirty-one orders audited at the point of sealing, Friday and Saturday:

Nothing was being forgotten at the plating stage. The mains were right in thirty-one out of thirty-one. The errors were all in the assembly of things that come from somewhere other than the pass:

  • Drinks, from the fridge behind the bar — 7 of the 9 errors
  • Sides finished separately — rice, naan, chips held in a warmer across the kitchen
  • Desserts, from a different fridge, sometimes added by a different person
  • Sauces, dips and cutlery, which the till lists as items and nobody treats as items

A single pattern explains almost all of it. The person bagging an order could see the mains and could not see anything else. A bag containing two mains looks complete. The ticket says it is not, but the ticket is reconciled from memory while three other orders are being assembled.

One more contributing factor, which was the reason Fridays were worse: when an order was part-assembled and set aside waiting for one component, the bag got sealed by whoever came back to it, which was frequently not the person who knew what was outstanding.

The fix, which has nothing to do with checking harder

Telling people to check the ticket is the intervention everyone tries and it does not survive a Friday. Three structural changes instead:

1. Nothing gets sealed until it is complete, and incomplete orders cannot be closed. The bag stays open, rolled down rather than taped, and a physical clip holds the ticket to the opening. An open bag is a visible signal that something is outstanding; a taped bag asserts completeness. This one change caught most of the set-aside errors.

2. The items that come from elsewhere go in first. Counter-intuitive, and it worked. Drinks, sides and desserts are collected and placed in the bag before the mains, because they are the ones at risk of being forgotten. The mains are what everybody remembers; loading them last means the bag is never sitting there looking finished.

3. One person seals and reads the ticket out loud. Not a second checker — they did not have the labour for that and most sites do not. The same person, saying the item names aloud while touching each one. It sounds faintly ridiculous and it is the standard error-reduction technique in every industry that cannot afford mistakes.

One supporting change: sauces, dips and cutlery were moved onto the ticket as numbered lines rather than appearing as modifiers in small text. An item you have to read in a different font is an item that gets missed.

What changed

Eight weeks, same site, same staffing:

Before (90 days) After (56 days)
Missing-item reviews 23 3
Mean rating of off-premise reviews 3.0 4.1
Collection time, order to handover 23:00 23:40
Comps and refunds for missing items £1,340 £210

Forty seconds slower per order. Twenty missing-item reviews avoided, a full star on off-premise, and about £1,100 a quarter back in refunds that were previously being handed over for forgotten rice.

The forty seconds is the real trade, and it is the reason this does not get done. Every change above makes the bag take slightly longer, and in the middle of a Friday rush that cost is immediate and visible while the benefit is a review that does not get written three days later.

What to do with your own numbers this week

Twenty minutes, and it is the whole point of this piece:

  1. Pull ninety days of off-premise reviews.
  2. Sort them into four piles: missing, wrong, cold, late.
  3. Average the star rating of each pile.
  4. Multiply count by the gap from your target rating.

Most operators find the ranking is not what they expected, and nearly all find that late is their most common complaint and their least damaging one. That is the finding that redirects the quarter.

Where this is not your problem

If your missing-item count is already low, the ordering above will look different and your attention probably does belong on timing. The notification itself is a surprisingly common culprit there — pressing "ready" before the order is complete converts a short wait into a betrayal.

Third-party couriers complicate attribution. A bag sealed correctly and delivered to the wrong address produces a review that reads identically to yours. If your missing-item reviews are concentrated on one platform and your audit at the point of sealing comes back clean, that is a different conversation with a different counterparty.

Cold food is not usually a kitchen problem either, and it has its own distinctive cause — more often the plate and the route than the stove.

Read-aloud decays. It was still happening at eight weeks because the bag could not be sealed without it. Any version relying on diligence alone will be gone by the second busy Saturday.

OMMU splits off-premise complaints by failure type and tracks the mean rating of each, which is the table that reorders your priorities. Ninety reviews and four piles on a desk produces the same answer — the error is treating "bad takeaway reviews" as one thing when the four causes differ by nearly two stars.