Where errors are born
Four doors a wrong order walks in through
Almost every wrong order enters through one of these. Knowing which door dominates your own operation is the difference between a targeted fix and a lecture to the kitchen.
- 01
Menu copies that drifted apart
The website, the app and each delivery aggregator hold their own copy of the menu, edited on different days by different hands. A dish renamed on one channel, a spice level that exists on another, an item still orderable where it was withdrawn — the guest orders from a menu the kitchen no longer runs.
- 02
Modifier chaos
"No onion", "extra sauce", "split the meal" — each channel expresses modifiers in its own vocabulary, and the mapping into the POS flattens or drops them. The order arrives technically complete and practically wrong, and no one in the kitchen ever saw the request that was lost.
- 03
Kitchen display gaps
Items on one ticket split across stations without a way to reassemble them; a ticket bumped as done while a drink is still in the fridge; a modifier printed in a corner nobody reads at speed. The screen is where an order is either kept whole or quietly loses a piece.
- 04
The handover at the pass
The food is right and the bag is wrong: two orders sealed at the same minute swap couriers, a bag leaves before its second bag is ready, the ticket never said the order had six items. The last metre of the operation undoes everything the first thirty did correctly.
Diagnose first
Count the misses before paying to fix them
The instinct when ratings dip is to buy something — a new kitchen screen, a checklist app, a consultant's training day. The disciplined move is cheaper and comes first: build a taxonomy from the evidence you already hold. Refund reasons, remake logs, aggregator dispute exports and the free-text complaints guests leave all say which door each error came through, once someone classifies them. A fortnight of that classification, done honestly, usually redraws the problem: the operation that was sure it had a kitchen discipline issue discovers most misses are modifier mappings and bag assembly, which are fixed by configuration and ticket design rather than by exhortation.
The taxonomy also sets the baseline that makes every later fix accountable. Change one thing — a modifier vocabulary, a pass procedure — and read the same classification a month later. Without that reading, accuracy work is a sequence of purchases each declared a success by the people who proposed it; with it, the operation knows which fixes paid and which door to work on next.
The last metre
From the pass to the courier without losing an item
Delivery moved the moment of truth out of the dining room and into a sealed bag. These are the checks that keep an order whole across its most error-prone metres.
A flow of four checks — a ticket built for packing, a counted completion check, seal and label with bag count, courier matched at collection — with a branch for a defined substitution path when an item runs out.
- 01
A ticket built for packing
The ticket states the item count and groups the order the way it will be bagged — drinks, sides, sauces listed as countable lines, not footnotes.
- 02
Completion, checked not felt
The bag is closed against the ticket's count, and the check is recorded — the difference between "it felt complete" and "six of six".
- 03
Seal and label
A sealed bag labelled with the order and its bag count — "bag 1 of 2" is a sentence that prevents a whole category of missing-item disputes.
- 04
The courier matched, not assumed
Collection confirms the courier against the order — a code, a scan — so two couriers arriving in the same minute cannot swap dinners.
When an item runs out mid-order
The escape hatch is a decision, not an improvisation: a defined substitution and refund path, offered to the guest before the bag leaves — because the silent substitution is itself one of the four doors.
Our role
The instruments we install and the habits the line keeps
- 01
The taxonomy and its baseline
We classify a period of your refunds, remakes and disputes by entry point, so the programme starts from your errors rather than from a generic playbook.
- 02
The systems fixes
Menu governance, modifier mapping, ticket and kitchen-screen design, the handover checks — configured and built into the systems you already run wherever they can carry it.
- 03
The reading a month later
The same classification, re-run after each change, so the operation renews or redirects each fix on evidence — accuracy work that cannot show its own effect is theatre.
The pass procedure, the packing habit and the standard the shift holds are the operation's to keep; we build the tickets, mappings and checks that make the right way the easy way.
Common questions
What buyers ask
Complaints say items are missing from delivery bags. Is the kitchen to blame?+
Less often than the complaint suggests. A missing-item dispute records where the failure was discovered — the doorstep — not where it entered. In practice the miss is frequently upstream of cooking entirely: a modifier that never reached the ticket, a drink that lived on a different station's screen, a second bag that stayed on the pass, two orders swapped between couriers at collection. That is why we insist on classifying a period of disputes before changing anything: if the dominant entry point is ticket structure or courier matching, retraining the kitchen fixes morale and nothing else. The bag-count and courier-match checks in the handover flow eliminate whole categories of these disputes without asking anyone to cook differently.
Why do order errors climb every time we add a delivery channel?+
Because each channel adds a menu copy and a modifier dialect, and both multiply the ways an order can silently change shape. A new aggregator means another place where an item can keep an old price, another vocabulary for "no onion" that must be mapped into your POS, and another integration where a mapping gap flattens a request. None of this is an argument against the channel; it is an argument for joining channels through governance rather than by copy-paste — one master menu projected outward, a controlled modifier list mapped once per channel, and a parity audit that catches drift before guests order from it. Operators who do that groundwork add channels without adding a new class of errors each time.
Do we need new systems to cut order errors, or better settings in the ones we own?+
Run the taxonomy first, and let it answer. In a large share of operations, the biggest error sources are governance and configuration: menu copies edited independently, modifiers unmapped, tickets structured in the order items were tapped rather than the order they are packed, completion left to feel rather than count. All of that is fixed inside the systems already owned. New software earns its place at specific gaps the taxonomy exposes — a kitchen display that genuinely cannot group an order, a POS integration that genuinely cannot carry a modifier — and buying it before the classification exists means paying to move the problem rather than to remove it.
In this cluster
Trace your errors with us
A diagnostic that classifies a period of your refunds, remakes and disputes by entry point, and names the two fixes with the largest share.
Start the conversation
