- Market
- Saudi Arabia
- Written for
- COO, Head of Digital
- Languages
- EN · AR
The drift
Four ways one order stops meaning the same thing everywhere
Each failure below looks like a one-off mistake at the branch. Seen together they are one condition: several systems holding copies of the same facts, with nothing appointed to keep the copies honest. Wrong items packed in the bag is a different failure with its own page; this one is about channels that no longer agree.
- 01
Prices that diverge by channel
A price is changed at the till, or a Ramadan menu goes live on the website, and the aggregator copies are left where they were. The gap survives until a guest, an auditor or a settlement file notices — and by then it has been charged hundreds of times.
- 02
The stock-out race
The kitchen runs out of a dish at eight; the channels keep selling it until someone finds the right tablet and the right setting. Every order taken in that window is a refund, an apology and a guest who ordered dinner and received a phone call.
- 03
Status updates that mislead the guest
The guest's screen says preparing while the ticket never reached the kitchen, or on its way while the bag is still at the pass. A status invented from timers instead of events reads as reassurance until the doorbell fails to ring, and then it reads as a lie.
- 04
The order retyped at the counter
Where a channel is not wired into the POS, staff read orders off a tablet and key them in again. From that moment two records of the same order exist, and every void, discount and end-of-day count inherits the difference between them.
The architecture
One menu record has to outrank every copy of it
Channel drift is not cured by discipline. A chain can brief every branch to update every tablet, and the drift returns within a month, because the architecture invites it: five systems each accept edits, so five versions of the menu exist by design. The durable fix inverts the arrangement. One record — items, prices, translations, availability — is appointed the source of truth; every channel becomes a subscriber that receives updates and is forbidden to be edited directly. From then on a price change is one edit that propagates, not five edits that race.
Ownership matters as much as the record. The master menu belongs to the operation — a named role that approves changes — not to whichever vendor's console happens to be open. A POS can hold the record, a headless menu service can hold it, and either works; what does not work is ambiguity, where marketing edits the website, the branch edits the till and the aggregator copies belong to nobody. The synchronisation layer we build enforces the hierarchy the operator chooses: it does not choose it, and it cannot save an operator who declines to choose.
Status deserves the same treatment as the menu. What the guest is shown should be assembled from events the systems actually emitted — ticket printed, kitchen fired, marked ready, courier collected — with honest gaps where no event exists, rather than a progress bar animated on a timer. A guest tolerates being told the kitchen is busy far better than being told a comforting fiction that the doorbell then contradicts.
The pipeline
The route a menu change takes to reach every channel
Once one record outranks the copies, publishing becomes a pipeline with observable steps — and drift becomes an alarm instead of a discovery.
A pipeline: edit the master record, validate, publish to each channel, confirm and reconcile, watch for drift — with a branch for marking an item out of stock everywhere in one action.
- 01
Edit the master record
One place accepts the change — item, price, Arabic and English names, availability window — under the role that owns it.
- 02
Validate before publish
Missing translations, prices that fell to zero, items with no photo — caught here, where a mistake costs a correction rather than a refund.
- 03
Publish to each channel
App, web and each aggregator connection receive the change through their own interface, translated into each platform's shape by the layer, not by staff.
- 04
Confirm and reconcile
Each channel's acceptance is recorded; a channel that rejected or timed out is retried and flagged, so a silent failure cannot masquerade as success.
- 05
Watch for drift
A scheduled comparison reads each channel back and diffs it against the master, because the honest assumption is that drift will try again.
The one-action stock-out
When the kitchen runs out, one action in one place marks the item unavailable everywhere — the race between the kitchen and the channels is ended by design, not won by whoever reaches the tablets fastest.
Our role
The synchronisation layer we build and the record you govern
- 01
The drift audit
We read every live channel on the same day and put the differences in front of you — prices, items, availability, status behaviour — because the scale of the drift, seen once, settles the argument about whether the work is worth doing.
- 02
The canonical model and its pipeline
We build the master record, the validation, the per-channel publishing and the read-back comparison — wired to the interfaces each platform offers merchants, with retyping removed wherever a connection exists to replace it.
- 03
Status told from events
We map what the guest sees to what actually happened — fired, ready, collected — and where a stage emits no event we show an honest range rather than an animated guess.
Who may edit the master menu, and which channel wins when history disagrees, are operating decisions we help you name and then enforce in the layer — the architecture carries the policy; it does not replace it.
Common questions
What buyers ask
Why do our aggregator prices not match the till?+
Because they are separate copies, each edited on its own schedule. A price change made at the POS does not travel anywhere by itself; each aggregator copy waits for someone to log into that platform's merchant console and repeat the edit, and under operational pressure some copies are missed. The mismatch is therefore not a fault in any one system but a property of the arrangement — several editable copies, no appointed master. The fix is architectural: one record owns prices, the channels subscribe to it through the interfaces the platforms provide, and a scheduled comparison reads each channel back so a missed update becomes an alarm the same day instead of a discovery at settlement.
How do we stop channels selling a dish the kitchen has run out of?+
Make the stock-out one action instead of five. Today the kitchen's knowledge that an item is finished has to travel by voice to whoever holds each tablet, and every minute of that journey sells food that does not exist. Wired properly, marking an item unavailable in one place propagates to the app, the website and each aggregator connection automatically, and the same action reverses when the item returns. Two details decide whether it works in practice: the action must be reachable from where the kitchen actually stands, in one tap, or it will be skipped mid-rush; and the propagation must be confirmed per channel, because a stock-out that silently failed to reach one platform recreates the original problem on the channel you check least.
Should the POS or the ordering platform hold the master menu?+
Either can, and the choice matters less than making it. The POS is a natural master where it already holds items, prices and modifiers well and offers interfaces the other channels can subscribe through; a separate menu service is the better master where the POS is old, where several POS brands coexist across branches, or where the menu carries content the till has no field for — photography, translations, channel-specific descriptions. What matters is that exactly one is appointed, that every other surface becomes read-only against it, and that the decision names a person as well as a system: a master record with no owner drifts more politely than five copies, but it still drifts.
In this cluster
Trace the drift with us
A working session that inventories every place a menu or an order is edited today and scopes the single record the channels would subscribe to.
Start the conversation
