What needs you today,ranked by money.
Alerts and proposed decisions land in one place, ordered by the money at stake — and they come to you, so nobody has to go hunting.
A brief in the inbox is opened and questioned: marql ranks the store against its like-for-like peer group, shows which categories carry the revenue gap down to single products, and lists the actions in priority order.
Decision inbox · demo network
today · 07:10
- ONGOING
- 6
- EARLIER THIS WEEK
- 3
- EARLIER
- 2
Margin down 4.2 pp on dairy
5 locations · money summed · gross margin
USEFULNOISEMARK READASK AI
Discount leakage above plan
Westside · average basket
Stock-out risk on 340 SKUs
North Park · stock-out share
Labour cost drifting up
Harbour · labour cost %
Lunch combo outperforming
City Center · repeat rate
Four steps, and none of them is a dashboard.
The point of the Inbox is that nobody has to go looking. The work arrives already sorted, and leaves with a name and a date on it.
Detectors read your data
Cloud connectors land data every few minutes, events become signals hourly, and the full scans and the decision cycle run daily. Honest intervals, not a claim of real time.
One problem becomes one row
A repeat of the same problem at the same location updates the row it already has instead of printing a new card, and when a whole family repeats across the network it collapses into a single line with the money summed.
Money and persistence decide the order
Rows are ranked by money at stake, severity and how long the problem has been running — and whatever is still firing sits in Ongoing with its day count, so a two-month state stops looking new every morning.
You take control of it
Assign an owner and a horizon of 3, 7 or 14 days and the row becomes a task with the metric captured before it. Decisions with a measurable target metric go on into the proof loop and end up in the Impact Ledger.
Scored against this location, not a flat rule.
MonTueWedThuFriSatSun
Tuesday, an ordinary day
Tuesdays run about 5% below the week's average here, and this one sits in the middle of its own normal. The tightened flat rule fires on it anyway — and on Wednesday and Thursday.
Saturday, 16% below its own normal
Saturdays run 16% above the average here, so a drop of the same size lands at an ordinary Tuesday's height. A flat −20% rule never reaches it, and tightening one until it does costs three false alarms a week.
A revenue row opens at 3σ from this location's own normal and at least 15% below it. Saturday is 16% and about 12σ. Illustrative figures.
Everything here exists to make the list shorter.
An inbox nobody can finish is a feed, and a feed gets ignored. Collapsing, grouping and ranking are the product, not the decoration.
Urgent first, standing conditions apart
Needs you now leads the list. A condition that keeps firing sits in Ongoing with the number of days it has been running, wins collect under Upside, and everything untagged falls into Today, Earlier this week and Earlier. Inside a section the biggest money is on top — the urgent lane keeps the order the scoring gave it — and what kind of problem it is rides on the row as a detector tag: margin, stock-out, cost.
What is genuinely new stops being buried under what has been true for two months.
Out of stock at Madrid · Malasaña — 1 SKU
Margin drop at Madrid · Retiro
Cash banking shortfall at València · Ruzafa
15th day running
Purchase cost rising: Zumo de naranja
29th day running
An owner and a date
Taking control assigns an owner role and a horizon, creates the task, and captures the metric before and after it.
Every decision has an owner against it, and you can see what happened after.
Ampliar la ventana de medición de tráfico
transactions · +5.1% predicted · 2 805 €
owner: operations lead · task created
inconclusive — the task closed 12 days past its date
16
proposed
1
in progress
3
measured
It comes to you
A morning email once your organisation switches it on, and browser push for anything critical once you allow it in your browser — including a single reminder if a critical row sits 72 hours with nobody against it.
You find out without opening a dashboard.
stale-critical reminder
critical · open 72 h · no task against it → one push to owners
once per incident · at most 3 a day
Exactly how much control it takes.
An operator who has been sold automation before wants to know exactly how much control it takes. This is the honest answer.
It proposes; it does not write to your POS.
Prices, orders and stock stay in the systems that own them.
Hands-off acceptance is off by default.
Switch it on and it is still capped: a fixed list of decision types, a money ceiling, a daily limit and a kill switch.
Detectors are switched on per organisation.
What a new network sees on day one is the core set, not the full catalogue.
A row with no measurable target metric stays a task.
It is tracked and assigned, but it does not enter the proof loop.
The Decision Inbox is in beta.
It is labelled as such inside the product.
The questions that come up in the first demo.
A threshold alert fires on a flat rule — the same number for a kiosk and a flagship. Here each location is scored against its own normal, which the section above draws. Rows are then ordered by money, collapsed to one per problem, and most carry a proposed action rather than only a notification. Your own threshold rules can still be added on top, and they arrive in the same place.
Cloud connectors bring data in every few minutes. Events turn into signals hourly. The full scans, the network patterns and the decision cycle run daily, and some pattern detection is weekly. We publish those intervals rather than calling it real time, because an operator who is told 'instant' and sees a four-hour-old number stops trusting everything else on the screen.
Because the design target is a list you can finish — the collapsing above is most of the answer. Long-running issues also move out of the way of what is new, and the cap limits rows rather than incidents, so one loud detector cannot take the whole list. Nothing is hidden: every participant keeps its own detail page, and muting or retuning a detector stays a human decision.
Roles decide it. An owner and a COO see the network; an analyst has read-only access; a store manager sees their own locations; a franchise partner sees theirs, read-only. Read-only roles see the state of a decision — who owns it, when it is due — without the controls to change it.
Yes. Signals are stored language-neutral and rendered per user in six languages, including Romanian and Bulgarian, so the same row reads correctly for a head office and a local manager.
It becomes a task with an owner and a due date, and the location's metric is captured as it stood at that moment. When the horizon closes, the metric is read again. Where the row was a measurable decision, the outcome goes through the proof loop into the Impact Ledger — and a task that was never completed is closed as inconclusive rather than credited.
Ready to see
the euros it found?
The walkthrough is built around the tills and accounting you already run. Leave your details and pick a time.