A dashboard tells you the shelf is empty.
It does not tell you who was asked to fix it, whether they did it, or what they said when they didn't. That gap — between knowing and someone actually having done something — is where retail margin goes.
Retail is on our roadmap and not yet built. This page says exactly what exists today, what does not, and what closing that gap would involve. We would rather you read that here than discover it in month three.
Every part of the chain optimises locally while the group underperforms globally.
Buyers optimise margin. Planners optimise stock cover. Distribution centres optimise throughput. Store managers optimise the floor in front of them. Suppliers optimise their own order book. Each is rational on its own terms, and each has valuable information the others never see.
So a shelf sits empty while the same item sits in a distribution centre forty kilometres away. A promotion runs at three quarters of the estate and nobody knows which quarter missed it or why. A markdown gives away margin that a transfer would have saved. Shrinkage is recorded accurately, monthly, and acted on by nobody in particular.
None of that is a reporting failure. The numbers are usually correct. It is a failure of the step after the number: the instruction, the owner, and the record of whether anything actually happened.
A retail group is not a set of departments. It is one system that nobody can see whole.
Sourcing decides what arrives. Supply chain decides where it goes. Stores decide what happens to it in front of a customer. Strategy decides what it should have cost. Each holds information the others never see, and the losses collect in the space between them — in decisions that belong to no single function and therefore to no single person.
Seven functions, each optimising well on its own terms. The lines between them are where the money is lost — and where nobody is accountable, because the decision belongs to no single function.
Platforms are sold to retailers as whole-enterprise transformations covering every function at once. We are not that, and pretending otherwise would waste your time in the first meeting rather than the fourth. Here is each function and what we would say about it.
SUPPLY CHAIN & LOGISTICS
Out-of-stock that becomes an owned instruction rather than an alert. Supplier issues routed to a named person with a deadline. The loop itself is built and running in our supply-chain work today; retail's vocabulary is what is missing.
STORE OPERATIONS
Allocation, loss prevention, price and promotion compliance. Every one is a decision made repeatedly, executed by someone far from head office, and almost never reconstructable afterwards. This is the sharpest fit for what we have built.
SOURCING
Supplier performance becomes instructions that close against a recorded outcome, so adherence is computable. We do not do raw material planning or sourcing strategy, and we would not pretend to.
STRATEGY & PLANNING
Demand forecasting, assortment simulation, scenario planning. Genuinely valuable, genuinely not what we have built. You have vendors for this and they are better at it than we would be.
MARKETING & SALES
Campaign optimisation, segmentation, dynamic pricing. Outside our lane entirely.
CUSTOMER EXPERIENCE
Loyalty, omnichannel, unified customer view. Outside our lane entirely.
SUSTAINABILITY & CIRCULARITY
An append-only ledger is a natural fit for emissions and waste attestation, since a record written at the time cannot be quietly revised later. We have not built it, so we are listing it as an idea rather than an offering.
Allocation under scarcity
Limited stock, more stores than units. The decision is made every week, usually reactively, and afterwards nobody can reconstruct why this store got twelve and that one got none.
The empty shelf with stock in the network
Not a forecasting problem — a routing and follow-through problem. Somebody must be asked to move it, and the asking must be recorded, owned and closed.
Markdown authority
Margin given away is a decision with a signature on it, or it should be. Who approved this cut, against what alternative, and what did the last hundred of these actually recover?
Shrinkage with a name attached
A loss number nobody owns is a report. A loss that raises an instruction to a named person, with a date and a recorded outcome, is a control.
Price and promotion compliance
Head office sets one price. Two hundred stores implement it in two hundred ways. The gap between what was instructed and what is on the shelf is measurable, and almost nobody measures it.
The override
A store manager overrides the replenishment system, and is often right. Without a record of who overrode what and how it turned out, the organisation learns nothing from either the good calls or the bad ones.
The loop those six decisions need is already built, in our supply-chain work. An exception arrives carrying the action it recommends. Issuing it creates an instruction owned by a named person against an SLA. Closing it records an outcome from a fixed list rather than a comment box — which is what makes “83% executed as instructed, and here are the three systematic reasons for the rest” a number you compute instead of a feeling you have.
- Alerts that carry an executable recommendation
- Instructions with a named owner and an SLA
- Closure against a controlled vocabulary, so adherence is computable
- Every change written with its audit record in the same transaction
- Approval thresholds, where the approver cannot be the proposer
- Your own database, with access enforced by the database itself
- A model that acts through the same gate as a person, or not at all
Retail's vocabulary. None of it is declared, and until it is, this is a roadmap page rather than a product page.
- Store, region, format and trading calendar
- SKU, range, planogram and space
- Distribution centre, transfer and replenishment order
- Promotion, price change and markdown authority
- Shrinkage event, count and investigation
- The bindings to your point-of-sale, ERP and warehouse systems
That list is a declaration, not an engineering programme. It is the same exercise that already stood up a hospital ward and a wastewater treatment works on one unchanged core.
Not a whole-enterprise transformation, not another dashboard, and not a replacement for your ERP.
We are not proposing to model your entire enterprise or simulate it. You already know your out-of-stock rate; adding a fourth screen that also knows it changes nothing. What is missing is the step afterwards — an instruction that exists, belongs to someone, has a deadline, and closes with a recorded outcome that can be counted later. Your ERP keeps running and keeps being the system of record. This sits above it and holds the decisions it was never designed to hold.
And it is deliberately unglamorous. Retail in this country also runs through power interruptions, cold chain at risk, and store networks where the person closest to the problem is furthest from the system. A recommendation nobody is accountable for is worth nothing on a Saturday in a store with no power. An instruction with a name and a deadline is worth something.
Retail gets built in the order serious conversations arrive.
Bring one of the six decisions above — whichever one costs you most and is hardest to reconstruct afterwards. Thirty minutes, the live system running the loop for a different industry, and a straight answer about what declaring retail would take.