Operating model · Free · No call required
The Multi-Location Retail Operating Model
Multi-location retail marketing fails on an ownership seam far more often than on creative. This is the operating model underneath it: who owns the join between the catalog and the store map, the four moves that stop spend landing on what a store cannot sell, and the store-tier protocol that survives a district manager disagreeing with it.
Published August 30, 2026· 10-minute read · Written so you can run it without hiring anyone
The ownership split
Company-owned retail has no franchisee to negotiate with, which is why operators assume it has no governance problem. It has a harder one: three functions each hold part of a decision, and none of them owns the decision.
Merchandising owns what is true
The item master. Which SKUs exist, what they are called, what they replace, and when they are discontinued. On-hand by location, and the refresh cadence of that number. Markdown state and the calendar behind it.
Marketing cannot hold this and should not try. Every retailer that has attempted a marketing-side product database has ended up maintaining two item masters that disagree, and the disagreement surfaces as an ad for something the store stopped carrying in March.
Marketing owns what is said
The language layer over the catalog — per-SKU descriptions, per-store page copy, brand voice, and the gate that decides what publishes. Channel structure. The measurement layer, including the store-level join.
This is the half that does not exist in the PIM. A PIM holds structured attributes and enforces none of the sentences built from them, which is why per-SKU quality degrades quietly as the catalog grows.
Neither owns alone: the join
The map from SKU to store to catchment. Which items are sellable, where, this week — and therefore what may be advertised into which geography.
This is the single most expensive unowned artifact in multi-location retail. Merchandising has the inventory and no view of demand; marketing has the spend and no view of on-hand. Nobody is accountable for the join, so spend keeps landing on items the nearest store cannot fulfil, and the failure reads as a channel problem rather than an ownership one.
Four moves that stop spend landing on what you cannot sell
Every one of these is an ownership change wearing operational clothes. None of them requires a new vendor, and none of them works as a report that somebody is supposed to read.
Make on-hand a gating input, not a report
Inventory should decide what is eligible to run before a campaign is built, not explain the result afterwards. A weekly export that marketing reads is not a gate. The eligibility rule has to sit in the system that produces the campaign, or it will be honoured whenever there is time and skipped whenever there is not.
Set the threshold per item, not per company
A single company-wide floor — suppress below N units — either wastes long-tail depth or advertises a display model into a sell-out. Thresholds belong to velocity and lead time. A slow item with six weeks of cover is safer to promote than a fast item with eleven units on hand.
Give every store a reason to trust the number
Store teams stop reporting inventory accuracy the moment a correction changes nothing they can see. Show the store what its on-hand accuracy did to its own traffic and its own BOPIS completion, at its own address. Accuracy is a marketing input, so marketing has to close that loop rather than assume operations will.
Treat returns and cancellations as content signal
A return reason is a sentence about the product description that a customer wrote for you. Returns data almost never reaches the people who own the copy, so the description that caused the return stays live and keeps causing them. Route the reason codes to the language layer on a fixed cadence and the worst offenders surface without anyone auditing a catalog.
The store-tier protocol
Not every store deserves the same spend, and everyone knows it. The reason it goes unsaid is that the conversation is a personnel problem before it is a budget one, which is exactly why a published protocol resolves it faster than a good argument.
1. Tier stores on what they can convert, not on what they sell
Revenue rank tells you where the volume already is. Tiering for marketing needs headroom: catchment size against current share, fulfilment reliability, and staffing stability. A high-revenue store with no headroom is a poor place to add spend.
2. Write the tier into the systems, not the deck
The page generator, the local campaign structure and the budget allocation all read the same tier assignment. A tier that lives only in a slide is a tier that the next quarterly plan quietly ignores.
3. Publish the criteria before the first store is moved
A store that learns its tier from a budget cut experiences a judgement about its team. The same decision, taken against criteria published a quarter earlier, reads as arithmetic. The criteria matter less than their being visible first.
4. Re-tier on a fixed calendar, and only then
Continuous re-tiering invites lobbying from every district manager who had a strong month. A fixed cadence — twice a year is usually enough — keeps the map stable and the conversation about the criteria rather than about the exception.
The store conversation
Five things to say, in this order. The order matters more than the wording.
- 1“Here is what your store gets, and here is what stays yours to run.” Store leadership hears centralisation as a loss of judgement. Name the boundary before naming the benefit.
- 2“Here is your number today.” Their traffic, their BOPIS completion, their local pack position — at their address, against their own baseline. A chain average persuades nobody who runs one building.
- 3“Here is what we need from you, and how often.” Usually inventory accuracy and hours accuracy. State the real cadence. Understating the ask is the fastest way to lose store cooperation in month two.
- 4“Here is what happens if nothing changes.” Not a threat. The counterfactual is that spend keeps arriving for items the store cannot fulfil, and the store keeps absorbing the customer disappointment.
- 5“Here is how the tier moves.” Name the measure that earns more investment, so the model reads as a ladder rather than a verdict.
The boundary comes first because a store team that expects to lose judgement stops listening before the benefit arrives. Most versions of this conversation that fail, fail by opening with the benefit.
What this pairs with
This is the governance half. The build half — which agent goes first, what it depends on, and a 90-day sequence with an acceptance test on every phase — is the deployment blueprint. Together they are the whole method.
If you would rather have someone run it alongside your team, the assessment maps your current position against this model.
Ready to talk now? Request the 30-minute consultation. Scope and pricing are settled on the call.