Centralised vs. fragmented stock management in retail

One store sells out of an item while the shop on the next street — from the same chain — has three boxes piling up in the stockroom. The problem is not demand. It is where the stock lives and, more importantly, who controls it. At the end of this guide you will find a decision matrix and a 12-point checklist to audit your stock management model in under two hours.

The debate has been framed wrongly for years

Everyone debates "centralised vs. fragmented" as if it were a system architecture choice. It is not. The real operational question is: who has authority over the stock at any given moment?

A chain can have a central warehouse and stock that is completely fragmented in practice — when each store acts as if its inventory were sovereign and a transfer between points of sale takes three days with manual approval. We have seen this repeatedly in INFOS projects: the system says "centralised", the operation says otherwise. The operations director shows the pretty diagram. The store manager shows the WhatsApp where they have been requesting a top-up for two days with no reply.

Retail in Portugal grew by 4.7% in 2024 (INE). That growth does not forgive inventory inefficiency — it increases pressure on SKU turnover and exposes chains that still manage stock store by store as if each point of sale were an independent company. Growing 4.7% with the same stock model that worked with half the number of stores is the recipe for chronic stockouts and tied-up capital at the same time.

What you need before you start

Before touching any configuration, gather seven pieces of information. Without them, any diagnosis is speculation:

• Up-to-date map of all stock points: stores, central warehouse, transit warehouse, consignment.
• Sales history by store, minimum 12 months, by SKU.
• The real lead time of the main supplier — not the contractual one, but what actually happens in the third week of October.
• Definition of who approves transfers between stores: store manager, regional director or automatic system.
• Confirmation that each store's POS reports stock to the back office in real time — or the exact lag with which it does so.
• List of the references with the highest stockout rate over the last 6 months, by store.
• Clarity on the current replenishment model: push (central decides), pull (store requests) or hybrid.

Step 1 — Diagnose where the stock really lives

Open yesterday's inventory report. Count how many stores have negative or zero stock on a reference that has sold in the last two weeks. Then count how many stores have more than 60 days of cover on the same reference. If the difference between the minimum and maximum cover across stores is greater than 4x, your stock is fragmented in practice — regardless of what the system architecture says.

There is a diagnostic error that comes up regularly: teams look at average stock and conclude everything is fine. A store with zero units and another with 40 give an average of 20 — which resolves neither situation. The average hides exactly the problem worth finding. Always report the minimum, maximum and standard deviation of cover by reference, never just the average.

• Identify the 10 references with the greatest dispersion of cover across stores.
• Calculate the opportunity cost of stockouts: missing units × average margin.
• Record how many transfers between stores were made in the last month and the average execution time.
• Check whether safety stock is defined per store or globally — and whether it is defined at all.

Step 2 — Choose the right model for your operation

There is no universally superior model. There is the model appropriate to your demand profile, your logistics network and the maturity of your retail management software. Use the table below to position your chain.

Step 3 — Define the replenishment rules before touching the software

This is the step most projects skip. The team configures the system before having the rules. Then the system does exactly what it was configured to do — and it is wrong. Correcting the configuration after the system is in production costs three times more than defining the rules beforehand.

The replenishment rule that is not documented does not exist — what exists is the judgement of the store manager who happens to be on shift that day.

Define the reorder point by product family, not by individual store — stores in the same chain rarely have demand profiles distinct enough to justify completely independent parameters. Set the safety stock based on the standard deviation of weekly demand, not on an arbitrary percentage chosen in a meeting. Decide which references are managed on push and which on pull. Fix a maximum cover limit per store — above that limit, the system proposes a transfer to another store or a return to the central warehouse. And document who can override an automatic system suggestion and with what justification: without this control, the replenishment system becomes optional in practice.

Step 4 — Audit real-time visibility

During an implementation project with a regional food retail chain with 18 stores, it was discovered that the central system received stock updates with a 24-hour lag. Replenishment decisions were being made on yesterday's data. The central warehouse was sending goods to stores that had already sold off the excess — and refusing shipments to stores that had run out in the meantime. No one had noticed because the daily report arrived in the morning and looked up to date.

Check your lag now in three steps: open the stock of a reference in the central back office; call the store and ask for an immediate physical count; if the difference is greater than 5% in value, you have a synchronisation problem, not a stock model problem. The right model with the wrong data produces wrong decisions with more confidence — the worst possible scenario.

For chains with an active online channel, this lag is fatal: the customer buys online, the store has no stock, the promise fails. MAXIRETAIL operates with real shared stock — not just synchronised with a delay — which makes ship-from-store and BOPIS operationally viable without oversell.

Step 5 — Build the internal decision case

Take this matrix to the meeting. Fill in the "current situation" column honestly — it is the only exercise that matters before any investment decision.

The five mistakes that destroy the model after implementation

Centralising the stock without centralising the decision. The warehouse becomes central but each store still requests what it wants, when it wants. Fix this with a fixed replenishment calendar and remove the ad hoc request initiative from the stores for automatically managed references.

Configuring the safety stock once and never reviewing it. Seasonality changes, suppliers run late, demand evolves. A and B references need reviewing at the start of each season — at least twice a year. A safety stock calculated in January for a summer reference is a science-fiction number in July.

Treating transfers between stores as an operational exception. In chains with real shared stock, transfer between stores is a normal balancing mechanism, not an emergency. Create a standard process with a 48-hour SLA and run it weekly. When the transfer only happens in a crisis, it happens too late and with a disproportionate logistics cost.

Implementing centralisation without preparing the store team. The store manager who has lost autonomy over their stock will resist the system — not out of ill will, but because they feel they have lost control over their own performance. The answer is not a training session on the new process. It is showing them the dashboards that prove their store has fewer stockouts and better turnover with the new model. The numbers win the autonomy argument.

Confusing integration with synchronisation. Many systems "integrate" POS and back office through files that run in the early hours. That is not real-time stock — it is yesterday's stock with today's appearance. For ship-from-store or BOPIS to work without failing on the promise to the customer, the update has to be transactional, not batch. Ask your software supplier what the exact mechanism is — do not accept "real time" without knowing the interval.

What the numbers reveal after the diagnosis

If the Step 1 diagnosis revealed cover dispersion greater than 4x between stores, the problem is not solved with a meeting. It is solved with real-time stock visibility, automatic replenishment rules and a POS that does not work in island mode. MAXIRETAIL and the KORA Inventory Suite handle shared stock in retail chains with exactly this logic. For the broader context of omnichannel operations in Portugal, the guide for store chains in Portugal develops the framework this article does not cover. And to understand where process automation changes the cost argument, the article on process automation in Portuguese industry gives the operational perspective missing from most retail analyses.

Sources

• INE — Instituto Nacional de Estatística. Turnover Index in Trade, 2024. Available at: www.ine.pt
• INE — Instituto Nacional de Estatística. Trade Statistics 2024. Available at: www.ine.pt

Frequently asked questions

What is the practical difference between centralised and fragmented stock management?

The difference is not in the system, but in the authority. A chain can have a central warehouse and fragmented stock in practice if each store acts autonomously and transfers take days. The reverse also occurs: a "centralised" system that actually operates in a decentralised way. What matters is who controls the stock at each moment and with what speed.

How do I know if my stock is really fragmented?

Open the inventory report and compare the maximum and minimum cover of the same reference across stores. If the difference is greater than 4x, the stock is fragmented in practice. Ignore the average — a store with zero and another with 40 units give an average of 20, which resolves neither situation. Always report the minimum, maximum and standard deviation.

Which stock model is best for a chain with seasonal demand?

The hybrid model is ideal for seasonal or regionalised demand. It allows the central warehouse to manage the base, uniform assortment, while stores keep local stock for seasonal variations or local preferences. Pure centralised stock runs the risk of poor allocation in these scenarios.

How many stores justify moving to centralised management?

With fewer than 5 stores, centralised management introduces unnecessary overhead. From 10 stores with a common assortment onwards, the gains from centralisation become clear and measurable. Between 5 and 10 stores, assess demand uniformity and logistics capacity before deciding.

What is the "reorder point" and how should I define it?

It is the stock level that triggers an automatic replenishment. Define it by product family, not by individual store — stores in the same chain rarely have demand so distinct as to justify completely independent parameters. Base it on the standard deviation of weekly demand, not on arbitrary percentages.

What information do I need before changing the stock management model?

Seven pieces of information are essential: a map of stock points, a 12-month sales history by SKU, the supplier's real lead time, the definition of authority over transfers, confirmation of real-time POS reporting, the references with the highest stockouts, and clarity on the current model (push, pull or hybrid).

What is the cost of inefficient stock management?

Two simultaneous losses: chronic stockouts (lost sales, dissatisfied customers) and tied-up capital (excess stock in other stores). With 4.7% growth in Portuguese retail in 2024, inventory inefficiency quickly exposes chains with obsolete models. Calculate the cost: missing units × average margin.