Can a hundred pharmacies manage stock the same way?

One policy, a hundred decisions

Picture a chain of a hundred pharmacies where the order is decided by people on the spot. One owner, one brand, one strategy. But every day the stock decision is taken by a hundred different people.

Each has their own experience, their own attitude to risk, their own sense of how much is enough. One is afraid of an empty shelf. Another is afraid of surplus. A third looks only at last week’s sales. A fourth trusts the supplier’s representative. A fifth simply does not want to change how they have always worked.

The company has one policy. It has a hundred systems wearing the same logo.

Which is why the real purpose of automating the order is not speed. It is this: a good daily decision should stop depending on who happens to be working in that pharmacy today.

Three pharmacies, three different businesses

Without a single system, different stock models appear inside one chain very quickly. In one pharmacy high availability has been bought with surplus. In another the stock looks light, but sales and customers are being lost. In a third there is a lot of money in goods and still not enough of the right range.

Occasionally you find a pharmacy where availability is high and turnover is good. But if that result rests on one strong individual, the company does not yet own a system. The person leaves, and the result can leave with them.

A chain cannot build a competitive advantage on the hope that a rare all-rounder will happen to be standing behind every counter. Scale exists when the result can be repeated in every location regardless of the individual.

Why the owner’s instruction does not change the system

The owner sees too much money in stock and says: reduce it. One pharmacy cuts too hard and runs into stockouts. Another removes a few lines. A third changes almost nothing. Within a month or two most drift back to their old behaviour.

Then the opposite problem arrives: availability must go up. The simplest local answer is to order more. Availability rises — and so do days of stock.

In both cases the direction from the top may have been right. What was missing was the mechanism that turns direction into the same daily decision everywhere.

Without that mechanism, a management policy is only a request.

Could you just hire the perfect pharmacist for every location?

To manage ordering well by hand, a person has to understand the customer, analyse demand and stock, see the system, allow for slow movers and supply constraints, and hold the discipline every single day.

If such a specialist really exists, is manual recalculation of every order the best use of their attention — or should it go where a machine cannot take a standard decision?

If the result depends on a rare person, the company does not own a management system. It owns a fortunate exception.

Five levels, and one question that separates them

1. Manual ordering. The person decides what and how much to order.
2. The system recommends. It calculates; the person overrides freely.
3. Automatic order with manual changes. The system creates the order; it matters that you can see who changed it, why, and with what result.
4. Automatic order with exception management. Standard decisions stay with the system; people intervene on a genuine exception.
5. Centralised network policy. The company manages the rules, the rights, the parameters and the recurring causes.

Between level two and level four sits one question: who takes the decision by default? If an employee can overwrite the calculated order without giving a reason, you have not automated ordering. You have added a screen with a suggestion on it next to the manual process.

Automation amplifies whatever logic you put into it

You can hand the order to a machine and simply automate the habitual way of calculating it. That is worth being explicit about, because the two common architectures behave very differently.

Starting from the forecast

The most widespread model begins by estimating future sales, then converts that estimate into required stock and an order. The forecast can be improved endlessly — seasonality, trends, promotions, day of week. The better the forecast, the better the decision inside that model.

But the principle does not change: first predict demand, then turn the prediction into stock. Two consequences follow. A sharp acceleration or slowdown reaches the calculation only after new data has shifted the estimate of the future. And the higher the service level required, the more the system has to insure itself against forecast error.

If the whole market uses the same paradigm, a better forecast improves your result — but it does not necessarily open a gap between you and the best of your competitors.

Starting from the state of the stock

The forecast approach asks: how much do we expect to sell? Managing by stock state asks a different question: is the current stock adequate relative to where the system wants it to be? The object being managed is the state of the stock, and the target becomes the system’s response to that state.

The target here is not a static norm and not simply “average sales × days”. The system moves it, by rules, when stock keeps landing too low or too high.

This does not rule out forecasting as a tool for budgeting, purchase planning or promotions. For daily replenishment, it is a different control system. In practice the two can be combined — a forecast system also looks at stock state, and state-based control can use an estimate of future demand. The difference is which signal is in charge: expected demand, or the deviation of the actual state from the required one.

Competitive advantage is sometimes created not by a more accurate forecast, but by a different architecture of feedback.

This is the principle Horizon is built on. In specific projects, availability has been lifted above 99% — in some cases to roughly 99.5% — with days of stock under constant control. Those are results of particular implementations, not a universal guarantee.

350 pharmacies, and twenty that said the system was wrong

In a chain of about 350 pharmacies, orders to the warehouse had been generated fully automatically for around nine months, with almost no human intervention in the operational decision.

Then about twenty pharmacies began to argue actively that the automatic order was managing them badly. They went to the owner and produced individual situations as proof.

We compared the data. Availability in those pharmacies was very high — the same as in the rest of the chain. Days of stock were not:

The problem was not that customer service is impossible under automation. Those locations wanted to hold noticeably more stock than their actual demand justified. For the owner, the argument changed shape. Instead of “who is right, the system or the pharmacy?” the question became: if the service is comparable, why does one group need roughly 40% more days of stock?

Opinions can be debated. Availability and days of stock can be measured.

Why automatic ordering is sometimes felt as a threat

Manual ordering is not only work. It is also the right to decide which goods enter the pharmacy and in what quantity. Automation changes that right.

In some companies local decisions are influenced by supplier representatives, promotional programmes, long-standing relationships, local arrangements or personal incentives. But even with no financial motive at all, a person may defend a familiar autonomy and dislike a system that makes their decisions visible.

So resistance does not prove the algorithm is bad. Sometimes it shows that someone is losing the ability to shape stock at their own discretion. The way out is to convert opinions into a measurable question: what happened to availability and to days of stock?

Automation should remove repeated decisions, not management thinking

There is a dangerous illusion — that if the machine places the order, nobody needs to think any more. Mature automation does the opposite. It removes thousands of repetitive decisions so that people can work on the things that need judgement:

• why this exception arose
• why one supplier systematically creates a problem
• why one pharmacy behaves differently from the rest
• why the system keeps needing the same manual correction
• whether to change a parameter, a process, or the policy itself

The routine order is the system’s job. The recurring cause of exceptions is management’s job. The pharmacist keeps the work where a human genuinely matters: the patient, spotting a real exception, explaining a cause, reporting a local event.

Questions for the owner

• Who takes the decision by default — the system or the person?
• How often do staff change the calculated order, and is the reason recorded?
• What happens to availability and days of stock after manual changes?
• Which reasons for intervening keep repeating?
• Are we fixing a single order, or the process that keeps producing the same exception?

Related: How to automate manual ordering · Horizon vs the alternatives · What high availability costs

The thinking in this article draws on The Evolution of the Pharmaceutical Market, a book by our founder Serzas Gutsaga.