Open Stock2Track on a Monday morning and look at the order it puts things in.

At the top, in red: sixty-three products have fallen below reorder level. Urgent. Impossible to ignore.

At the bottom, in a calm blue, filed as information: four hundred and thirty-nine items have not sold in ninety days, with the capital they are holding printed beside them.

The second is by far the larger number in rupees, and it sits at the bottom in the quiet colour for a reason that has nothing to do with software design. Running out is an event. Not selling is the absence of an event — and neither a shopkeeper's memory nor a notification list puts an absence first.

This is about why your head gets that backwards, and what to put on the screen so it stops mattering.

Your memory is doing its job. That is the problem.

In 1973 Tversky and Kahneman showed that people judge how often something happens not by counting but by how easily examples come to mind. Usually that works, because common things are easy to recall. Now apply it to a shop.

A customer asks for something, you do not have it, and he is standing in front of you looking irritated. Vivid, recent, unpleasant. It files itself.

A carton at the back that nobody has asked for in three months files nothing. There is no episode to remember, because nothing happened. It is not that you remember stockouts more strongly than dead stock — dead stock leaves no memory to retrieve at all.

So when you sit down to buy, your head hands you a list of things that once ran out and absolutely nothing about things that never moved. Reliably. Every time.

Which is precisely why dead stock is on the screen with a rupee figure attached. Not as a report you have to think to run — as a line that arrives whether you asked for it or not, saying four hundred and thirty-nine items, this much money, ninety days. It is there because it is the one thing your memory will never raise on its own.

The stockouts you remember are the cheap ones

This is the part that should change how you think about buying.

The most thorough study of retail stockouts — fifty-two studies, 661 outlets, seventy-one thousand shoppers — found a worldwide out-of-stock rate of about 8.3 per cent, and recorded what customers actually do when the thing they came for is missing:

What the customer does	Share
Buys it at another shop	32%
Substitutes a different brand	20%
Substitutes the same brand, different size	20%
Delays the purchase	17%
Does not buy at all	11%

Read it twice, because the important thing is not in the numbers.

Forty per cent substitute, right there in your shop. That is the case that creates a conversation at your counter — the memorable one. And it costs you almost nothing, because you still made the sale.

The other sixty per cent walk out, put it off, or quietly give up. No conversation. No record. Usually nobody tells you.

So it is not simply that stockouts are visible and dead stock is not. Within stockouts too, the visible ones are the cheap ones. Your instinct has been trained, diligently and in good faith, on the least expensive part of your losses.

Here is the honest bit: no software sees the customer who walked out. Stock2Track does not either. What it does is stop the stockout happening — a reorder level on every item, checked against live stock daily, and a line at the top of your screen when sixty-three have been crossed. You are not being told about the sale you lost. You are being told before it becomes one.

The same study found that roughly seventy per cent of stockouts start with the shop's own ordering, not with the supplier. This one is yours to fix.

Some of your instincts are excellent. Some were never trained at all.

Most articles on this subject tell you your gut is unreliable. That is lazy, and the research does not support it. Simple rules — ignore most of the information, use one good cue, decide — frequently match or beat elaborate models, and they win under exactly the conditions a small shop has. In one study, managers' rule of thumb for spotting repeat customers (has he bought recently?) got 83 per cent right against a sophisticated model's 75.

But there is a framework for when that stops being true. Robin Hogarth divides the situations we learn from into kind and wicked: a kind one gives feedback that links what happened back to what you did, accurately and often; a wicked one gives feedback that is poor, misleading or missing. You get good at things in kind environments and never at things in wicked ones — and it feels the same from the inside either way.

The split runs straight through one shopkeeper doing one job.

His instinct about the counter is trained in a kind environment. Which customers haggle, whose driver turns up late, what moves in the first week of the month. Fast, direct, unmistakable feedback. That instinct is probably excellent and no software should touch it.

His instinct about dead stock was never trained, because that environment has never once sent him a signal. There was nothing to learn from.

Same man, same shop, same thirty years — and no way to feel the difference. That is the whole brief for what a system should and should not do. Leave the counter alone. Cover the silence.

Two thousand lines is not a judgement. It is arithmetic.

Robyn Dawes put the division of labour better than anyone, in 1979: people are good at working out which things matter, and bad at combining many different and incomparable things into one answer.

That insults neither side. You know which factors matter — season, the wedding rush, which supplier is slow. What nobody can do is hold a shelf life, a lead time, a landed cost, ninety days of movement and the cash tied up for two thousand items at the same time and rank them against each other. That is not a memory task at all. It is arithmetic on a table, and a table does it perfectly the first time.

Which is exactly what the dead-stock line is: not "you have some slow movers" but four hundred and thirty-nine specific items, sorted, with the money each is holding. Click it and you get the list. That is the arithmetic done, and it is the part your head was never going to do.

Two errors are worth naming because they are systematic rather than random, so being careful does not fix them:

You hedge toward the middle. People ordering under uncertainty pull quantities toward average demand and away from the amount that would make the most money — and how the decision is presented changes the size of it. Seeing the numbers helps in a way that trying harder does not.

You are anchored, and experience removes the awareness rather than the bias. Estate agents shown the same property with different asking prices valued it differently. Fifty-six per cent of amateurs admitted the asking price had swayed them; only twenty-four per cent of the professionals did. The experts were anchored and less likely to notice.

Your anchors are the supplier's suggested quantity, the case-pack size, what you ordered last time. Which is why the purchase history in Stock2Track is per item, per supplier, per date. When you are about to agree a rate, the last four rates you paid are on the screen rather than in your memory, where the anchor lives.

Decide once, not two thousand times

Here is the thing this software actually does, and it is worth being precise about, because it is not what people assume.

Stock2Track does not decide anything. We have said that plainly elsewhere on this site and it remains true: software reminds, calculates, compares and records. You decide.

What changes is when you decide.

You sit down once, in a quiet hour, and set the rules:

• A reorder level on each item — what "low" means for this product, given how fast it sells and how long your supplier takes.
• A credit limit on each customer — how much you are willing to be owed by this person.
• A period that counts as dead — ninety days, or whatever suits your trade.
• Who is allowed to give a discount, and how much, before anybody is standing at a counter under pressure.

Then Stock2Track applies those rules to every item, every customer and every day, and puts the results on one screen.

You have not handed anything over. You have moved the decision away from the counter, with a supplier waiting and a queue behind him, to a desk, calmly, in advance. It is the same argument we made about discounts at an exhibition: the number should come from a rule set beforehand rather than from whoever happens to be holding the calculator.

Two thousand decisions a week becomes four decisions a quarter, plus a list to look at.

What the screen actually says on a Monday morning

This is worth walking through, because every line on it is a thing that used to depend on somebody remembering.

Sixty-three products below reorder level. The stockout that has not happened yet. Compared against your own rule, not somebody's idea of low.

Twenty-eight customers with overdue payments, with the total attached. Not "some people owe us" — the number, and who. Receivables are the classic thing that drifts, because nobody feels the absence of money that was never in the account.

Four suppliers with payments pending, with the payable total. The other side of the same coin, and the one that quietly damages a relationship you rely on.

Two indents pending fulfilment, and one stock transfer despatched and not accepted at the other end. This is the gap goods disappear into when a business has more than one location. Something left here and has not arrived there, and until somebody notices it is nobody's problem.

Two credit notes not yet refunded. Small money, and exactly the kind that goes stale.

Leads with overdue follow-ups, and leads with no activity in seven days. The sales equivalent of dead stock: nothing is happening, so nothing reminds you.

And four hundred and thirty-nine dead stock items, with the capital they are holding.

Look at that list again and notice what it has in common. Almost every line is an absence — something that did not happen, did not arrive, was not chased, did not sell. These are precisely the things a human being cannot notice, because there is nothing to notice. A screen does not have that problem. Nothing happening is just another query.

There is also an owner's view separate from the working dashboard, which matters more than it sounds: the person who needs the ninety-day view is not the person billing at the counter.

When you should overrule it, and when you are fooling yourself

Any article telling you to follow the system every time is wrong, and there is a good test for the exception.

Paul Meehl called it the broken leg problem. A formula says a man is ninety per cent likely to go to the cinema tonight; then you learn he has just broken his leg. Nobody sensible sticks with the formula. Meehl was also specific about what a real exception looks like: it rests on an objective fact, almost perfectly connected to the outcome, needing no theory to understand.

Real broken legs in a shop: the road to that market is closed for six weeks. Your supplier's factory has shut. A large wedding order lands next Tuesday. The competitor two streets away has closed down.

"I have a feeling this will move" is not a broken leg. It is the same memory that never saw the four hundred and thirty-nine.

There is unusually good evidence about which overrides help. Across more than sixty thousand real forecasts at four companies: large adjustments improved accuracy — when somebody genuinely knew something the system did not, they were right. Small adjustments made things worse. And upward adjustments were much more often wrong than downward ones.

The direction your memory pushes you — order more, because you remember running out — is the direction in which human adjustment is most often wrong.

Override rarely, override big, and be most suspicious of yourself when the override is upward. This is also why Stock2Track hands you a list rather than placing orders on your behalf: the override needs to be available, and it needs to be a decision somebody made.

Where the system is wrong, and what stays yours

The opposite case is real too.

Researchers studied nineteen thousand item-and-store combinations in supermarkets where an automatic system proposed orders and managers routinely overrode it. The managers smoothed their ordering, pulling orders off peak days, and carried about ten per cent more inventory as a result. On the system's terms, they were wrong.

But their smoothing cut the swing in handling workload by more than forty per cent. Where labour costs more than storage, that may have been the better trade all along. The system had no idea who was available to unload on a Tuesday.

The system owns the arithmetic. You own the constraints it cannot see — pack sizes, shelf space, who is working, when cash actually arrives.

Two honest limits

The dead-stock statistics you will see elsewhere are made up. Look for a figure on what share of retail inventory is dead and you will find twenty to thirty per cent quoted everywhere. Every source is a company selling software to fix dead stock, and not one says where the number came from. We are not going to give you one. Your own ninety-day list is the only figure that matters and it takes a click.

But dead stock is worse than idle cash, and that part is measured. When one retailer cut the lines it carried across forty-two categories, sales went up by an average of eleven per cent. The slow movers were not merely sitting there — they were getting in the way of what sells. Overstocking does not only lock up money; it manufactures stockouts.

And a system is not magic. One study of 370,000 records across thirty-seven stores of a fully computerised retailer found sixty-five per cent of them were wrong. Software plus sloppy counting is still sloppy counting. What a system gives you is a record that can be checked; the discipline to check it is still yours.

The short version

• Your memory stores drama, not proportion. Dead stock creates no memory to retrieve, because nothing happens.
• The stockouts you remember are the cheap ones. Around 40% of customers substitute in your shop — visible, and it costs you almost nothing. The 60% who leave make no sound.
• You get good at things where feedback is fast and clear, and never at things where it never arrives. Both feel identical from the inside.
• Your judgement about one line or one customer beats any system. Your judgement about two thousand at once is arithmetic, not judgement.
• Decide once, as a rule — reorder level, credit limit, what counts as dead — and let the screen apply it daily. That moves the decision away from the counter; it does not hand it over.
• Nearly every line on that screen is an absence: stock not reordered, money not collected, goods not received, an item not sold. Absences are what people cannot see and queries find easily.
• Override for an objective fact you can name, not a feeling. Upward overrides are most often wrong — the exact direction your memory pushes.

If you want one thing to do this week, it is not to buy software. Pull the list of everything you have not sold in ninety days and add up what it cost you. Most owners have never seen that number, and whatever it is, it will tell you straight away whether the rest of this is worth your time. If you want help getting that one figure out of whatever you run now, ask — it is a short conversation and we do not mind if the answer sends you away happy.

---

Where the figures come from: the out-of-stock rate, the consumer-response breakdown and the store-level cause share are from Gruen, Corsten and Bharadwaj, Retail Out-of-Stocks: A Worldwide Examination of Extent, Causes and Consumer Responses (2002); kind and wicked learning environments from Hogarth, Educating Intuition (2001); the conditions for trusting intuition from Kahneman and Klein, American Psychologist, 2009; availability from Tversky and Kahneman, Cognitive Psychology, 1973; the limits of human information integration from Dawes, American Psychologist, 1979; simple heuristics from Gigerenzer and Gaissmaier, Annual Review of Psychology, 2011, and the customer-base comparison from Wübben and von Wangenheim, Journal of Marketing, 2008; hedging toward mean demand from Schweitzer and Cachon, Management Science, 2000; anchoring in experts from Northcraft and Neale, Organizational Behavior and Human Decision Processes, 1987; the broken-leg problem from Meehl, Journal of Counseling Psychology, 1957; forecast adjustments from Fildes, Goodwin, Lawrence and Nikolopoulos, International Journal of Forecasting, 2009; supermarket ordering overrides from van Donselaar and colleagues, Management Science, 2010; assortment reduction from Boatwright and Nunes, Journal of Marketing, 2001; phantom products from Ton and Raman, Production and Operations Management, 2010; and inventory record accuracy from DeHoratius and Raman, Management Science, 2008. These are studies of retail and of decision-making in general, not of your shop — they show the shape of the problem, not your numbers.