Two suppliers sell you the same item. One quotes two rupees less, and he has had the business for three years.

Ask yourself when you last checked whether he is still the cheaper one. Not whether his rate is still lower — whether he is still cheaper, which is a different question and a harder one.

Because somewhere in those three years his delivery went from four days to six, and then to nine in the weeks you needed him most. The four per cent he used to give became three, in a conversation nobody wrote down. Two consignments went back on quality and one debit note is still sitting unadjusted. None of that is on the invoice. The two rupees is.

Nobody is grading your suppliers, and they know it

This is worth saying plainly, because it is not an accusation. It is a mechanism.

Economists have a decent model of why long business relationships hold together without anyone going to court. Baker, Gibbons and Murphy set it out formally in 2002: a relationship stays honest as long as the value of all the future dealings is worth more to both sides than what either could grab by defecting once. Robert Macchiavello's 2022 survey of the same question in developing economies adds the part that matters here — where formal contract enforcement is slow and expensive, these informal relationships do more of the work, not less. That is not an Indian failing. It is why a market where nobody sues anybody functions at all.

But it has a failure mode, and it is quiet.

If a supplier believes nothing is being measured, the cost of a small defection falls to almost nothing. Not a betrayal — a slip. Two days late on a delivery nobody was timing. A rate that went up in March and nobody mentioned it. A short consignment of eleven pieces against an order of a hundred.

Each one is too small to be worth a phone call. That is precisely why there are so many of them, and why three years later you are buying from somebody quite different from the person you thought you had a relationship with.

Four things drift, and only one of them has a number

Look at what actually goes wrong with a supplier:

• The rate creeps, or the discount quietly narrows.
• The delivery takes longer, and — worse — becomes less predictable.
• The quality drifts, and rejections go up.
• The settlement slows: returns not taken back, debit notes not adjusted.

Every one of those costs you money. But only the first one is written down anywhere, because the rate is on the purchase entry and the other three are in somebody's memory.

So the rate is the only thing that ever gets argued about. Your supplier knows exactly which number you are watching, and he is careful with that one.

The arithmetic nobody does: what "late" actually costs

Here is the part that turns a feeling into a figure.

The stock you are forced to keep as cover is not a matter of taste. There is a standard formula for it, in every operations textbook — Chopra and Meindl's is the usual one — and it says the safety stock you need depends on two things about your supplier: how long he takes on average, and how much that varies.

The second term is the interesting one. In the formula, lead-time variability is multiplied by your average demand squared. That has a consequence worth stating in plain words:

An unreliable supplier on your fastest-moving line costs you far more than an unreliable supplier on a slow one — not a bit more, disproportionately more.

Which is the opposite of how most buying decisions get made. The fast line is where you negotiate hardest on rate, and it is exactly where unreliability is most expensive.

One honest qualification, because the research is more careful than the sales pitch. Chopra, Reinhardt and Dada showed in 2004 that the standard formula can over-encourage managers to chase variability specifically, and that cutting a supplier's average lead time is often worth more than cutting its variability. Both matter. Neither appears in his price.

And this does work in practice, at scale. When Walmart forced its suppliers' delivery performance up, it reported that it could hold the same in-stock levels with less inventory. The supplier got more reliable; the buyer's money came out of the godown.

What the large buyers do, and the part that is portable

Big retailers measure this formally, under the name on-time in full — did the whole order arrive, complete, on the day it was promised.

Walmart's programme is the best-documented example, and the numbers are instructive. From August 2017 suppliers had to deliver the full order on the must-arrive date 75 per cent of the time, or pay a penalty of three per cent. The bar went to 85 per cent, then to 87 per cent in March 2019, and by September 2020 to 98 per cent.

But the detail worth your attention is not the threshold. It is what the suppliers were actually scoring when the measurement started: some of the largest were reported as low as 10 per cent, and as late as March 2019 the food and consumables category was running at around 40 per cent against a 97.5 per cent requirement.

Read that again. These are enormous, professional companies, and they were nowhere near — not because they were incompetent, but because nobody had ever measured them before. Your supplier is in exactly that position today.

None of this is new, either. Gordon Dickson catalogued twenty-three supplier-evaluation criteria in 1966. You do not need twenty-three. You need four.

The four numbers worth keeping

1. On time, in full. Of the last twenty purchase orders to this supplier, how many arrived complete, by the date promised? One percentage, per supplier.
2. What you actually paid, by item and by date — not what his rate card says, and not what you remember agreeing.
3. What went back. Rejections and returns as a share of what he sent.
4. Debit notes raised against debit notes settled, and how long the gap was.

Notice that none of these require new work. Every one falls out of paperwork you are already producing — the purchase order, the goods receipt, the debit note. The only difference is whether they are recorded in a way you can compare, or written on a pad and then remembered selectively.

That is the entire gap between a shop that knows its suppliers and one that has opinions about them.

Rank them. Do not grade them.

There is a useful experimental finding here that most people get backwards.

Shokoohyar and Katok tested two ways of scoring suppliers in 2022. One gave each supplier a fixed target to hit. The other ranked suppliers against each other. Suppliers put in more effort under both — but noticeably more under ranking. Competition did work that a target could not.

So the conversation to have is not "you need to be at ninety-five per cent." It is:

"On delivery, you are third out of four."

That is a different sentence and it produces a different response. It also happens to be the easier one to defend, because you are not asserting a standard — you are reporting a fact about people he considers his equals.

The honest limit: a score needs somewhere else to go

Measurement only has teeth if the supplier believes you could actually leave.

A study published this year tracked 78 suppliers across 42 months at a large manufacturer and found what you would hope: suppliers who improved were less likely to be dropped, and suppliers whose competitors were doing well were more likely to be. Relative performance genuinely drives decisions.

But the same study found the link weakens sharply when a component is scarce or specialised — because a buyer will not drop a supplier he cannot replace, and both sides know it.

Which is the real argument for keeping a second source, and the economics here is unusually clear. Roman Inderst showed in 2008 that committing to a single supplier is only rational for a buyer who controls a large fraction of that supplier's market. You do not. Almost nobody reading this does. Burke and colleagues reached a compatible conclusion from the operations side a year earlier.

A second supplier is not disloyalty. It is the thing that makes your first supplier's score mean something.

What this is actually worth — the honest number

Most articles would stop at "you will save money". Here is the real evidence, including the part that cuts against the pitch.

Matthew Grennan and Ashley Swanson studied what happened when hospitals joined a database that let them see what other hospitals were paying for identical products, and published it in the Journal of Political Economy in 2020. The result: savings of 3.3 to 3.9 per cent on the big-ticket items and 1.6 per cent on commodity items — among buyers who had been paying high prices.

Now the part that matters more. Across all buyers, the average effect was about three dollars a unit and was not statistically significant. Every bit of the gain sat with the people who had been overpaying and did not know it.

That is the honest promise of this whole exercise, and it is worth more than an exaggeration:

If you have been buying well, measuring will mostly tell you so. If you have been buying by feel for six years, this is where the money is. And you cannot tell which of those you are until you look.

It is also worth knowing how wide the spread on an identical product can be. In an earlier study of one heavily-used medical device, Grennan found prices across hospitals varying by around thirteen per cent around the mean — for a physically identical item — and moving from the lower quarter of that range to the upper one was worth roughly three hundred thousand dollars a year to an average-sized hospital. What separated the buyers was not only their size. It was how well they negotiated.

You cannot negotiate well from memory. You can negotiate well from a list of every rate you have paid for that item for two years.

One more reason, and this one has a date on it

Returns and debit notes used to be a commercial nuisance. Since last year they are also a compliance dependency, and the change is recent enough that many traders have not caught up.

Three things worth knowing:

When you return goods or reject a consignment, it is the supplier who issues the credit note, under section 34 of the CGST Act — not you. Your debit note is your record; his credit note is what moves the tax.

There is a deadline, and it is now the same as the one for claiming input credit. A credit note has to be declared by the 30th of November following the end of that financial year, or the annual return, whichever comes first. That changed from September with effect from October 2022. Input tax credit under section 16(4) runs to the same date. One date to remember instead of two.

And since 1 October 2025, his credit note only reduces his liability if you have reversed the matching input tax credit. That came in through Notification 16/2025–Central Tax. It means a return that your books and his books disagree about is now his problem as well as yours — and he will come back to you about it, months later, when neither of you can remember the consignment.

Separately, the 180-day rule is worth knowing precisely, because it is usually described wrongly. If you have not paid a supplier within 180 days of his invoice, you pay back or reverse the input credit, with interest. But since October 2022 that reversal is proportionate to the amount unpaid, not all-or-nothing, and you get the credit back when you pay him. Part-paying an invoice does not cost you the whole credit.

"We will sort out the debit notes later" now has a date attached to it.

How Stock2Track does this part

Little of this is exotic. It is bookkeeping you are largely doing already — the difference is whether it is kept in a form you can compare.

• Every rate you have paid, by item, by supplier, by date. So when a supplier's rate moves you see it in the week it moves, and when you sit down to negotiate you have the history in front of you rather than in your head.
• The same item across suppliers, side by side. Which is the comparison that makes a rate meaningful at all.
• Purchase order against goods receipt. What was ordered, what arrived, when, and what was short. On-time-in-full is not a feature you switch on; it is what falls out of recording those two things against each other.
• Debit notes raised, and whether they have been adjusted against payment, with ageing — so the ones going stale are visible while something can still be done about them.

What the software contributes is not intelligence. It is memory that does not flatter anybody.

The short version

• Suppliers drift. It is not dishonesty, it is what happens when a relationship is not being measured — and the economics of that is well understood.
• Four things drift; only the rate has a number attached, so only the rate gets argued about.
• An unreliable supplier costs you real money in the stock you must carry to cover him, and disproportionately more on your fast-moving lines.
• Keep four numbers per supplier: on-time-in-full, rate history, rejections, debit notes settled. All four come out of paperwork you already produce.
• Rank suppliers against each other rather than against a target. It demonstrably works better.
• A score only has teeth if you could actually buy elsewhere, which is the real case for a second source.
• Be honest about the payoff: if you have been buying well, this confirms it. If you have been buying by feel, this is where the money has been going. You cannot know which without looking.
• Returns and debit notes now carry deadlines and a dependency on your supplier's filing. That one is not optional.

One last thought, since you are also somebody's supplier. Everything above is being done to you by your own customers, or it is not — and the four numbers your buyers would keep about you are worth knowing before they mention them.

If you want to start, do not start with software. Take your last twenty purchase orders to your largest supplier and work out, by hand, how many arrived complete and on the promised date. It takes an afternoon. Whatever that percentage turns out to be, it will be the most useful thing you learn about your business this month — and it will tell you straight away whether the rest of this is worth automating.

---

Where the figures come from: the safety-stock relationship is standard operations-management material, set out in Chopra and Meindl's Supply Chain Management and analysed in Chopra, Reinhardt and Dada, Decision Sciences, 2004; the Walmart delivery-compliance figures are from contemporaneous trade reporting between 2017 and 2020 and should not be read as current requirements; the supplier-evaluation criteria are from Dickson, Journal of Purchasing, 1966; the scorecard experiment from Shokoohyar and Katok, Journal of the Operational Research Society, 2022; the supplier-termination study from O'Connor, Romero and Asiaei, International Journal of Production Economics, 2026; the sourcing results from Inderst, RAND Journal of Economics, 2008, and Burke, Carrillo and Vakharia, European Journal of Operational Research, 2007; the price-transparency findings from Grennan and Swanson, Journal of Political Economy, 2020, and Grennan, American Economic Review, 2013; and the relational-contract material from Baker, Gibbons and Murphy, Quarterly Journal of Economics, 2002, and Macchiavello, Annual Review of Economics, 2022. The GST points refer to sections 16 and 34 of the CGST Act, rule 37, and Notification 16/2025–Central Tax; the October 2025 change is recent and tax positions move, so treat all of it as orientation and let your own tax adviser tell you what applies to you. We build software, not legal advice.