Winning badly: the price that was paid for somewhere else

Go back to the first number in this course

A negotiation feels finished the moment the price is agreed. This lesson follows one that was won cleanly, and then follows the year that came after it. Nothing improper happened, and no term of the contract was broken. What changed was the part of a supplier's service that was never written down and never paid for. That part is handed out every morning, to whichever customer looks worth the trouble.

Lesson 1 opened with Talpitiya quoting USD 7.05 for autumn 2025. Nobody asked where that number came from. It is the subject of this lesson, and it was decided eighteen months before it was said.

September 2023: the instruction

Larkmead appointed a group sourcing director, Rhodri Penhale, in the summer of 2023. His first act was an instruction to every buying department: 5% out of the supply base for autumn 2024.

It was not an unreasonable instruction. Larkmead's own costs were rising. Its intake margin had slipped for three seasons in a row. And a flat percentage across the whole supply base is the only instrument a group director actually has. He cannot negotiate 340 styles himself, and a target that varies by supplier is a target nobody can audit. Every buying office in the world has received this instruction, and most have issued it.

Thursday 12 October 2023

Talpitiya quoted USD 6.55 for the autumn 2024 LM-6274. Larkmead pushed to USD 6.22, a cut of 5.04%, on the nose.

Talpitiya's cost to make that season was USD 5.72. At USD 6.55 the margin was USD 0.83, which is 12.67% of the FOB. At USD 6.22 it is USD 0.50, or 8.04%.

Nuwan said one sentence in that meeting that Imogen wrote down and did not act on: below about nine per cent of FOB we manage an order rather than make it. Then he accepted, because that autumn Talpitiya had 62 idle line-days in the window, and an order at 8.04% is better than a line running on nothing.

The win: USD 0.33 a piece, USD 15,180. Booked, reported, and correct in every system Larkmead runs.

Why nobody could have been accused of anything

What follows is not retaliation, and it is important to be precise about that. A course that teaches buyers to expect punishment teaches them the wrong thing. Talpitiya did not breach a single term. Autumn 2024 shipped on its contract date. On-time-in-full for the season, meaning the share of orders that arrived on the agreed date with the whole quantity, was 100%. There is no debit note, no claim, no quality event and no complaint on the file.

What changed was discretion: the part of a supplier's service that is not in the contract, has never been priced, and is handed out every morning by people deciding whose problem to solve first.

Four things that happened, all of them provable from Larkmead's own records

The samples slowed down. The usual sample turnaround on the Talpitiya account went from 4 working days to 11. Nothing was late against a contract, because sample turnaround was not in the contract. The consequence was this: three autumn 2025 styles were not ready for Larkmead's own range presentation on 6 March, and were placed instead with Hettipola Apparels at USD 0.44 a piece more on 21,000 pieces, which is USD 9,240.

The proposals stopped. In autumn 2022 and autumn 2023, Talpitiya had brought Larkmead three cost proposals nobody had asked for. A wider knitting machine running the same quality (USD 0.13). A collar construction change (USD 0.07). And a carton size change that took a pallet layer out of the freight (USD 0.04). That is USD 0.24 a piece. In autumn 2024 and spring 2025 it brought none.

This is the quietest loss in the whole account, and the largest one that repeats. Unasked-for improvement is unpaid work. A supplier at 12.67% funds it out of the expectation of a relationship. A supplier at 8.04% funds it out of nothing, so it stops. USD 0.24 across 46,000 pieces is USD 11,040 a year of cost reduction that simply stopped arriving. And it stopped without a meeting, an email or a decision anybody could point to.

The repeat was declined. In week 6 of the autumn 2024 season, LM-6274 was selling 34% ahead of plan. Duncan asked for 12,000 pieces in 42 days. Talpitiya declined: no capacity in that window.

Larkmead's plan values a full-price piece of LM-6274 at a contribution of USD 6.90. Twelve thousand pieces of demand that could not be met is USD 82,800.

The next quote opened higher. On 10 April 2025 Talpitiya opened autumn 2025 at USD 7.05, which is 13.34% above the autumn 2024 settled price. Break it down against Talpitiya's own bulletin, which Imogen asked for and received:

USD 0.48 a piece is USD 22,080, and it is not a negotiating position Imogen can talk down, because it is not padding. It is a supplier refusing to run an account at 8.04% for a second year and pricing itself back to where it can. And it arrives in the one place a buying system cannot inspect: the opening number.

The account, totalled honestly

The provable items alone are 2.79 times the win.

Now the honest part, and it belongs in the lesson rather than in a footnote. You cannot prove Talpitiya would have taken the repeat at a different price. They said they had no capacity, and factories genuinely do run out of capacity in November. What Larkmead can show is that Talpitiya shipped a 14,000-piece repeat for another customer out of the same weeks, visible on the shipping schedule during Larkmead's own December quality visit. That is evidence of a choice, not proof of a motive. Treat the USD 82,800 as the shape of an exposure rather than as a bill. The other three lines need no such caution. They are Larkmead's own invoices, Larkmead's own sample log, and Talpitiya's own written bulletin.

Distributive and integrative, priced against each other

Two negotiations on the same account, eighteen months apart.

The autumn 2024 negotiation was distributive, which means a fixed amount of money moved from one side of the table to the other. It moved USD 15,180, and then USD 42,360 moved back over the following year.

The autumn 2026 negotiation in lesson 2 was mostly integrative, which means it created value that did not exist before the meeting. A mill's contract band was worth USD 0.18 a piece to nobody until 114,000 pieces were committed against it. Forty-seven idle line-days were worth USD 40,686.40 to nobody until a buyer moved a date. And then, having created it, the two sides argued about how to split it. That is the distributive half, and it is much easier to win when the thing being split did not exist an hour earlier.

A buyer who only knows the distributive game is not playing badly. They are playing a smaller game, and the shrinkage is invisible, because it arrives as an opening quote rather than as a loss.

When to stop: three tests, none of them moral

"Do not squeeze suppliers" is advice nobody can act on. Here are three questions that have answers.

1. What is their floor, and have you asked for it? Nuwan gave his in a sentence: about nine per cent of FOB. Ask every strategic supplier, once, at a review rather than in a negotiation, and write the answer in the account file with the date. They will tell you, because it is in their interest that you know. A floor you have been given and crossed anyway is a decision. A floor you never asked for is an accident.
2. What share of their output are you? Larkmead is 11% of Talpitiya's annual production. That is enough that they will take your order, and not enough that they will restructure around you. Below roughly a tenth you are a customer among many, and your discretionary service is a place in a queue. How that concentration is mapped, in both directions, is course 27.1.
3. What else can they do with the window? This is the test that actually predicts the outcome. Talpitiya said yes in October 2023 because it had 62 idle line-days and no alternative. It said no in November 2024 because it had a full book and a choice. A price won against idle capacity is a price won once. It holds only for as long as the capacity stays idle, and the day it does not, the price returns, with the discretionary service leaving alongside it.

There is a fourth question. It is not a test, but it is worth asking out loud in the room: what will this price do to what you send me next year? Most suppliers will answer it honestly, and the answer is usually more useful than the last cent.

Act as a retail sourcing analyst with no stake in whether last year's negotiation looks good. I want one price win followed all the way through the year that came after it. The negotiation: supplier [SUPPLIER], style or programme [CODE], quantity [PIECES], their quote [PRICE], settled [PRICE], date [DATE], and what I know or can guess about their cost to make [AMOUNT]. Their stated margin floor, if they ever gave me one: [PERCENT]. My share of their annual output: [PERCENT]. Whether they had alternative work for that window: [YES OR NO AND WHAT]. What happened afterwards, as far as I can document it: usual sample turnaround before and after [DAYS AND DAYS], any styles moved to another supplier and at what price difference [DETAIL], any repeat or in-season top-up requested and the answer [DETAIL], the number of cost proposals they brought without being asked in the two seasons before and the two after [NUMBERS AND VALUES], and their opening quote the following season with any breakdown they gave me [DETAIL]. Do the following. First, state the win in dollars and as a share of their margin before and after. Second, build a table of what the year cost, separating what I can PROVE from my own records from what I am guessing, and refuse to present a guess as a fact. Third, break the following season's opening quote into input movement, declared cost inflation and margin rebuild, and tell me which part is negotiable and which is not. Fourth, apply three stopping tests to the original decision: their stated floor, my share of their output, and whether they had an alternative use for that window. Say which test would have predicted the outcome. Fifth, tell me which parts of that year's cost were distributive, meaning money moved between us, and which were value destroyed that neither side got. Sixth, give me the four things to measure each season that would have shown this a year earlier, and say where each one lives in a normal buying office's systems, or that it lives nowhere.

What to take away

A negotiation does not end when the price is agreed. It ends about fourteen months later, when the next quote arrives. By then it has been settled in four currencies: the price, the service you no longer get, the ideas that stop being offered, and the number the supplier opens with.

Larkmead won USD 15,180 in October 2023 and could show it in a report the same afternoon. The USD 42,360 it paid back was spread across a sample log, three purchase orders with a different supplier, a stack of proposals that were never written, and one line of a bulletin. No system joined them up, which is why the same instruction will be issued again. Unless somebody in the buying office does the joining by hand, once a season, in about an hour.