The reference that wins the account

Six minutes on the telephone

On 21 April 2026, Ruth Ellender spent six minutes on the telephone with Kajsa Wennerberg, a merchandiser at Ostlund, the Nordic casualwear brand that is Serapion's largest customer. She asked three questions. She then wrote in her assessment file that Serapion was recommended for a trial order.

Fourteen pages of capability pack, a factory visit, an audit, seventeen samples and eleven months of work. And the account turned on six minutes with somebody who works for neither company.

This is not a sentimental point. It is a structural one, and it has a consequence you can test: the thing that wins your next account is how you behaved in your worst week at your current one.

The three questions, and what each one tests

Notice what is missing. Nobody asks about machines, capacity, certificates or price. All of those were filtered out earlier and mechanically, in the stages of lesson 1. By the time a buyer picks up the telephone, your capability is assumed. The only open question is behaviour under stress. That is the one thing a factory cannot demonstrate on demand, and cannot fake in an interview.

That is also why you cannot prepare for a reference call. You can warn your referee, and you should. The next section is about what happens when you do not. But you cannot change the answer. The answer was written months ago, by something you did.

The answer that was bought fourteen months earlier

September 2025. Ostlund order OS-2288 was running late. The cause was a lab dip (a small trial dyeing sent for colour approval) that sat on Ostlund's own desk for nine days. Serapion had the contractual right to take those nine days back on the delivery date. It had the emails. It would have won the argument.

Instead Nashwa Hilmy air-freighted 4,200 pieces at Serapion's own cost. Air freight cost USD 1.47 a piece more than sea, so the bill was USD 6,174.00. It was booked as a loss against the margin on OS-2288, where it made that order look worse than it was, and nobody looked at it again.

That is the answer Wennerberg gave to the third question, fourteen months later. It is the reason Brindlemoor's file says recommended.

So where does the USD 6,174.00 really belong? Not in the margin on OS-2288. It is a cost of winning an account that did not exist yet. By the second season, Brindlemoor is worth USD 49,020 of contribution. That is a return of 7.9 times, on a decision no accounting system will ever call an investment.

This is the honest version of a cliché. "Look after your customers" is not advice. It is a slogan. What produces the sentence one buyer repeats to another is a specific, dated, priced act: absorbing a cost that was not yours, at a moment when you could have argued and won. Factories under-invest in this for one reason. It is booked as a loss in the month it happens, and it pays out in a different account two years later.

The reference that was neutral

Brindlemoor called three references. The second was Steinacher, a German workwear brand Serapion had shipped to for two seasons.

Nobody at Serapion warned them. The merchandiser who knew Serapion had left eight months earlier. So the call reached her replacement, who opened the vendor file and said, honestly, that there were no issues recorded.

A neutral reference reads as a negative. Ellender's note said the second reference was uninformative, and the consequence was written into the trial: a second pre-production visit, and 100% inspection on the first shipment, at Serapion's cost. That inspection cost USD 1,860. On 14,600 pieces that is USD 0.13 a piece. It is the difference between a contribution of USD 1.25 and the USD 1.12 the order actually earned.

An email nobody sent cost thirteen cents a piece.

A reference is a person, and people leave

Serapion has six accounts it would happily use as a reference. But it does not have six referees. It has six seats, and the person in each one changes.

Half the reference list is people who have never seen Serapion do anything difficult. That is not a relationship problem. It is arithmetic. At an average tenure under three years, a reference list decays completely inside a decade, and half of it turns over every two years.

That makes the maintenance cheap and obvious. Almost nobody does it:

1. Keep the list as names with dates, not companies.
2. Telephone each name once a quarter, about nothing in particular.
3. When a name changes, the clock starts again. Treat that account as unusable for references until the new person has seen you handle something.
4. Ask before you give. A referee who is expecting the call gives an answer. One who is not reads out a file.

What to take away

1. By the time a buyer calls a reference, capability is assumed. The call is about behaviour under stress, and it is the last gate.
2. The three questions are about disclosure, ownership and generosity. None of them can be answered by anything you write about yourself.
3. The answer was written months earlier. Absorbing a cost that was not yours, at a moment when you could have argued and won, is what produces it.
4. Book that decision as what it really is: a cost of winning accounts you have not met yet. Otherwise you will keep treating your best sales spending as a loss.
5. A neutral reference reads as a negative and is paid for in extra inspection and extra visits. Warn every referee, every time.
6. A reference list is people, not companies, and it decays. Keep names and dates, call them quarterly, and start the clock again when the seat changes.