A bootstrapped services business with a team of PhD model builders. They had taken clients off the large consulting firms, and they had taken them on quality.
They were also discounting. Six-figure contracts where the firms they were beating charged seven figures over five years.
The obvious diagnosis is that they did not know what they were worth. It is the wrong diagnosis, and I want to deal with it first because I nearly made it.
The chief executive knew exactly what the work was worth. The discounting was not ignorance. It was cash.
Bootstrapped company, net ninety day terms, and a discount for payment up front smooths a gap that will otherwise decide whether payroll clears. Anyone who has run a services business without outside money has made that trade, and it is a reasonable answer to the constraint in front of you.
The model she was running said: cash in, against team burn. On that model the discount was defensible and she was right.
The model was complete. It was also missing a term.
The second term is the one nobody had priced. Every engagement assumed the client arrived with data you could use. Fortune 50 companies believe they are data ready. Most of them are not, and you find out after the contract is signed.
When that happened, the team cleaned it. That time came out of the same people who were scheduled on the next engagement, and the one after that.
With six or seven customers in a pipeline and a small team, the bottleneck moved down the line and arrived at a customer who had done nothing wrong.
A single dirty dataset does not delay one project. It forms a queue.
Put the second term back in and the same contract looks different. The discount was still solving the cash problem. It was also buying delivery cost that nobody had counted, at a price set before anyone knew what the data looked like.
A discount that fixes cash flow and creates a delivery cost
is not a
pricing decision.
It is a deferred one. And it comes due in a quarter you have already committed, to a customer who is not the one who got the discount.
Which meant the answer was not a price list. It was plumbing.
So the work was to hire a data engineering team and build reusable pipelines, because what you are actually selling in a services-to-product transition is repeatability, and price follows from that rather than the other way round.
Alongside it, contractual: caveats in the master agreement covering the time cost if the data turns out not to be clean. Unglamorous, and it is the clause that stops one customer's mess becoming another customer's delay.
You cannot charge for predictability you cannot deliver.
Two other things came out of the same read.
The buyer was usually the chief marketing officer, and a CMO does not care about your model. They ask four questions: what was your approach, can you explain it simply, why should I have confidence in it, and how do I apply it.
A team of PhDs answering a different question is not a communication problem. It is a product problem.
And the vertical mattered more than anyone had said out loud. Financial data has no standardisation. Retail and consumer goods data does. Same team, same work, completely different economics depending on which door the client came through. The strength of the business was not model building. It was model building where the input was clean.
What the business was for got said in two words, in the end. Not analytics, not forecasting. Decision intelligence. The noun is trust, and it came from watching what a CMO actually needed rather than from what the team was good at.
It became Continuum: behavioural economics, prediction and optimization in one platform rather than four services. Conceptualised, built through its first analytical dashboards, and then handed to the person on the team with the strongest product instinct. I kept mentoring them after I left, which is the only ending I actually want.
Contract sizes went up. Not cleanly. The company was still bootstrapped, and when the data-readiness bottleneck hit two customers at once the discount for payment up front came back out, because it was still the right answer to that constraint.
The difference was that by then everyone could see what it cost.
Six weeks does not change a culture and I would not claim it did. It changed what was in the model.
If you are discounting to solve something, it is worth knowing what the discount is buying on the other side of the contract.