Why we don't send a proposal without a diagnosis

Three quotes and the cheapest wins is a sensible-sounding way to buy, and for many purchases it is. The model works when the thing being bought is defined precisely enough that all the quotes concern the same work. In expert services that condition is rarely met, because producing the definition is the very work being bought. Answering a request for quotation then means putting a price on something nobody has examined yet. We do not do that, and this text is the reason written out in full, so you can judge it for yourself.

The price attaches to the representation, not the problem

A request rarely arrives in its original form. It has travelled through the company, picked up a category and finally a name, and the name often points in the wrong direction. A proposal made directly against the request prices the latest representation of the problem rather than the problem. The quote can be arithmetically precise and still concern the wrong thing, and the precision only makes the error harder to spot.

Competitive bidding does not correct for this. It rewards it. The winner is whoever priced the request most literally and asked the fewest questions.

A guess meets one of two fates

A supplier pricing an unknown protects themselves in one of two ways. The first is padding: a cushion for surprises gets baked into the price, and you pay for an insurance policy nobody tells you exists. The second is underpricing, and it is the more expensive of the two. Auction theory has a name for the mechanism, the winner’s curse: when several parties guess at the same unknown value, the contest tends to be won by whoever guessed most wrongly.¹

An underpriced project does not stay unprofitable. It becomes a business model. The margin gets recovered through change orders, so every surprise uncovered along the way is revenue for the supplier and a negotiation for you. In that model a dispute is not an accident. It is a built-in feature waiting for the first ambiguous boundary.

The average lies about the tail

How large is the risk of the unknown, then? The largest measured dataset comes from IT projects: across a sample of 1,471 projects the average budget overrun was 27 percent, which sounds manageable. The number that matters is a different one. One project in six overran its budget by 200 percent on average.² The risk does not live in the average but in the tail, and the tail is at its fattest precisely when the problem has not been defined.

A fixed price does not remove that risk. It only decides who carries it, and at what interest rate. A fixed price that skipped the diagnosis means someone agreed to carry tail risk blindfolded, and risk carried blindfolded always gets priced in somewhere.

An order in which money meets knowledge

Our order is written on the services page, and it has three steps. First a free 25-minute conversation, whose anatomy we opened up separately. If a direction emerges but the framing needs real work, the next step is a paid scoping phase whose output belongs to you whether or not we continue. A proposal comes only once we know what is being fixed, and it is phased so that you can stop after any phase.

This order has one consequence worth saying out loud: if the scoping shows the problem to be smaller than the request, the proposal shrinks with it. That is not charity but arithmetic with a longer horizon than a quarter. A proposal given after knowledge is a promise. Before it, it is a bet, and the stakes are yours.

A checking question for the buyer

You are left with one question, and it works in any industry: did the bidder ask anything they did not already know the answer to? If the proposal was born without a single such question, you know whose problem it priced.

From words to done.

Sources

1. Thaler, R. H., “Anomalies: The Winner’s Curse”, Journal of Economic Perspectives, 1988. https://www.aeaweb.org/articles?id=10.1257/jep.2.1.191
2. Flyvbjerg, B. & Budzier, A., “Why Your IT Project May Be Riskier Than You Think”, Harvard Business Review, 2011. Average overrun 27%; one in six projects overran by 200% on average. https://arxiv.org/abs/1304.0265