The race to the cheapest decision

Consider a bank shopping for a fraud engine, or an agency choosing a system to decide who qualifies for a benefit. The buyer knows exactly what it wants: a system whose decisions are sound and, when challenged, defensible. That is the quality it is paying for. It is also, at the moment of purchase, the quality it cannot see. Two vendors sit across the table. Both demo cleanly, both quote accuracy figures, both promise diligence. One of them has spent real money on being able to account for every decision it makes — records of what it saw, the rules it applied, and enough state to reconstruct the outcome later. The other has spent that money on the bid. From where the buyer sits, they look the same. Only one number distinguishes them, and it is not the one that matters.

This is not a story about foolish buyers. It is a story about what a rational buyer does when the quality it wants is unobservable at the point of sale. Markets can only reward the qualities they can price, and they can only price the qualities they can see. When the decisive quality is invisible, competition does not stop; it relocates. It moves onto whatever dimension the buyer can read — and the most legible number in any procurement is price.

Buying on the only number you can see

Soundness and defensibility are not features you can inspect on a spec sheet. You cannot verify them by watching a demo, because a demo is a curated set of cases the vendor chose. You cannot verify them from an accuracy percentage, because the number is self-reported and says nothing about the decisions that will land on real people under conditions no demo covered. What you actually want to know — will this system make defensible calls, and can it show its work when one is contested — is precisely the thing that stays hidden until long after the contract is signed, if it ever surfaces at all.

So the buyer does what buyers do with unassessable quality: it stops trying to assess it and competes on what remains. Every dimension that cannot be verified is quietly set aside, not out of indifference but out of necessity, because a difference you cannot confirm cannot justify a difference in price. What survives the winnowing is the handful of attributes that are legible on paper, and the most legible of all is cost. The competition collapses onto it.

Now watch what this does to the vendor who invested in accountability. That investment was not free. Building a system that keeps an examinable record of each decision, that can be replayed, that can stand up when someone contests it — that is engineering effort, and it lands in the price. The cheaper competitor skipped it, and skipped the cost with it. If the buyer could see the difference, the premium would be a bargain: it is buying a decision the institution can defend. But the buyer cannot see it. All it sees is that one bid costs more and it cannot articulate what the extra money buys. In public procurement the effect is sharper still, because the rules were written to prevent favoritism and waste, and they do it by privileging the low, responsive bid. The very safeguard that keeps purchasing honest also strips out the buyer's discretion to pay for a quality it cannot document — and accountability, unpriced and invisible, is exactly such a quality.

The corner competed away first

Follow the logic one more turn and you arrive at something worse than the accountable vendor merely losing a deal. You arrive at a market that teaches every vendor to stop building accountability at all.

If the record-keeping, the replayability, the contestability all cost money, and if none of them can be seen at the point of sale, then they are the first thing a vendor under price pressure will cut. Not the last — the first. A visible feature, cut, shows up as a gap on the comparison sheet and costs you the bid. An invisible one, cut, shows up nowhere; it simply lowers your number and helps you win. Rational competitors will therefore compete away the accountability spend before they touch anything a buyer might notice, precisely because it is the corner nobody at the table can see being cut. The saving is banked immediately, in a lower bid; the cost of the missing account arrives years later, on someone else's desk, when a decision is challenged and there is nothing to produce.

A buyer who cannot see which decisions are sound will buy the cheapest, and the cheapest is the one that saved its money by being unable to say.

This is the lemons problem seen from the buyer's side of the table. Akerlof told it from the seller's — the honest seller of a good used car withdraws because he cannot get paid for quality no buyer can confirm, and the market fills with lemons. From the procurement side the same asymmetry runs the same way. The buyer, unable to distinguish the accountable system from the unaccountable one, cannot rationally pay more for the former; the vendor, unable to be paid for accountability, cannot rationally supply it; and the market settles on the cheapest unverifiable thing on offer. The good drives itself out. What makes the procurement view more alarming than the used-car view is the object being sold. These are not cars. They are the engines that will decide who is flagged as a fraud, who is found eligible, who is denied — decisions that land hard on people who never signed the contract and cannot see the corner that was cut in their name.

Making accountability a purchasable requirement

The failure here is informational, and so is the fix. You do not correct a race to the bottom by asking vendors to be more virtuous; you cannot moralize your way out of a mispriced market. You correct it by making the missing quality visible and required at the point of purchase, so that competition has something to compete on other than price — so that the vendor who refused to cut the corner is not punished for it.

Concretely, that means procurement standards that demand accountability as a condition of the bid rather than leaving it to the conscience of the seller. A specification that requires each decision to carry an examinable record — the evidence actually consulted, the rules in force at the time, enough state to replay the decision and test whether it holds — converts an invisible virtue into a visible line item. Once the requirement is written into what is being bought, the vendor who skimped no longer looks cheaper; it looks non-compliant. The bid that omits the record is not a lower bid for the same thing. It is a bid for a different, lesser thing, and the difference is finally legible on the page where it belongs.

This is the same move that assurance, insurance, and warranty make everywhere else that quality is hard to observe. A warranty lets a buyer reward the seller who is willing to stand behind the product, because the willingness is now a term of the contract rather than a private hope. Certification lets a buyer pay for a standard met, because a third party has made the meeting of it visible. Required, examinable records do for decision systems what those instruments do for cars and buildings and audited accounts: they give the buyer a reason it can defend for paying the accountable price. The Decision Receipt is one shape this can take — a per-decision record built to be inspected and replayed — but the deeper point is structural and does not depend on any one implementation. When accountability is a requirement, the competition is over quality that can be seen, and the accountable vendor is competing on the thing it is actually good at rather than being underbid by a rival whose only advantage was the corner it cut.

None of this asks the market to value accountability more than it does. It asks only that the market be allowed to see what it is buying. A buyer who can read soundness on the page will pay for it, the same way a buyer who can read a warranty will pay for the warranty. The race to the cheapest decision is not a verdict on what institutions care about. It is a verdict on what they are shown. Change what the bid must reveal, and you change which bid wins.

— Dispatches · Summit Cognitive