Theranos and Walgreens: launch on a famous board’s word, or test first?

The room

The psychology: escalation of commitment

Escalation of commitment describes what happens after a first decision. Once people have backed a course of action, they tend to read new information through the need to show that the original choice was sound. Bad news becomes a reason to try harder rather than a reason to stop, and the pressure is greatest on the person who made, or championed, the first call. In Barry Staw’s classic experiment, the people who put the most fresh money into a failing project were the ones personally responsible for choosing it.

Two related shortcuts often feed the escalation. Authority bias leads people to give extra weight to a claim because of who stands behind it, and social proof leads them to treat other people’s commitment as a sign of quality. Research on young biotech companies found that prominent partners and investors raised a venture’s valuation most when its quality was hardest to observe. In other words, endorsements matter most exactly when nobody can yet see inside the product.

The Walgreens decision shows both forces working together. By 2013 the partnership was three years old, and a validation study that never happened had become a delay to work around rather than a finding in its own right. Meanwhile a board that came to include former secretaries of state, senators and generals supplied reassurance on a question none of them had been asked to answer: whether the tests agreed with an accredited laboratory. Each year of commitment made the missing data easier to tolerate and harder to demand.

Why smart people fall for it

Three mechanisms turned a promising pitch into a commitment that was hard to question.

1. 1

   Prestige stood in for proof

   A board that grew to include former cabinet secretaries and generals signalled quality to everyone around the deal. A board’s reputation, however, is not a method-comparison study.

2. 2

   Every year raised the exit cost

   By launch the partnership had run since 2010. Each meeting, payment and public plan made stopping look like an admission that earlier judgments had been wrong.

3. 3

   Missing data went unpriced

   The evaluation that never took place and the hard-to-compare tests were treated as delays to manage. They were themselves the most important information on the table.

Was trusting Theranos a reasonable call in 2013?

In fairness, the decision had real logic. Bringing basic health services into pharmacies was an attractive strategy, a finger prick is easier for many people than a needle, and the first chain to offer cheap, fast testing across thousands of stores could have shaped the market for years. Large companies regularly partner with start-ups whose technology they cannot fully inspect, because the start-up’s intellectual property is the reason for the deal.

Nor was the board’s prestige meaningless. Directors with long public careers have reputations to protect, and investors read their presence as a sign that someone serious has looked inside. The research by Stuart, Hoang and Hybels shows that markets do reward such endorsements. It also shows the reward is largest where quality is hardest to see, which is a description of how people decide under uncertainty, not proof that the endorsers checked anything.

What turned a defensible bet into a costly one was the specific gap. The question that mattered for a pharmacy chain was narrow and testable: do these results match those of an accredited laboratory on the same customers? Walgreens had tried to get that answer in 2011 and, according to later reporting, had not been able to. Going ahead meant accepting the partner’s description of its own product in a field where results guide medical decisions.

The make-up of the board made it a weak substitute for that answer. Fortune’s 2015 review of the twelve directors found that six had no medical or technology background and none had formal accounting, auditing or legal expertise. A board assembled for reach and reputation can be excellent at opening doors and still be poorly placed to test an assay. Reading it as a scientific endorsement was a category error, and an easy one to make.

What does the research say about prestige and persistence?

Staw’s 1976 study put 240 business students in the role of a corporate financial officer. Some made an initial research and development allocation themselves; others inherited it from someone else. Some then learned the chosen division had improved; others that it had declined. When the students had made the original choice and it had gone badly, they put the most new money into it. Personal responsibility, not the numbers, drove the extra investment.

That pattern maps closely onto partnerships. The executives who argue for a deal own its success in a way that later reviewers do not. Each step forward, a pilot store, a press announcement, a further payment, is easier to justify than reopening the first decision. Escalation rarely looks like one reckless gamble. It looks like a series of reasonable next steps, each taken because the last one was already taken.

Prestige accelerates the process. Stuart, Hoang and Hybels studied 301 venture-backed biotechnology firms and found that ventures with prominent partners and investors went public faster and at higher valuations, and that the effect faded as firms aged and became easier to judge on their own record. Endorsement works as a substitute for direct evidence. That is useful when evidence is genuinely unavailable and dangerous when it is merely unrequested.

Put together, the two findings explain why the Theranos partnership was hard to stop. The people who had backed it had personal reasons to keep backing it, and the famous names around the company gave everyone a respectable reason to feel the evidence question had been handled elsewhere. Neither effect requires anyone to be careless. Both simply move attention away from the one test that would have settled the matter.

How would a modern partner validate a breakthrough claim?

Start by separating the commercial question from the technical one. Whether a partnership is good business depends on demand, margins and brand fit. Whether a diagnostic test works depends on one thing: agreement with a recognised reference method on the same samples. The second question should be answered before the first is allowed to create momentum.

Make that answer a contractual precondition. Name the independent laboratory, the number of paired samples, the tests to be compared and the level of agreement that counts as a pass. Agree in writing that the launch date moves if the data do not arrive. A partner confident in its technology loses little under these terms; a partner who refuses them has told you something important.

Design the pilot so it can fail. A small number of stores running split samples, with results checked against an accredited lab before they reach customers, produces evidence within weeks. Decide in advance what result pauses the rollout, and give that decision to someone who did not sponsor the deal. Escalation is weakest when the person judging the evidence has nothing to defend.

Finally, ask what each endorser actually examined. A board, a famous investor or a glowing magazine profile can tell you that powerful people are enthusiastic. Only data can tell you that the product does what it claims.

The manager's playbook

1. Name the claim that decides everythingWrite down the single testable claim the deal depends on, such as agreement with an accredited lab, before discussing anything else.
2. Make independent evidence a launch conditionPut third-party validation, with a named lab and sample size, into the contract as a precondition rather than a follow-up.
3. Set exit criteria before you commitAgree in advance what result, delay or refusal pauses or ends the deal, so stopping later is a rule rather than a confession.
4. Separate the sponsor from the evaluatorThe executive who championed the partnership should not be the one who judges whether the evidence is good enough.
5. Read endorsements for what they coverAsk what each famous backer actually examined. A reputation for judgement is not a review of the data.

Questions people ask