Strategy Decision Cases

Crisis response: 70% of these decisions were never reversed, and that is the problem

BP minimized and deflected after Deepwater Horizon. The response held. So did the criminal, civil and reputational costs. Seventy-one crises on record, and why persistence is the wrong thing to want here.

Read this rate backwards

Across the 71 verified crisis decisions in the record, 70% were never reversed. On every other page here, a high persistence number is at worst neutral. On this one it is the finding.

A crisis response that is never reversed usually means the company picked a posture in the first week and held it while the facts moved. The record is explicit about what that looks like. BP's choice after Deepwater Horizon is summarized as minimize and deflect; the verdict is that the minimizing response stood and produced the criminal, civil and reputational costs that followed. Anheuser-Busch InBev's response to the Bud Light boycott was a both-sides neutral statement and the removal of executives; the verdict is that the non-answer stood and the documented multi-year decline followed from it. Meta framed Cambridge Analytica as a hypothetical risk; the framing stood until regulators penalized it.

Three decisions that held. Three companies that would have been better off changing their minds.

The cheap exit that was not one

3M took a narrow quiet settlement on PFAS in 2018. The record's verdict is that the cheap narrow exit was superseded by a $6 billion settlement after the bankruptcy shield was dismissed. Credit Suisse overrode risk for revenue in 2021; it did not survive independently, ending in a UBS rescue takeover two years later.

What these have in common with BP and Bud Light is not the severity of the underlying event. It is that the first response was sized to the version of the problem that was visible in week one, and nothing in the organization was set up to resize it.

What the ones that worked actually did

Apple appears twice and both entries are worth separating from the pattern. In 2016 it refused the FBI's request and fought publicly; it never built the software, the order was vacated, and its security architecture stayed intact. In 2019 it rationed minimum concessions on independent repair, releasing access while continuing to fight bills that banned parts pairing.

Neither is a communications strategy. Both are positions with a stated boundary — this is what we will give, this is what we will not — decided against the thing being protected rather than against the volume of the criticism. That is the difference between a posture that holds because it was chosen and one that holds because nobody re-opened it.

How to run this decision

Drawn from what separated the records that stuck from the ones that were undone. A step nobody in the record can be seen taking is not here.

  1. 01

    Separate the event from the exposure in the first 48 hours

    The event is what happened. The exposure is what it could cost if the worst credible version is true. 3M priced the event. The exposure was $6 billion.

  2. 02

    Name what you are protecting, in one line

    Apple was protecting a security architecture, and everything it conceded and refused follows from that. If the answer is the share price or the quarter, you will make concessions that cost more than they buy, because there is no boundary to measure them against.

  3. 03

    Put a date on re-deciding, not on responding

    Most of what goes wrong here is a week-one posture surviving into month six. Set the date now, while it is cheap, for when the response gets re-opened against what is then known. Write it in the calendar of someone who is not running the response.

  4. 04

    Give the person who can change the answer no operational role

    The team executing a response cannot also be the team that judges whether it is still the right one. Every held-and-failed case in this record had a response that nobody senior was tasked with contradicting.

  5. 05

    Say what you will not do, publicly and early

    A boundary stated in advance is worth more than a concession made later, and it is the thing that stops the response drifting. Apple in 2016 said what it would not build, and that was the position for the duration.

  6. 06

    Price one more quarter of not acting

    This is the question the record keeps answering expensively. Whatever the response costs, compare it to the cost of the same decision taken three months later with the same facts. In this corpus the delta is almost never small.

What it looks like when this is going wrong

  • The response was sized to the week-one version of the facts and has not been resized.3M, 2018. BP, 2010.
  • The statement is written to avoid offending any group, and therefore commits to nothing.Bud Light and Anheuser-Busch InBev, 2023.
  • The issue is described internally as a hypothetical or potential risk.Meta, 2018.
  • The people running the response are also the people judging whether it is working.The eleven reversed responses on record.
  • A risk function was overruled for a revenue reason and the override was not escalated.Credit Suisse, 2021.