The Both-Sides Formula
Navigate to the real story by combining accounts instead of picking a partisan version.
- Difficulty
- Moderate
- Time to result
- ~weeks to results
- Steps
- 6
- Confidence
- 72%
Stone describes his working method for Amazon's pandemic year, where two finished narratives were already circulating: organised labour and internal critics casting the company as reckless and indifferent, and Amazon casting itself as the flawless pandemic hero. His formula is to treat both as artefacts rather than evidence, then talk to as many people as possible across every camp, combine their accounts, and navigate to the story that survives all of them. The output is deliberately unbalanced in places: Amazon did hire virologists, did get lost in the confusion of March 2020, did make mistakes and did fire whistleblowers. He explicitly refuses both the critic and the hagiographer roles, aiming instead for a story that critics and employees can each read and take something from.
Origin
Extracted from Acquired, where Brad Stone described how he reported Amazon's COVID response for Amazon Unbound and called the method 'basically the formula'.
Core principles
- 01In the moment, everyone is telling a simplistic version of the story.
- 02The two loudest accounts are both constructed, not observed.
- 03Volume of sources beats access to any single source.
- 04Combining accounts is different from splitting the difference.
- 05A good account lets opposite readers take different things from it.
How to run it
- 1
Write down both partisan versions first
State each side's story in its strongest form before you report. For Amazon's pandemic year that meant the reckless-employer account and the flawless-hero account, side by side.
Pro tip Use each camp's own vocabulary; the loaded words tell you what they are defending.
- 2
Assume neither version is the story
Treat both as evidence about the tellers rather than about the events. The simplistic version circulates first because it is useful to someone.
Watch out Your own prior usually matches one of the two versions; name it early.
- 3
Talk to as many people as you can, in every camp
Cover current executives, departed executives, front-line employees, organisers and outside advisers. Departures are the highest-yield group because they carry specifics without a current-quarter script.
Pro tip Keep interviewing after the story feels settled; the reconciling detail usually arrives late.
- 4
Collect decisions, not verdicts
Ask what was decided, when, by whom, and what was known at the time. Verdicts cannot be reconciled across sources; dated decisions can.
Pro tip Anchor every account to a specific week — March 2020 confusion looks different from July 2020 hindsight.
- 5
Combine the accounts rather than choosing one
Lay conflicting versions of the same event against each other and keep what more than one source independently supports. Contradictions that survive get reported as contradictions.
Watch out Combining is not averaging; do not manufacture a midpoint no source describes.
- 6
Publish the version that keeps the inconvenient facts
Include what damages both narratives, and resist supplying the moral verdict. Different readers should be able to take different things from the same account.
Pro tip If neither camp can quote you cleanly, the reconciliation probably worked.
Watch out Expect both sides to read the balanced account as evidence of the other side's influence.
In the wild
Critics inside and around the company, particularly in organised labour, were characterising Amazon as irresponsible and indifferent to employee risk. Amazon was simultaneously presenting itself as an unfairly maligned pandemic hero that had made no missteps. Stone reported both and kept the facts that broke each: the company hired world-famous virologists to advise it, and it also got lost in the confusion of March 2020, made mistakes, and fired whistleblowers whose departure other executives thought was wrong.
→ An account that neither camp can quote cleanly, and that readers with opposite priors both found factual.
Rather than choose between 'marketplace globalisation is great for customers' and 'marketplace globalisation destroyed Western sellers', Stone went back to the individual sellers Amazon had celebrated in past Bezos shareholder letters and took their temperature years later. Nearly all had become disgruntled; one who had adapted his product mix was still doing well and was notably less embittered.
→ Both claims survive in the same chapter, with the seller-level detail showing which sellers each applies to.
Common mistakes
Balancing by splitting the difference
Averaging two narratives produces a claim nobody made and no source supports. The method is reconciliation of specifics, not a midpoint.
Treating access as the story
Waiting on the one central subject can hold the work hostage. Stone never got Bezos and argues the book did not suffer, because the disciplined subject would have recited polished, already-published answers.
Letting one embittered source set the arc
A single vivid account of grievance is easy to build a narrative around, which is exactly how the simplistic version gets written.
Is it for you?
Best for
Writers, analysts and researchers covering a company or event where organised narratives already exist.
Not ideal for
Fast turnaround commentary where there is no time to reach sources on every side.
From the transcript
“I think that is easier than it looks because in the moment everyone is telling a really simplistic version of that story.”
“And so talking to as many people as I could and combining their accounts and looking at both sides and trying to navigate the real…”
“it's really I try to tell the warts and all story and then hope different people can take different things out of it”
From the episode
Special: Amazon Unbound (with Brad Stone)