How we'll publish predictions, including our misses

Anyone can explain the past. Brand consultants are very good at it. After a company stumbles, there's always someone who saw it coming.

We think that's too easy. So every outside read we produce ends with a prediction on record: a specific, observable outcome, a check date, what would make us right, and what would make us wrong. It's written down before the outcome is known.

Some of these predictions will be wrong. That's the point. A method that's never at risk of being wrong isn't telling you anything.

Each prediction ends in one of three outcomes, and we define them before the check date:

• Supported: the outcome we named happened, within the time we gave.
• Contradicted: the outcome that would prove us wrong happened instead.
• Inconclusive: the evidence is mixed or insufficient, or conditions changed enough (a merger, a recall, a new owner) that the test no longer means what it did. We publish these too, with the reason.

Recording predictions also keeps us honest about confidence. If we can't name what would prove a finding wrong, we don't have a finding yet. We have an opinion.

When the check date arrives, we look, and we publish the result either way. The public record is planned to open on March 20, 2027. Until results come in, we have no predictive track record to show, and we won't claim one. You can follow it on our Predictions page.