A year-end audit of every prediction I made
I keep a public scorecard. Here is where I was right, wrong, and merely lucky.
Every prediction post should come with a follow-up that grades it honestly. This is mine. The rule here is simple: no quiet edits, no retroactive reframing.
The good calls
I called the Prebid consolidation and the retail-media surge, and both played out close to how I described them. Neither was an especially brave call in hindsight — both trends were already visible to anyone looking.
The bad calls
I got two things clearly wrong:
- I badly underestimated how long third-party cookies would limp along in practice.
- I was simply wrong about CTV pricing, which stayed messier and more fragmented than I predicted.
A prediction you never grade is not analysis. It is marketing for your own judgement.
The lucky calls
At least one call I got right for the wrong reasons. I predicted a soft CPM quarter — and there was one — but I expected it to come from weak demand. It came from a supply glut instead. Right line on the chart, wrong mechanism.
What I am changing
The pattern in my misses is consistent: I overestimate how fast the industry changes its plumbing and underestimate institutional inertia. Next year I am weighting that inertia more heavily, and I am writing fewer predictions with hard dates attached.