How the analysis gets made

How we analyze the news

Each day we pull from a wide set of trusted ad-tech and media sources. Turning that firehose into one sharp read takes five steps:

1. Pull the day's stories. We collect from publisher feeds, trade press, company blogs, and primary sources, then extract the discrete stories inside each item — the actual claims, companies, and numbers — and set aside filler.
2. Group what's really the same story. The same news breaks across a dozen outlets with a dozen headlines. We compare every story to every other by meaning — not by matching keywords — and mathematically cluster the ones describing the same event. That collapses ten versions of one announcement into a single thread, so you read the event once, with the full picture, instead of ten overlapping summaries.
3. Weigh what matters. Each story thread is scored for importance to an ad-tech operator. Routine noise is demoted; the moves that change deal economics, market structure, or someone's roadmap rise to the top and earn deeper treatment.
4. Run it past a council of perspectives. The stories that matter most go through what we call the Decision Council. Instead of one flat take, the same event is examined through several distinct expert lenses — for example a Skeptic who steelmans the case against, an Operator who asks what breaks when you actually execute this on a Tuesday, a CFO who looks at the real cost at scale, a Customer who asks whether anyone actually wanted this, and a Long-Term Thinker who asks whether it still looks smart in three years. Depending on the story we swap in others — a Market Analyst for M&A and earnings, an Engineer for capability claims, a General Counsel for regulatory moves.
5. Find the tension, then take a stand. We don't average the lenses into a both-sides shrug. We surface the sharpest disagreements between them, identify what the decision actually hinges on, and synthesize a clear view. When — and only when — there's real conviction, the analysis ends with a single, falsifiable prediction in plain English.

That last step is the one we're accountable for. Every prediction with conviction goes on the public scoreboard and is graded right, wrong, or inconclusive when the outcome is known. No quiet edits, no memory-holing the misses.

How we analyze podcasts

Industry podcasts are where a lot of the real signal lives — but they're long, and most of each episode is banter and ad reads. We listen so you don't have to.

Each episode is transcribed in full, then distilled through the same operator lens into a one-page brief with a fixed structure:

• TL;DR — two or three sentences capturing the whole episode's thrust, specific enough that you know whether it's worth your time.
• What was covered — the discrete topics, with named companies, people, and numbers. Ad reads and host chatter are dropped.
• Notable claims & predictions — the sharpest "if true, this matters" lines, attributed to who said them.
• Fact check — where a sharp claim doesn't hold up, we say so, with the accurate version. We keep a high bar: we only call something false when we're solidly confident, and when it's borderline we name the speaker's incentive and what they left out rather than allege a falsehood.
• Names mentioned — hits from our watchlist of companies, people, and topics worth tracking.
• Why it matters for operators — the connection back to ad-tech strategy, operations, and buying/selling behavior. If the impact is low, we say so rather than forcing relevance.

The result: the substance of a 90-minute conversation in the time it takes to drink your coffee.

Why we keep score

Analysis is cheap when no one checks it later. The scoreboard is our answer to that: every confident call is dated, tracked, and graded in the open. It's how we earn trust — and how you can tell whether our read on the market is actually any good.

Get the analysis in your inbox →