Notes on building GTM systems that compound.

Every GTM team we meet treats their CRM like the engine. It isn't. It's the dashboard. Here's the architectural shift that separates teams that scale from teams that drown.

A static prompt is a one-shot consultant. A learning agent reads its own replies and rewrites its instructions every Friday. Here's the loop, and the eval scaffold that keeps it honest.

One number collapses three independent signals into noise. Fit, engagement, and contract value want to be modelled apart — then composed at the moment of decision, not before.

If positioning lives in the brand bible, it's already lost. The teams who compound treat positioning as a rule the system has to obey — every email, every ad, every demo, every hire.

A new VP joining hits 100% intent and decays over 60 days. A funding round hits 100% and decays over 14. Most outbound is fired against signals that decayed weeks before the email left.

Six providers, ordered by hit rate × cost × latency. Most teams pay for the wrong vendor first because they don't measure marginal coverage. We do — and we publish the numbers.

Most founders post when they feel like it. The ones who win post on a cadence the system schedules — and review every draft against three rules: lived, contrarian, useful.

Direct mail, hand-cut landing pages, and SDR follow-up don't need an agency. They need a system. Here's how one operator runs a 300-account quarter without a single status meeting.

The fix wasn't the ad. The fix was the audience. We replaced LinkedIn's interest targeting with enriched CRM segments, and the same banners suddenly performed ten times harder.

Six queries, three pivot tables, one verdict. We run this on every new engagement before we touch anything else. Most teams fail the third query — usually because of one bad field.

Prompt generators produce average output for average teams. The unfair advantage is in the second-order instructions — the things you'd never ask a generator to encode because you didn't know they mattered.

Most trial follow-up is a four-email drip. The compounding teams treat trial as a research operation: enrich the user, model the team behind them, then send a message that names what they're trying to do.

Coverage, accuracy, and price by region — NAMER, EMEA, APAC. The vendor every consultant recommends is third in NAMER, sixth in EMEA, and last in APAC. Here are the actual numbers.

You don't get to invent a category in 2026. You get to occupy one with more clarity than the next three contenders combined. Here's the test we run before any positioning workshop starts.

The first AI-native RevOps hire isn't a configurator. They're a software author. The hiring loop hasn't caught up — here's the take-home test we use to find them.

Most content programmes confuse the two. A teardown changes a buyer's mind. A 90-second clip changes a candidate's. Treat them as separate functions with separate KPIs.

If you can't name the seven people at every target account, your target list isn't targeted — it's a database. The teams that pull this off cap it at thirty and rotate quarterly.

Hot leads are abundant; what's scarce is the patience to qualify them before sales squeezes them dry. The counter-intuitive lifecycle move that lifted our close rate eleven points.

The bar moved. A reasonable subject line and a personalisation token used to be enough. Now the floor is a paragraph that proves you've read the company's last earnings call.

Prospects don't watch demos until they trust the founder. The cheapest way to build that trust at scale is to publish a 40-minute conversation a month. Here's how to make it count for revenue.