Practices
Practices, attributed
One page, one approach, one named practitioner, one source you can open — plus our reading of why it works and where it breaks, marked as ours.
How it works reported, sourced
Why it works RevGuild analysis
Where it breaks named limits, not a pitch
5 practices · 5 practitioners · every quote links to a source we fetched
Usage-based compensation models
Graham Collins names seven usage-based commission models. We run the arithmetic, the rep behaviour each one produces, and the point where each breaks.
Solves: paying full commission at signature when the revenue arrives monthly and may never fully arrive
CLAFTS (Claude Drafts)
Anthropic's Jared Sires has Claude pre-draft his inbound customer replies into his Gmail drafts folder. Nothing auto-sends — his work becomes approving a queue.
Solves: reply volume grows with the account book, and eats the hours the seller needs for selling
Five-dimension scoring rubrics, tuned on a test territory first
Anthropic's Travis Bryant scored 4,000 accounts overnight against two written rubrics, tuned on a test territory first. How it works, and where it breaks.
Solves: territory planning where account priority is either a gut call or a model score nobody can inspect
To-go coverage
Dave Kellogg's to-go coverage — divide pipeline by what is left to sell, not the full quarter target, so the ratio still means something in week six.
Solves: pipeline coverage ratios that are only meaningful on day one of the quarter and quietly mislead every week after
AI-native growth engineering team
Fyxer's growth engineering team ran 360 experiments in a year. The payoff wasn't volume — it was affording the test nobody wants to run.
Solves: the tests that would move the funnel most are the ones nobody can afford to be wrong about in public