Sequencing and week one

Three horizons ordered by dependency rather than ambition, and the seven things I'd need in the first week before any of it is worth starting.


Ordered by dependency, not by ambition. Nothing in Horizon 2 works if Horizon 1 is skipped, and the reason to publish the order rather than keep it is that sequencing is the part of a plan that gets quietly reversed under pressure. Written down, a reversal has to be argued for.

Horizon 1 — instrument before spending

Funnel event schema through to first credit purchase. Per-surface community response measurement, right-censored, so that a two-hour-old thread with no reply is counted as new rather than abandoned. An answer-engine citation baseline panel, built before anything is optimised against it. A retrospective on the coupon cohort, which is the one activation that has already run and is still readable. And an inventory of the cloud-marketplace listings — whether any exist yet, and whether what exists is indexed — because a listing is an answer-engine surface and the baseline panel is incomplete without it.

INFERRED None of this requires new budget, and all of it is prerequisite to defending any. Attack that on one axis: if marketing touch cannot be joined to usage revenue at account level, Horizon 1 stops being a marketing project and becomes an engineering dependency with a different owner and a different timeline. That is the fourth question below, and it is the one whose answer I would want first.

Horizon 2 — prove one channel and one harvest

One paid channel taken to statistical significance behind a pre-registered holdout, rather than four channels taken to anecdote. In parallel, one high-volume community topic harvested into canonical public pages, with citation lift measured against the baseline panel from Horizon 1. Two proofs, one paid and one earned, read off the same funnel definition.

INFERRED Running them in parallel is deliberate and it is the argument of this whole document in miniature: if the paid proof and the earned proof are measured on different funnels, the comparison between them is unavailable, and the question of how to split the next dollar has no evidence behind it.

Horizon 3 — scale what cleared, kill what didn’t, then hire

Apply the kill rule as written, without relitigating it once the failing channel has a champion. Put a standing experimentation cadence on the credit-purchase flow. Run events as a calendar with a retention metric attached rather than as a series of nights.

Partner recruitment at volume belongs here too, and so does the partnerships hire. Signing systems integrators before the pipeline definition exists produces contested deals rather than revenue, and the relationship capital that recruitment needs is the one input on this page I can’t supply myself. Both wait for a working measurement, for the same reason everything else does.

Hiring comes last on purpose, and it is the ordering the horizons above already argue for: a team built around a proven engine inherits a measurement, and a team built ahead of one inherits a forecast. I’d rather be judged on the first.

What I’d need in week one

UNKNOWN Every item below is something I do not have and cannot get from outside. They are listed so that nothing else in this document reads as though it assumes an answer.

INFERRED If the answers to the first three show ARPU is bimodal — a long tail of small accounts against a short head of production apps — blended CAC is the wrong metric and the paid section needs rewriting around segments before a dollar moves. Naming that now costs me the appearance of certainty. Discovering it in month two would cost a quarter.