Four weeks of standardised, weekly outcome tracking — built to move every account as far up the ladder as it will go, and hold the floor while we do it.
In July we lost 8 accounts and net churn hit 8.51% — the highest of the year. The part that should change how we respond: the exits came out of the green base. Green fell from 74 to 65 against 9 offboardings. Red and amber barely moved.
We were not losing accounts we were already worried about. We were losing accounts that looked healthy — because "healthy" only ever meant "no signal reached us." Working the red list harder would not have caught any of them. By the time an account reaches escalation the decision is usually made: cancellation requests saved 0 of 3 in July.
Green accounts are also where every referral and expansion we have ever had came from. Referrals were historically our largest growth engine, and they have gone entirely organic.
Two problems, one answer. Working green accounts harder is simultaneously the churn defense and the referral offense. A green account is not a won account — it is an account where nothing is currently wrong. That is a floor, not a finish line.
So this is not a churn sprint. Zero preventable churns is the floor. The purpose is the climb above it.
Every standing outcome sits at a level. This is the bar the team is coaching toward, and the thing both EAs and AMs need to be fluent in.
Because your EA is on it. Reliable, manual, dependent on one person remembering.
Not on memory. Documented, repeatable, survives a handover.
It happens on its own. Your EA handles only the exceptions.
It is not a definition of automation. The level describes what the client experiences, never the mechanism. Automation is the usual road to L3, not the meaning of it. And it is not a grade. Level is a state that moves over months. We measure movement, not altitude — most of the bench will not reach L3 everywhere, and an account that climbs L1 → L2 on three outcomes is the sprint working.
Where an outcome can't reach L3, the gap is either a skill template we can supply, or the honest basis for an upsell conversation. AI & Automation is no longer one of the standing outcomes — it was never a tenth thing you do, it is the dimension that separates L1 from L3 on all of them. Standing outcomes are now eight.
Tiered the way the ladder is. Holding the floor is necessary and not sufficient.
Signal coverage — the share of accounts where every signal artifact actually exists. A good result on partial coverage is luck, so this gets stated before any other number.
Preventable churns. Tier movement. Outcome rows on track. Divergence between what the EA sees and what the client feels.
Two hard targets, both on the work rather than the outcome: every account has a named 10/10 path, and every account had a week where someone actively pushed a level. Both are checkable. Level movement, referral and expansion conversations are then reported as counts, each with what caused it.
Level movement is the headline number and it deliberately carries no target. Levels move over months, not weeks, so a target would raise the count by marking levels up rather than by making accounts better — the same trap we avoided with wow moments.
The ask is a moment every week on every account: something the client did not expect and would notice. A saved deadline, a problem caught before they saw it, a small thing done unasked. It does not have to be big; it has to be unexpected.
We deliberately do not count them. If we asked for a weekly number it would go up without anything changing for clients, because people would start reporting ordinary good work to have something to submit. Instead, every Account Brief carries a standing answer to one question: what would be a wow moment for this client? Specific, named, refreshed when it has been used. An AM who cannot write one does not know the client well enough yet, and that is worth knowing.
What we measure is what those moments cause — accounts that climb a level, and clients who start talking.
Referral counts report at 90 days rather than inside the sprint, since the lag between a wow moment and a referral is longer than four weeks. A reporting note, not a reason to aim lower.
No new reports. The grading lives on a new tab in the EA Role Scorecard the client already has. SODs and EODs continue unchanged — they answer what are you working on; this answers what did you deliver.
Every account tiered red / yellow / green against a shared definition. Account Briefs updated where something moved.
Marks each standing outcome on track, at risk, off track or paused, with one line of context. Files a short note to their AM.
Catches anything urgent and preps the EA for the client conversation. Deliberately does not judge yet: the client hasn't spoken.
The client grades the same outcomes in their own words, in their own column.
Now both grades exist. Where they disagree is the signal — and it is what the AM works next week. Live conversation only where something triggers it.
About 10 minutes per EA per week, and three hours per AM in weeks 1–2 falling to two across a book of 17–22 accounts. AM capacity was already flagged as at its limit in the July EOM, so the design deliberately adds a tab rather than a report, and keeps most of the weekly loop asynchronous.
The extra hour early is chasing EAs who haven't filed — a same-day message in week one, a live conversation in week two. That was missing from the first estimate (thanks Wela). It front-loads onto exactly the weeks the motion feels most fragile, so it is budgeted rather than discovered. And it is not overhead: a missing grade is the blind signal we are trying to surface, so the chase is the coverage work. If it has not dropped by week three, the rollout conversation did not land — that is a finding about the rollout, not about AM capacity.
Eight accounts first. We widen based on what those eight teach us, rather than instrumenting the whole book on a date and hoping.
Each AM picks one or two and writes their own brief; I join the actual EA conversations and client gradings, live. The bar gets set by demonstration rather than description, and coaching happens in the moment instead of a week later through a report.
Not a fixed date for a book-wide switch. At the gate we decide what changes and how many accounts join next, using real numbers on what the motion cost and whether EAs actually adopted it. If it needs another cycle at eight, it gets one.
The sprint can no longer claim it made the book visible. Coverage becomes a measure of the enrolled set, not of all 81 accounts, and un-enrolled accounts stay uninstrumented on purpose. That means July's failure mode is still live on most of the book through September — a green-base exit on an un-enrolled account is one this sprint would not have caught.
Accepted, because a motion that half-works across 81 accounts surfaces nothing anyone can trust, and both AMs who have reviewed this flagged adoption rather than design as the thing that decides it. One thing does not phase: all ~81 accounts still get tiered every Monday from day one. Tiering is cheap and it is the book-wide signal we keep while grading rolls out.
Seven artifacts. Templates are standardised; judgment is left to the AM.
What the end-of-sprint report measures against.
| Measure | July 2026 |
|---|---|
| Active clients | 86 at month end · 81 as of Aug 4 |
| Offboarded | 8 terminated (1 buyout) |
| Net churn rate | 8.51% — highest of 2026 |
| Escalations | 11 total · 4 saved · 4 churned · 3 pending |
| Save rate (closed escalations) | 50% |
| Health distribution | Red 5 Amber 12 Green 65 |
| AM workload | Wela 24 · Bits 22 · Mel 20 · Chaddy 17 · Cesar 3 |
If you saw an earlier version, three things moved.
It stopped being a churn sprint. Zero preventable churns is the floor now, not the point. The measures are tiered — can we see it, are we holding, are we climbing — and the headline number is accounts that moved up a level.
The ladder is new. Every outcome sits at L1, L2 or L3. That is the bar we are coaching toward, and it is a growth path for the EA as much as a standard for the client.
Referrals are in. A standing goal of three per client, conversion not required — that is sales' job. Counts report at 90 days because the lag is longer than four weeks.
Genuinely open, and better answered now than in week three.
Running this against your real book — what breaks first?
The design now assumes three hours a week per AM in weeks 1–2, two after, and ten minutes per EA. Where is that still wrong, and by how much?
What would make the brief worth opening on a Friday rather than one more thing to fill in?
Where does this collide with how you already run your accounts — and which of the two should give?
Previewed with Chaddy, Mel and Colin. EA rollout conversations run Aug 19–21; all four AMs start together on Aug 24.