A1
observation
A coach was pulled toward teaching, not commission.
held fixed
The performance bar and the scope: developing others.
outcome
He met the bar on the way to something he already wanted.
Substrate 02 · People
A team
that decides.
Hold the bar.
Leave the method.
People act without being told. The work is to find the condition that changes their behaviour, then stop reaching into how they get there.
THE QUESTION What can someone do when you have removed the barrier, and what must remain theirs if the capability is going to leave with them?
The decision loop · not a management style
01 · Observe
Listen for the constraint inside what people say. Do not diagnose first.
02 · Wait
Let the pattern recur. A useful idea that has happened once is still an idea.
03 · Promote
Hold the bar, fund the reps, keep the consequence. Leave the method open.
04 · Carry
The person keeps the capability. The title is a formality.
Equinox · result
111% / 44 mo
To budget for 44 consecutive months after six years of underperformance.
Equinox · standard
7 of 8
Coaches finished above 100 sessions per month. One did not. The consequence was exercised.
LulaFit · capability
~35 → ~85%
Close rate for 30–40 remote contractors after the selling standard was built.
Find. Hold.
Promote. Carry.
This is the system. The same four moves have been applied to a gym team, remote sellers, and a team building agents. The practices change. The decision pattern does not.
applied to
01Findthe condition that is actually there
02Holdone thing fixed
03Promoteonly after recurrence
04Carrylet capability leave with them
Equinoxthen · coaches
constraint
Fewer opening hours. Ask: what can we control?
invariant
The bar, funded reps, and consequence. Their method stayed theirs.
pattern
The assessment → first-session interval held across enough leads to copy.
result
111% to budget for 44 months. Four moved on to manage elsewhere.
LulaFitbetween · remote sellers
signal
30–40 contractors needed a shared selling standard at a distance.
invariant
Value-based, question-based selling. Practice was funded; delivery stayed theirs.
pattern
Webinars, in-person sessions, and real practice made the capability repeatable.
result
Close rate moved from approximately 35% to 85%.
Agent teamnow · still running
failure
Handing people finished agents did not take. The artifact was the wrong intervention.
invariant
Same evals, constraints, and architecture. Their methodology stays theirs.
status
Still being tested. The pattern is live, not commemorated.
open
How to hold the conditions while the capability is built in someone else’s way.
the thread same method · different substrate · different resultthe discipline never promote the method
FEATURED APPLICATION · EQUINOX
From nearly shut down to one of the company’s best-performing teams.
The department had underperformed for six years. None of the six coaches had ever cleared the 90-session mark. I reset what it meant to be a coach, built the processes around that standard, and turned fewer opening hours into a question: “what can we control?”
The result was a team where seven of eight coaches finished above 100 sessions, and the department delivered 111% to budget for 44 consecutive months.
7 of 8
above 100 sessions
111% / 44 mo
to budget
team result · not a single-player lift
HOW I OBSERVE PEOPLE · OPEN / POROUS
Elicit.
Don’t instruct.
Three practices, one job: make the information you need cheap enough to surface. The instrument travels; the substrate changes.
the instrument · trained
What I listen for
- Motivation that already exists, but has not been attached to the work.
- Criticism that is expensive to give, so only the confident offer it.
- Progress that has become invisible because the barrier is in the way.
the rule · portable
What the observation changes
- Ask the fixed question before bringing the fixed answer.
- Make feedback routine, not brave.
- Remove the barrier before adding an incentive.
FIELD TEST
When the domain expert is reviewing rather than rewriting, the translation layer is working. The same test applies to a coach, a founder, or a team building agents.
HOW I CONTROL CONDITIONS · CLOSED / NARROW
One invariant.
Three freedoms.
A management intervention is only clean when you can name what is yours before you touch what is theirs.
column A · mine
Hold fixed
- The bar. What counts as done, stated before the work starts.
- The scope. How much surface the person owns.
- The consequence. Who absorbs the result if the bar is not met.
column B · theirs
Leave free
- Sequence. The order they work in.
- Style. How they coach, sell, or build.
- Motivation. What moves them. Find it; don’t replace it.
A2
observation
The Monday meeting was doing two jobs badly.
held fixed
Business and development each got their own time.
outcome
The coaches ran Wednesday themselves. The diagnosis came from red / yellow / green.
A3
observation
The eighth coach could not be brought up.
held fixed
The bar did not move. Reps were funded. Method stayed his.
outcome
Let go. Seven of eight is the result; the eighth is the proof.
THE TEST
Write Column A in one sentence before you intervene. If you find yourself editing Column B, you are reaching sideways. Handing someone a finished solution is a control move disguised as help.
The interval
that predicts.
When you are handed only trailing metrics, don’t hunt for a better metric. Look for the gap between two events already being recorded.
Two events.
One usable gap.
The blank version is portable. The filled version proves it was not invented after the result.
Equinox · then
Assessment → first free session
Both events were already dated. Under a week closed significantly better. Coach communication compressed the gap; the pattern was copied across the division.
Now · still running
Campaign due → campaign sent
Slip correlates with retention and satisfaction. It is readable weeks before the trailing metric moves. Measured, correlated, and watched—not yet promoted into an enforced rule.
Measured is not promoted. A pattern earns a rule only after it recurs across enough cases to survive being wrong.
Trailing metric · readable after the fact
Trailing metric · arrives too late
Trailing metric · no action left
the interval · leading and shortenable
Find two events your system already records. Measure the gap. Change the behaviour that shortens it tomorrow.
Proxies weren’t
good enough.
The method moved from people to machines. The control plane did not.
People · then
The proxy worked. I didn’t trust it.
Thirty to forty remote contractors were almost never seen working. In-person audits, taking their classes, and attendance data ran the business well enough—and not well enough to trust.
Machines · now
Build the recorder first.
When Claude Code changed its interface and the reasoning went behind a wall, I refused another proxy. The recorder came first—thirteen days before the system it now watches.