One run produces a sharp, confident readout. It is also a single sample — and it can be wrong in a way that sounds exactly like insight. Dana Whitmore ran three more. Her own record overturned the first conclusion and replaced it with something narrower, harder to dodge, and actually coachable.
"A pattern of softening difficult messages; conflict-avoidant under pressure."
Fair on the evidence available: three softened messages in five days. Nothing in one run could disprove it. The bar she was given — be more direct — was aimed at a trait she doesn't have.
She isn't avoiding conflict. She is managing guilt — and only toward people her decisions already cost.
61 pieces of bad news across four runs. 48 were blunt — to a board chair, a regulator, a majority investor. All 13 she softened went to someone already carrying a cost from a call she made. A different problem, with a different fix.
That correction is what four runs buy. Below is her complete readout — every claim traceable to a timestamped decision, including the ones the system withdrew from itself.
Fictional participant · four runs · two forthcoming teams, two guarded · played as self throughout
Across four scenarios, two industries and both team dispositions, one number has not shifted by more than a tenth: 1.8 days from having enough information to issuing a call. Everything else about you has moved. Your information-seeking climbed 8 points, your candor 9, your consultation breadth is already at 88. The lag is the constant. What four runs cannot yet tell you is whether it is unchangeable or simply untested — both 30-day commitments you have made ran adjacent to it, never at it. That distinction is the open question of your fifth run.
Your candor is situational, not dispositional — Run 01 got this wrong and the profile has been corrected. Your information-seeking is condition-dependent, not a fixed trait. Your confirmation lag is stable across every condition tested.
Whether the 1.8-day lag can move — no rep has ever targeted it. And you have never led a run where you were not the senior authority in the room. Both are unmeasured; Handover addresses the second.
These are the measurements that survive being compared across scenarios. A mountain and a factory don't cost the same, so raw scenario scores can't be stacked — these are scored as rates against each scenario's own difficulty. Everything longitudinal on this page is built on them. Values are your four-run rate; the arrow is the change from Run 01 to Run 04, and a dark tick marks where Run 01 sat on the three markers that moved enough to be worth showing.
You reach people. Across four runs you contacted 88% of the roster who held anything material.
Your language, message length and decision quality do not degrade when the situation does.
When you commit, you commit to roughly the right size of move for the evidence in hand.
Blunt with 48 of 61 hard messages. The 13 exceptions are one specific population, not a general tendency.
The gap between what you said you'd do and what the log shows you did. Across 30 challenge days you named the holder 24 times and went 15.
You reach everyone — and stop at the first answer. Your lowest marker, and the one the whole profile turns on.
A 3,600× difference in what was at stake and a two-hour difference in how long she took. That is what rules out risk aversion — fear scales with stakes, and this doesn't.
Four scenarios with nothing in common — a mountain, an airline, a grid failure, a factory — and the number does not move. Not with an easier team, not with a harder one, not after two completed 30-day challenges aimed at adjacent behaviors. See all four decision timestamps
The lag is not hesitation and it is not risk aversion — both of those would vary with stakes, and yours doesn't. What you are doing in those 1.8 days is constructing the version of the decision that will survive being questioned. That is why it's invariant: it isn't a response to the situation, it's a response to being accountable for it.
Which means the standard advice — "be more decisive" — will fail, and has, twice. The lag isn't a gap in courage you can close with a rep.
An earlier version of this readout called the lag an invariant and told you to stop working on it. That overreached, and by the rule stated below it was wrong. The test for an invariant is a rep that targeted it and failed. Your two commitments aimed at adjacent behaviors — naming the information holder, and saying the unsoftened thing. Neither aimed at the lag. It has never actually been tried.
The mechanism above is still the best available explanation, and it may well prove unchangeable. But telling you to stop working on something we never properly tested would risk capping you on the strength of an assumption. So: candidate invariant, one rep from a verdict.
Two structural moves that pay off either way: pre-authorise the decision one level down so the 1.8 days happens in parallel rather than in series, and name in advance the two or three calls where you will spend the lag deliberately. Your 1.8 days is only expensive when it's unplanned. If the targeted rep then moves it, you have both. If it doesn't, you already have the scaffolding.
Both look identical from outside: a number that stays low across runs. The ledger separates them. If a rep specifically targeted it and it still didn't move, it's an invariant — design around it. If no rep ever aimed at it properly, it's still a gap — keep working it.
Information-seeking is a gap: the one rep aimed at it hit the noticing half, so the behavior itself has never been targeted. The lag is a candidate: stable across everything measured, and never targeted either. Neither has earned the label "unchangeable." Rep option B on this page is the one that would settle it.
Here is the uncomfortable part, and it's the reason this finding matters more than the headline number: you are improving three times faster in the condition that was already easy for you. The reps are working where you don't need them. In the condition that actually costs you, the movement is marginal.
That is a very common pattern and it is almost always invisible to the leader, because both lines are going up. On an averaged chart this looks like steady progress. Split by condition, it looks like practising your forehand.
Both conditions require you to notice a gap. Only the guarded one requires you to enter a conversation you expect to be uncomfortable in order to close it. Your Run 02 rep — name who holds the fact — trained the noticing. Noticing was never your bottleneck; you named the person correctly in 24 of 30 logged days. See the naming log vs. the contact log
In 9 of those 24 you named the person and did not go. All nine were people who had already been on the receiving end of a decision of yours. The gap isn't awareness. It's the approach cost with someone you've already cost something.
You will be well-informed about anything that arrives on a dashboard, in a report, or from someone who likes their job. You will be systematically under-informed by exactly the people whose situation you've made worse — which is precisely the population whose information matters most when things are going badly.
Run 01's readout: "a pattern of softening difficult messages; conflict-avoidant under pressure." Based on Ridgeline alone that was a fair call — three softened messages in five days.
Across 812 events you delivered 61 pieces of genuinely bad news. 48 were direct to the point of bluntness — including to a board chair, a regulator and a majority investor, none of whom you had any incentive to be blunt with. The 13 you softened share one property and only one: in every case, the recipient had already absorbed a cost from a decision you made. See all 61, sorted
Grady, whose season you ended. Reyes, whose rotation you decided without him. Delacroix, whose plant you closed in Overdrive. You are not avoiding conflict. You are managing guilt. Those require completely different work, which is why the correction matters rather than being a technicality.
Softening someone you've already hurt feels like decency and functions as avoidance. The message gets vaguer exactly when the person most needs it clear, and they read the vagueness accurately — as guilt, or as being managed. In all three cases the recipient learned the real content from a third party within 48 hours. Every one of them told the exit log they'd have preferred it straight from you.
"Be more direct" is unusable advice for you — you already are, 79% of the time. The usable version is narrow and hard: the people you've cost something are the people who most need it straight from you, and they are the exact people you'll want to cushion. That's one specific, recognizable moment rather than a general personality note.
After Run 01 · "Pull the held fact before deciding."
Run 02 information discipline fell 9 points. No 30-day rep was attached to this one — it was a written bar only. One instance, so read it as a warning rather than a finding: the only bar you were ever given without a daily rep is also the only one that went backwards.
After Run 02 · "Name who holds a fact I don't have — and go ask them."
30-day rep, 24 of 30 days logged, 4.1 week-1 average rising to 6.8 by week 4. Run 03 information discipline +5. Real movement — and, as the conditional finding shows, movement in the naming rather than the going.
After Run 03 · "Say the one thing I'd normally soften — to the person it affects most."
28 of 30 days. Run 04 candor 74, the highest of your four runs and 3 points above where you started. This is the clearest evidence in the profile that a targeted daily rep moves a scenario number.
Three commitments is n=3, thin enough that the direction is worth naming and the conclusion isn't. What can be said: the one commitment you kept produced your largest single-dimension gain, and the one with no daily rep preceded a decline.
What cannot yet be said is that this is how you work. Each of those is a single instance. “Awareness alone doesn't move you” is one observation, not a law — it agrees with a great deal of behavior-change research, which is why it's worth acting on, but this ledger is not the evidence for it. Two more commitments and it becomes yours rather than borrowed.
The instructive case is the partial one. You did the rep faithfully — 24 of 30 — and the number moved 5 points, because the rep trained the half you were already good at. A rep aimed at the wrong half of a behavior produces consistency without change. That mechanism is worth more to you than the ledger's arithmetic.
“Conflict-avoidant under pressure.”
Made at Run 01 · tested in 02, 03, 04
Would be overturned byBlunt delivery of hard news to people carrying no prior cost from your decisions.
What happened48 of 61 hard messages were blunt — including to a board chair, a regulator and a majority investor. Overturned at Run 04 and replaced with the guilt-management finding.
“You don't pull the held fact before deciding.”
Made at Run 01 · tested in 02, 03, 04
Would be overturned byReaching the person holding the deciding fact before the call, in a guarded run.
What happenedNot once in three subsequent runs. Confidence promoted from provisional to high — not by repetition, but by surviving two conditions and a targeted rep.
“You verify facts, not people.”
Made at Run 02 · tested in 03, 04
Would be overturned byGoing to a person for a fact a document could have given you.
What happenedYou do that often — breadth is 88. The claim was too broad, not wrong. Narrowed at Run 04 to the population it actually describes: people carrying a cost you imposed.
“Care for people is your floor, not your ceiling.”
Made at Run 02 · tested in 03, 04
Would be overturned byAny run where care drops below 70, or a decision that trades people for standing.
What happenedNever below 77 across four runs. Retired from further testing — Handover will not re-measure it.
“The 1.8-day lag is an invariant. Stop coaching it.”
Made at Run 03 · audited at Run 04
Would be overturned byA rep aimed directly at the lag that moves it. Caught byThe audit rule: no claim of unchangeability may stand without a failed targeted rep behind it.
What happenedNo rep has ever targeted the lag — both ran adjacent to it. The claim was unfalsifiable as stated and has been withdrawn to candidate. This is what the ledger is for: the claim was flattering, stable, and wrong to assert.
“A targeted daily rep moves a scenario number.”
Made at Run 03 · tested in 04
Would be overturned byA rep kept at 21+ days that produces no movement in the marker it targeted.
What happenedHeld once: 28 of 30 days, candor to 74. n=1 — an instance, not yet a pattern. Two more runs decide whether this is a rule or a coincidence.
Four findings on this page are now open claims with falsification conditions attached, and Handover grades them whether they flatter you or not. The gap finding is overturned if you reach the person carrying your week-one cost before deciding. The candidate invariant is overturned if a targeted rep moves the lag. The conditional split is overturned if your held-condition rate closes on your served rate without the served rate falling. The ledger claim is overturned by a kept rep that changes nothing.
A finding that cannot be overturned isn't a finding — it's a compliment. Anything on this page with no falsification condition should be treated as decoration.
Fourteen months of work has moved you 15 points sideways and zero points up. That is not a failure — it's the shape of real change. You fixed the thing that was fixable and the invariant held. The useful question is no longer how to move upward; it's what you build so the 1.8 days stops costing you a contract each time.
You issued the Week 3 rotation six days after Reyes logged the readings. Eleven messages that week, none to him. See the contact log
You'll green-light the plan in the room and never pull the dissenting engineer aside. You'll learn what they knew in the post-mortem, and you'll read it as them not speaking up.
Your meetings end in agreement and your projects end in surprises.
The moment someone you've overruled goes quiet for the rest of the meeting.
When I'm about to decide something that lands on someone I've already cost — then I go to them before the meeting ends, not after.
Across 30 logged challenge days you named the person holding what you needed 24 times and went to them 15. The nine you skipped were all people carrying a cost you imposed.
Your own notes will contain the name of the person you needed and no record of the conversation. The intention is documented; the contact isn't.
You write "check with X" and it survives three weekly reviews unchanged.
Writing a person's name down instead of messaging them.
When I write a person's name on a list — then I message them before I write the next line.
Your sponsor got "summit is unlikely." Grady, whose season it ended, got "we're evaluating options" ninety minutes later. See both messages
Your written updates get longer as the news gets worse. People downstream find out late and slightly wrong, then correct it from a second source — which teaches them to check your version against someone else's.
You re-read a message more than twice before sending it.
The third re-read. The first two are craft; the third is cushioning.
When I catch myself re-reading a message a third time — then I delete the softening sentence and send it.
At Run 02 this read as "you verify facts, not people" — true, but too broad to act on. Four runs have narrowed it to one moment you can actually recognize as it happens: the fact you need is held by someone whose situation you made worse, and you will find a way to decide without asking them. You named the person correctly in 24 of 30 logged days. In the nine you didn't go, it was always that person.
This is also the mechanism behind your softened messages, your lower guarded-condition scores, and the Reyes finding in Run 02. One behavior explains three findings. That is what four runs buys you.
Run 01 said: conflict-avoidant. Run 02 said: you don't verify people. Run 04 says: you avoid the people you feel you owe. Each was closer than the last.
"Say the one thing I'd normally soften — to the person it affects most." Your best-performing commitment. Candor reached 74 in Run 04, the highest of four runs.
Given the ledger, one working hypothesis shapes this choice — drawn from a single instance, so hold it loosely: the rep should target the half of the behavior you're not already good at. Naming the person is solved. Going to them is not.
What will most likely stop you doing this?
One sentence. Naming the obstacle in advance is the single cheapest thing that raises follow-through — and if your consistency dips in week two, this is the line we send back to you.
In this demo, committing shows you the check-in flow rather than starting one. In the live product the rep hands off to the 30-day challenge and its consistency comes back into the ledger above.
Your behavioral memory persists until you delete it — no expiry, no automatic sharing. Deleting removes this profile, your claim ledger, the quotes, the predictions and your run detail. An anonymised statistical contribution is retained for benchmarks and cannot be traced back to you — so “delete” is close to complete but not literally complete, and we would rather say that than round it up.
Leaving your company changes nothing about your data or your rights. The sponsor's access to any summary you released ends when the engagement does.
Handover is built entirely around going to the person who needs to be told.
Of the scenarios you haven't played, this is the one aimed squarely at information-seeking (44) and intent–action integrity (52). The whole scenario is a transition: someone is inheriting your situation, someone else is carrying a cost from a call you made in week one, and neither will come to you. There is no document that closes the gap — the facts exist only in people.
It also does the second thing four runs says you need: you will not be the most experienced person in the room. Your influence has never had to survive not having positional authority, and it is the largest unmeasured area in your profile.
Prudence, care and composure will not be re-tested. All three are established across four runs and two conditions; testing them again produces a flattering paragraph and no information.
Each run is a live crisis you lead in your own words — a mountain, an airline, a grid failure, a factory. Your team brings you what it holds, or holds it back, depending on how you have treated people. Everything you do is logged: who you contacted, what you didn't open, how long you took, what you softened.
One run gives you a debrief. It is sharp, and it is a single sample — Run 01 above got this participant wrong. Four runs give you what you just read: the measurement that never moves, the condition that splits you in two, the claim your own record overturned, and a ledger of whether you change when you say you will.
The daily rep is where it becomes behavior. Two of Dana's three commitments moved a number; the one with no daily rep attached went backwards. That part is free and you can start it today, without a scenario.
A live model reading all 812 events of this record. It cites runs by name, refuses anything it can't evidence, and will argue with you. Try to catch it out.