A complete story · demo data, real engine
Retention alarm — when regretted loss moves while totals look fine
The Quiet Exodus
Every headline number says calm; underneath, the people you can least afford to lose are leaving — and not for the reason everyone assumes. Every beat below opens on a pre-built scene from the deck, is confirmed by the matching engine against the data (never asserted), and toggles to the actual numbers. The full story template — its cast, its branches, the certainty ladder — is laid out beneath.
Act I — The World
where we standTHE REASSURANCE
Headcount (12 quarters)
Is the workforce stable?
the scene — pre-built from the deck
The story is: no story
Headcount (12 quarters) is flat across the window — the wiggles never clear the certainty cutoff. The honest reading is stability.
the handle · Report 'stable' out loud — the calm is real, and it's why what follows matters.
Act II — The Forces
what's moving— BUT —
THE ALARM
Regretted attrition — high performers, annualized %
Are we losing the people we can't afford to lose?
the scene — pre-built from the deck
Past the benchmark — and climbing
Regretted attrition — high performers, annualized % now stands at 15.1 — +58.9% above Industry benchmark (9.5) — and the trend is confirmed rising, not noise.
the handle · The corrective target is the benchmark line — date it, and report against it monthly.
— UNDERNEATH —
THE HIDDEN SPLIT
Regretted exit rate % by team (TTM)
Is it everywhere — or somewhere?
the scene — pre-built from the deck
Some of you are over the line
1 of 6 groups sit above Industry benchmark (9.5) on Regretted exit rate % by team (TTM) — the org-level average hides exactly who is over the line.
the handle · The alarm has an address: Platform. Diagnose there first — org-wide programs would waste five-sixths of their effort.
— MEANWHILE —
THE EXONERATION
Pay position (compa-ratio) vs exit-intent score
Is it comp — like everyone assumes?
the scene — pre-built from the deck
It doesn't track — the premise is false
Pay position (compa-ratio) vs exit-intent score: across 18 observations there is no meaningful relationship — r = 0.01. Knowing one measure tells you almost nothing about the other.
the handle · Stop reaching for the comp lever — pay position tells you nothing about who's leaving. The cause lives where the people live: with the work and the manager.
Act III — The Choice
what it's worth · what we do— WHICH MEANS —
THE BET
What is fixing Platform worth — and what do we do?
At 15.1% on 420 people, you lose ~63 per year; at $213K per event (1.5× of $142K), that is ~$13.5M a year.
Each 5 point reduction is worth ~$4.5M per year on your own numbers — which is the budget line any retention intervention should be judged against, and the number the meeting never has when it debates "is turnover a problem."
The recommendation: a Platform-scoped intervention — manager diagnosis first (the exoneration cleared comp), funded against a fraction of the $4.5M/yr prize.
- ⚖ fully-loaded cost per event = 1.5× median salary (the classic range is 0.5×–2.0×; pick yours and defend it)
- ⚖ events scale linearly with the rate; salary mix of leavers ≈ median
- ⚖ value = avoided events × per-event cost; no second-order effects (morale, knowledge) counted — so this is a FLOOR
what next? what else?
That’s the story this data confirms — four beats, every claim evidenced. More certainty is priced, not promised: a manager-level drill inside Platform · a stay-interview instrument on the high performers still here · one more quarter of history on the corrective target. Each step is a known cost against a $4.5M/yr decision.
4 Insight Cards · toolbox player contract v0.4.0
This story is a sequence of canonical Insight Cards — the same shape the ranked player queue (buildPlayerQueue) assembles for feeds. An authored epic keeps its order; the contract is shared.
The driving question: Are we losing the people we can't afford to lose — and why, really?
One story from the database — assembled from the shared deck of scenes. Browse the stories · browse the deck · bring your own data.
The story template behind this epic
The epic above is one playing of this template against designed demo data. The template itself — the complete story, the sequence of slots, the branches it can take — is the reusable record.
the story · before any data
the hero
A leadership team proud of the company people built — wants to keep the people who make it worth joining.
the villain
The quiet exodus — regretted loss that hides inside healthy totals
the problem, felt at three levels
- External · Regretted attrition among high performers is rising past the benchmark while headline turnover looks calm.
- Internal · The creeping fear that the best people know something the dashboard doesn't — and that the board will ask before we have an answer.
- Philosophical · People who give their best shouldn't become visible only when they resign.
the plan
- 1See the alarm and its address (the confirmed cards)
- 2Clear or convict the usual suspects (comp, manager, load)
- 3Fund a scoped intervention against the computed prize
if nothing changes
Do nothing and the curve steepens: each leaver loads the rest, the next resignation letter is from someone you'd counter-offer on the spot, and the story becomes 'why didn't we see this.'
success
The trend bends back under the benchmark and you own the chart that shows it bending.
From a leadership team that learns about its best people from resignation letters — to one that sees it coming and acts first.
the screenplay · the sequence of scenes it needs
Act I — The World
establish trust and stakes: the calm is realTHE REASSURANCE
the baseline calm
Is the workforce stable overall?
fills with
or any card of form: flat
the handle · Say 'stable' out loud — it's why what follows matters.
Act II — The Forces
the alarm, localized and exonerated— BUT —
THE ALARM
the alarm
Is regretted loss moving?
fills with
the handle · Set the corrective target at the line; date it.
— UNDERNEATH —
THE HIDDEN SPLIT
the address
Everywhere — or somewhere?
fills with
+ supporting: You have a two-speed organization
the handle · Go local: the alarm has an address.
— MEANWHILE —
THE EXONERATION
the cleared suspect
Is it comp, like everyone assumes?
fills with
the handle · Stop reaching for the comp lever.
Act III — The Choice
money, the bet, the close— WHICH MEANS —
THE BET
ROI-of-change calculator on stated assumptions
What is fixing it worth — and what do we do?
fills with
a computed step (calculator / decision) — not a single card
the handle · Fund a scoped intervention against a fraction of the prize.
branch points · the story forks on what the data confirms
after “exoneration” — the fork
if comp DOES track exits (the exoneration fails)
→ The Comp Review — the story becomes a compensation story — hand off to The Comp Review
after “split” — the fork
if no unit stands apart (the split fails)
→ The Quiet Exodus — stay org-wide: the DRIFT/culture variant of Act II (engagement-erosion beat)
the resolution · what this story exists to produce
A scoped retention intervention with an owner, a target, and a date — or an explicit decision not to act, on the record.
What next? What else?
More certainty is priced, not promised — the next steps, each a known cost against the decision:
- ↗ manager-level drill inside the hotspot
- ↗ stay-interview instrument on the at-risk cohort
- ↗ one more quarter against the corrective target
what it needs to be told
told differently by altitude
- exec — the three-act arc as rendered
- manager — your team's slice: the split beat + flight-risk 1:1 handles
- analyst — full drill + the cohort decomposition
provenance · Childrens 2013 turnover research (replayed) + the 2026 framework