The measurement system
Twenty-five facts roll into seven abilities, which roll into five HR scores. Every weight on this page is the weight the product uses. If a number ever surprises you, you can walk down the layers until you reach the single answer that moved it.
The design constraint was blunt: an HR director has to be able to defend any figure in this system to the person it describes. That rules out proprietary composites, and it is why the partial-data rule below is as conservative as it is.
Layer 4 · roadmap
Business linkage
Attrition against Turnover Risk, delivery pace against Team Dynamics.
Layer 3
Five HR scores
The five questions an HR director has to answer this quarter.
Layer 2
Seven abilities
Weighted clusters. Where growth and gaps become legible.
Layer 1
Twenty-five facts
Raw metrics straight out of the games. Immutable once written.
Click a fact, an ability or a score. The map lights everything it feeds and everything that feeds it — with the actual weights, from the same dataset the configurator runs on.
The map scrolls sideways on a narrow screen →
Twenty-five metrics, twelve games. Written once, never edited.
Not what the player said about themselves — what they did. Twenty-two of the twenty-five are deterministic formulas over recorded choices: same input, same output, every time. The other three read free text with a model.
22
Deterministic
A rule over recorded choices. Costs nothing to compute and never varies.
3
AI-scored
Free text read by a model. If it is unavailable, the cluster recalculates without it rather than blocking the session.
9
Inverted
A high raw value is the bad news. Polarity is declared, never inferred.
| # | Metric | Game | Week | Type | Ability |
|---|---|---|---|---|---|
| 1 | Decision Accuracy | Decision Sprint | W1 | Direct | Decision-Making |
| 2 | Pressure Delta | Decision Sprint | W1 | Direct | Decision-Making |
| 3 | Reasoning Quality | Decision Sprint | W1 | AI | Decision-Making |
| 4 | Signal Detection | Signal Hunter | W1 | Direct | Attention |
| 5 | Noise Resistance | Signal Hunter | W1 | Inverted | Attention |
| 6 | False Urgency Rate | Signal Hunter | W1 | Inverted | Attention |
| 7 | Deep Work Placement | Time Architect | W1 | Direct | Energy |
| 8 | Cognitive Endurance | Time Architect | W1 | Inverted | Energy |
| 9 | Fragmentation Score | Time Architect | W1 | Inverted | Attention |
| 10 | Locus of Control | Control Circle | W2 | Inverted | Stress Response |
| 11 | Escalation Judgment | Red Flag Filter | W2 | Direct | Stress Response |
| 12 | Sustainability Index | Recovery Architect | W2 | Direct | Energy |
| 13 | SBI Feedback Quality | SBI Feedback Sprint | W3 | AI | Communication |
| 14 | Feedback Toxicity | SBI Feedback Sprint | W3 | Inverted | Communication |
| 15 | Recognition Specificity | SBI Feedback Sprint | W3 | AI | Communication |
| 16 | Blame Avoidance | Difficult Conversation 360 | W3 | Inverted | Safety & Voice |
| 17 | Conflict Navigation | Difficult Conversation 360 | W3 | Direct | Communication |
| 18 | Voice Behaviour | Can You Say This Out Loud? | W3 | Direct | Safety & Voice |
| 19 | Psych. Safety Behaviour | Is It Safe to Say? | W4 | Direct | Safety & Voice |
| 20 | System Thinking | Go Look | W4 | Direct | Systems & Leadership |
| 21 | Blame Avoidance (team) | Go Look | W4 | Inverted | Safety & Voice |
| 22 | Focus Stability | Leadership Week | W4 | Direct | Systems & Leadership |
| 23 | Stress Recovery Rate | Leadership Week | W4 | Direct | Stress Response |
| 24 | Comm. Effectiveness | Leadership Week | W4 | Direct | Communication |
| 25 | Delegation Quality | Leadership Week | W4 | Direct | Systems & Leadership |
Each measurement is stored as an immutable snapshot carrying its raw value, its normalised value, the session and cohort it came from, and the version of the formula that produced it. Trend, personal best and personal average are derived from that history rather than stored.
A fact is a number. An ability is something you can say out loud to the person it describes. Each one is a weighted sum of its facts on a 0–100 scale where higher is always better.
Ability = Σ (metric.normalised × metric.weight)
| Ability | Composed of | Weight |
|---|---|---|
| Decision-Making | Decision Accuracy | 0.45 |
| Pressure Delta | 0.30 | |
| Reasoning Quality | 0.25 | |
| Attention | Signal Detection | 0.35 |
| Noise Resistance | 0.25 | |
| False Urgency Rate | 0.20 | |
| Fragmentation Score | 0.20 | |
| Energy | Deep Work Placement | 0.35 |
| Cognitive Endurance | 0.35 | |
| Sustainability Index | 0.30 | |
| Stress Response | Locus of Control | 0.35 |
| Escalation Judgment | 0.35 | |
| Stress Recovery Rate | 0.30 | |
| Communication | SBI Feedback Quality | 0.25 |
| Conflict Navigation | 0.25 | |
| Feedback Toxicity | 0.20 | |
| Recognition Specificity | 0.15 | |
| Communication Effectiveness | 0.15 | |
| Safety & Voice | Voice Behaviour | 0.30 |
| Blame Avoidance | 0.25 | |
| Psychological Safety Behaviour | 0.25 | |
| Blame Avoidance (team) | 0.20 | |
| Systems & Leadership | System Thinking | 0.40 |
| Focus Stability | 0.30 | |
| Delegation Quality | 0.30 |
The most important rule on this page, and the least glamorous.
Games are spread across four weeks, so after week one only Decision-Making, Attention and Energy have any data at all. A system that quietly filled the gap with a neutral 50 would produce a plausible number and a wrong decision. This one redistributes instead — and says so on the face of the figure.
Worked example · Burnout Risk after week one
Full formula
Burnout = (100 − Energy) × 0.35 + (100 − Stress Response) × 0.35 + (100 − Attention) × 0.30
Stress Response has no data yet, so:
available weight = 0.35 + 0.30 = 0.65
Burnout = (100 − Energy) × (0.35 / 0.65) + (100 − Attention) × (0.30 / 0.65)
→ reported as 52~, not as 52
| Inputs available | What the platform does | How it is shown |
|---|---|---|
| 0 of N | Nothing is computed | — (dash) |
| 1 of N | Computed but withheld — one input is not a score | — (dash) |
| 2 or more | Weights redistributed proportionally | 52~ (tilde) |
| All of N | Full formula | 52 |
Two consequences worth stating plainly. A preliminary score is for direction, never for a personnel decision. And a package that omits a whole week will leave certain scores permanently preliminary — which is not a limitation to discover in month three. See it for your own package →
Five questions, five numbers. Each one maps to a process you already run.
Who is ready for the next step? → Talent bench
Systems & Leadership × 0.35 + Decision-Making × 0.25 + Communication × 0.20 + Stress Response × 0.20
Weighted toward systemic and interpersonal capability rather than raw output, because output is what a promotion stops predicting.
Who is running out of road? → Wellbeing
Higher = worse(100 − Energy) × 0.35 + (100 − Stress Response) × 0.35 + (100 − Attention) × 0.30
At or above 70 this raises a named alert to a single HR contact. It is one of only two places in the entire product where an individual is named.
Who might leave? → Retention
Higher = worse(100 − Safety & Voice) × 0.40 + Burnout Risk × 0.35 + (100 − Communication) × 0.25
Kept separate from burnout on purpose. Someone can be exhausted and stay out of loyalty; someone else can be perfectly rested and leave because of the environment. Environment carries the heavier weight.
Where is the friction? → Team interventions
Safety & Voice × 0.40 + Communication × 0.35 + Stress Response × 0.25
Reported at team level as a median plus a spread plus named outlier positions — never as a bare average, because the average is exactly what hides the finding.
Where does the L&D budget go? → L&D planning
Needs a role profilemax_gap( Ability[i] , RoleProfile[i] ) across all seven
Not a level but a direction. It requires a filled role profile and all seven abilities — a largest gap computed over three of them is a different answer, not a rougher one.
Strong position. On the two risk scores, this band means low risk.
Needs attention. Read the direction of travel before the level.
Critical gap, or high risk on the two inverted scores.
This is the real Burnout Risk arithmetic, played by hand. Drag the three abilities and watch the score move — then simulate week one, when Stress Response has no data yet.
Every figure is shown against three references at once, plus a trend. The threshold for calling a change real is ±3 — below that the arrow stays flat.
Their own snapshot history
Better or worse than before. The only benchmark shown to the player first, and the only one that motivates without inviting a ranking culture.
Team median plus spread
Above or below colleagues. Useful, and dangerous alone — a whole team can be short on the same thing and all look fine against each other.
Role profile, set by you
The only benchmark that supports a promotion conversation, because it is the only one that encodes what the next job actually requires.
How it reads in the dashboard
Fifteen minutes, the real product, your scenario.