Scoring & Rating Bands

Scoring & Rating Bands

How a raw number — a 96% attendance rate, a 3.2-hour resolution time — becomes a fair 1.00–5.00 score, and how those scores roll up into one composite and a rating. This is the maths that makes every metric comparable.

The 1.00–5.00 scale & rating bands

Everything the module scores — each metric, each category, the composite — lives on the same continuous 1.00–5.00 scale (stored to two decimals). Five named bands divide it up:

Normalization — raw value → 1–5 score

Each metric defines five threshold bands in its own natural unit (see the catalogue). To score a raw value the engine finds which band it falls in, then interpolates within that band — so scoring is smooth, not a blunt five-step ladder.

1. Match the band

   The raw value is matched against the metric's five threshold ranges, each of which owns a slice of the 1–5 scale (Outstanding → 4.50–5.00, Exceeds → 3.50–4.49, and so on).

2. Interpolate inside the band

   Where the value sits within its band's range determines the exact score. A value at the very top of the Meets range scores near 3.49; near the bottom, near 2.50.

3. Clamp out-of-range values

   A value beyond the metric's authored range lands in the extreme band that owns that end — it can't “overflow” to a perfect 5.00 or an automatic 1.00.

4. Null → excluded

   If the raw value is null (no data, or a divide-by-zero), the metric returns null and is left out of the roll-up — its weight is handed to the metrics that did compute.

// within-band linear interpolation
position = (raw − band_min) / (band_max − band_min)     // ↑ higher-is-better
position = (band_max − raw) / (band_max − band_min)      // ↓ lower-is-better
score    = band_base + position × (band_ceiling − band_base)
score    = clamp(score, 1.00, 5.00)

Direction — which way is “good”

A metric's direction is the only thing that flips the interpolation:

Bell-curve (target-band) metrics

Three metrics — Overtime Utilization, Leave Utilization, Leave Balance — are healthiest at a moderate value, so their bands are arranged with the top score in the middle of the range. Both extremes score low.

Rolling up to the composite

The composite is a two-level weighted average, skipping nulls at both levels.

1. Category score

   Within a category, metric scores are blended by their metric weights — over the non-null metrics only:

   category_score = Σ(metric_score × metric_weight) / Σ(metric_weight)   // non-null metrics

2. Composite score

   Category scores are blended by their category weights — over the categories that produced a score:

   composite_score = Σ(category_score × category_weight) / Σ(category_weight)
   rating_band     = band(composite_score)

Worked example (General Staff)

category_scores:  ATT 4.08 (20%)   TSK 3.75 (20%)   PRJ 3.50 (10%)   TKT 4.20 (10%)
                  BHV 5.00 (10%)   LVE 3.60 (5%)    PRF 3.00 (10%)   TMS 3.80 (15%)

composite = (4.08·20 + 3.75·20 + 3.50·10 + 4.20·10
           + 5.00·10 + 3.60·5  + 3.00·10 + 3.80·15) / 100
          = 388.6 / 100
          = 3.89   →   Exceeds Expectations

Exceeds Expectations — which, for a General-Staff role, would generate a +1 salary step recommendation. See Salary Progression.

Data sufficiency & confidence

A score is only as trustworthy as the data behind it. The engine tracks data sufficiency:

coverage = (metrics_total − metrics_null) / metrics_total × 100

Fair-handling edge cases