The statistics behind the numbers

1. Control-group design

Effects are measured with a quasi-experimental design where the development in the affected teams is compared with comparable teams without the initiative. The control group shows what would have happened anyway, and the difference between the two developments is the effect of the initiative. We follow the methods literature that defines good practice for this design family, including its assumptions and pitfalls.

Dimick & Ryan (2014) · Wing, Simon & Bello-Gómez (2018) · Roth, Sant'Anna & Bilinski (2023)

2. Uncertainty and confidence

Every result is reported as an estimate with a 95% interval rather than a single number, and is classified into four confidence tiers, where "not enough data" is a possible answer. The approach follows the estimation literature: effect sizes and intervals over bare significance tests. A measured result is an association from a comparison, not a controlled trial, and is never presented as more.

Cumming (2013)

3. Benchmarks

Benchmarks build on published national statistics and industry analyses and are matched as closely as possible to the organisation's profile. Where a sufficiently specific breakdown does not exist, the closest available level is used, and the line says so. For competitive reasons we do not publish the exact composition of the base here; we are happy to present it in confidence.

4. From data to money

Absence and turnover are turned into money with a fixed model with conservative assumptions and discounting of future savings. The model and its assumptions are always shown next to the number in the product, line by line, so every figure can be recomputed there. The cost of turnover is well documented in the literature, both per departure and at organisational level.

Hinkin & Tracey (2000) · Park & Shaw (2013)

5. Presenteeism: always modelled, never measured

The cost of working while unwell only enters as a marked estimate based on multipliers from the research literature and never enters a documented saving. The literature consistently finds presenteeism costs at or above the direct costs of absence.

Goetzel et al. (2004) · Schultz & Edington (2007) · Kigozi et al. (2017)

6. Data-driven HR as a field

The approach follows the HR-analytics literature: quantitative evaluation as an integrated part of HR work, tied directly to business decisions rather than to reporting alone.

Marler & Boudreau (2016) · Rasmussen & Ulrich (2015)

7. Anonymity

Everything is computed at team level. Groups below the anonymity threshold are not shown, and individual scores do not exist in the product.