Start with a pilot, not a promise
The pilot programme.
Ninety days for workplaces, one ward over six months for the NHS, one semester for universities. Four KPIs agreed before we start, an A/B design built in, and the pilot fee credited against your first year if you continue. You risk a conversation, not a contract.
How the ninety days run
Plug in
Twenty minutes of onboarding for the operator, menus imported from a spreadsheet, health profiles set once (under three minutes each). Intervention and control groups assigned automatically.
Live service
People order as they always have. The AI Nudge Engine re-ranks each day's menu for the intervention group; the control group sees the standard menu. Every order is scored.
The read-out
A board-ready report: intervention against control on the four KPIs you chose, with significance testing. Then a continuation decision — yours, not ours.
What a successful pilot is measured on
These are the agreed targets a pilot is judged against — benchmarks modelled from peer-reviewed research. No pilot has completed yet, and we will never present a target as a result.
| KPI | Target (agreed up front) | Where it comes from |
|---|---|---|
| Active adoption | 60–80% of eligible staff ordering | Modelled vs the 20–40% ceiling of standalone apps |
| Diet quality (HEI-2020) | +3–5 points intervention vs control | Modelled from BIT/Nesta nudge trials (6–15% calorie reduction, p<0.001) |
| Healthy-meal selection | +20–26% | Modelled from canteen nudge literature |
| Canteen operations | −14–17% food waste · footfall up | Modelled from pre-ordering and demand-prediction studies |
The pilot OKR
Prove that the canteen you already run can be a measurable wellbeing asset.
Adoption target met — the platform reaches the people wellness apps never touched, including shift and night staff.
Diet-quality delta significant — intervention beats control at the 95% level on HEI-2020.
Board-ready evidence delivered — anonymised, exportable, procurement-compatible reporting your CFO accepts.
A clear continuation decision — the data makes year one an easy yes or an honest no.