Health Economics Metrics

Topic 60 of 76 · Software engineering and digital delivery

GDS Service Metrics

The UK Government Digital Service (GDS) Service Manual mandates four KPIs for every government digital service: cost per transaction, user satisfaction, completion rate, and digital take-up. Together they are the minimal economics of a public digital service — and the template NHS digital services inherit.

Why it matters

The GDS metrics encode the channel-shift business case that funded a decade of government digitization: the Digital Efficiency Report found digital transactions ~20× cheaper than phone and ~50× cheaper than face-to-face (local-gov figures: web £0.15, phone £2.83, face-to-face £8.62). But the savings only materialize when people complete the digital journey (completion rate) instead of the expensive channel (take-up) — the four KPIs are one economic model, not four dashboards.

The math

Cost per transaction = total service cost / completed transactions
Completion rate      = completed / started transactions × 100
Digital take-up      = digital transactions / all-channel transactions × 100
User satisfaction    = % satisfied+very satisfied (5-point, in-service survey)

Channel-shift saving = volume × take-up shift × (cost_old_channel − cost_digital)
… minus failure demand: (1 − completion rate) × fallback channel cost

Worked example

An NHS appointment-management service: 2M transactions/year, currently 70% phone (£3.20/call) / 30% digital (£0.25). A redesign lifts digital take-up to 55% and completion from 84% to 93%:

Take-up shift saving = 2M × 0.25 × (3.20 − 0.25) = £1,475,000/year

Failure-demand saving: failed digital journeys fall back to phone
  before: 2M × 0.30 × 0.16 × £3.20 = £307,200
  after:  2M × 0.55 × 0.07 × £3.20 = £246,400
  net £60,800/year — completion improvements protect the take-up gains

Satisfaction is the leading indicator: dissatisfied users revert to phone,
so a satisfaction drop forecasts take-up decay before it appears.

Software engineering connection

These four KPIs are a production-grade example of a cost-consequence table: one cost metric, three outcome metrics, never collapsed into a score. For product engineers the operational lessons: completion rate is a funnel-instrumentation problem (every abandonment point is findable and fixable); cost per transaction is cloud unit economics plus staff-assisted-channel costs; take-up is an equity metric in disguise — the users who can't or won't shift channels are disproportionately elderly, disabled, and deprived, so aggressive channel closure converts "savings" into access harm (see reach and equity). Publishing the KPIs (GOV.UK does, per service) is itself a mechanism: transparency disciplines forecasts the way benefits realization audits do.

Pitfalls

  • Take-up by coercion: closing the phone line lifts take-up and dumps failure demand on front-line staff; measure total-system cost.
  • Completion measured from page-2: starting the funnel after the drop-off point flatters the rate.
  • Per-transaction cost ignoring assisted-digital support and failure-demand handling.
  • Satisfaction surveys only at successful completion — the dissatisfied mostly never reach the survey.

Sources