Topic 30 of 64· Governmental performance and delivery metrics

Citizen Satisfaction Metrics

Citizen satisfaction metrics measure how people rate their direct experience of a public service — distinct from trust in institutions generally, and distinct from whether the service actually achieved a good outcome. A service can be well-liked and ineffective, or effective and disliked; the gap between the two is itself diagnostic information a delivery team should be watching.

Why it matters

Satisfaction is measured at two different altitudes that get routinely conflated. At service level, the UK's now-retired Performance Platform and today's GOV.UK service manual require a per-service satisfaction survey (typically a five-point "very satisfied" to "very dissatisfied" scale, administered at the point of transaction) as one of four mandatory service KPIs — see service standards and transaction metrics. At institutional level, the UK Civil Service People Survey measures employee engagement and experience across every central government department annually, and separately the OECD's "Trust in Government" program surveys public trust in national government across member states, tracking a long-run decline and recovery pattern shaped heavily by crises (the 2008 financial crisis and the COVID-19 pandemic both produced sharp, visible movements in OECD trust figures). The reason engineers building citizen-facing services need to hold satisfaction and outcome apart is a known failure mode in service design: a beautifully designed, easy-to-use digital form for a benefit claim can score very high satisfaction while the underlying policy — eligibility rules, processing backlogs, award amounts — leaves the claimant no better off. Satisfaction measures the interface; it does not measure the value delivered behind it.

The math

Net satisfaction = % satisfied (or very satisfied) − % dissatisfied (or very dissatisfied)
                    (neutral/no-opinion responses excluded from both terms, but counted
                    in the response base for calculating each percentage)

Satisfaction-to-outcome gap = satisfaction score − outcome achievement score
                    (both normalized 0–100; a large positive gap signals a service
                    that "feels good" but under-delivers on substance)

Trust index (OECD-style) = % of survey respondents answering "yes" to
                    "do you have confidence in [national government]?"
                    tracked as a time series, typically disaggregated by
                    age, income, and education

Worked example

Local authority council tax e-billing service: a satisfaction survey at the point of successful transaction shows 2,400 respondents: 1,650 satisfied/very satisfied, 250 dissatisfied/very dissatisfied, 500 neutral.

Net satisfaction = (1,650/2,400 × 100) − (250/2,400 × 100)
                  = 68.75% − 10.42%
                  = +58.3 net satisfaction

This looks strong in isolation. But the survey is only shown to users who successfully complete the transaction — a known measurement bias (see pitfalls below). Pairing it with the completion rate metric from service standards and transaction metrics shows completion is only 71%, meaning:

True population satisfaction is unmeasured for the 29% who abandoned the journey —
plausibly the most dissatisfied cohort, since abandonment is itself a strong negative
signal the survey never captures.

National-level illustration (structure of an OECD-style trust series): national government trust reported at 42% in year 1, falling to 34% in year 2 (a crisis year) and recovering to 39% in year 3 — a trajectory typical of the shock-and-partial-recovery pattern the OECD documents across member states following major crises.

Software engineering connection

Instrument satisfaction surveys at every meaningful exit point of a user journey, not only at successful completion — the single most common engineering mistake in this space, and one that silently converts a satisfaction metric into a survivorship-biased vanity metric. Where possible, pair the satisfaction score with a completion or outcome metric on the same dashboard so a team cannot celebrate rising satisfaction while completion quietly falls (see cost per transaction and digital inclusion for who gets excluded from digital satisfaction sampling in the first place — non-digital and assisted-digital users are systematically under-represented in in-service surveys). Satisfaction and trust data also feed directly into the legitimacy leg of Moore's strategic triangle, and belong on the "customer" and "legitimacy" perspectives of a public value scorecard — see trust and legitimacy metrics for the institutional-level counterpart to this service-level metric.

Pitfalls

  • Survivorship bias in point-of-completion surveys: users who abandon a journey never see the survey, so a high in-service satisfaction score can coexist with a low completion rate and a large invisible population of dissatisfied non-completers.
  • Treating satisfaction as a proxy for outcome: a well-designed interface for a poorly designed policy scores well on satisfaction and badly on outcome — always report both, never one as a stand-in for the other.
  • Small, unrepresentative samples reported with false precision: a satisfaction score from a few hundred self-selected respondents reported to one decimal place implies a confidence the sample size cannot support.
  • Ignoring demographic disaggregation: national trust and satisfaction figures that aren't broken down by age, income, disability, or digital access can mask sharply divergent experiences across groups — a pattern the OECD's own Trust in Government releases explicitly disaggregate for.

Sources