Topic 27 of 64· Governmental performance and delivery metrics

Outcomes-Based Accountability (OBA)

Outcomes-Based Accountability, also called Results-Based Accountability (RBA), is Mark Friedman's framework for separating two questions that public-sector reporting habitually blurs together: "is the population doing well?" (population accountability) and "is this specific program doing well?" (performance accountability). Conflating the two is, in Friedman's telling, the single most common reason well-run programs get blamed for population trends they never had the power to move.

Why it matters

Friedman set the framework out in Trying Hard Is Not Good Enough (2005), arguing that most public reporting either drowns decision-makers in population-level statistics no single agency controls (teen pregnancy rate, unemployment rate, life expectancy) or drowns them in program-level activity counts (clients seen, referrals made) that say nothing about whether anyone's life improved. RBA's contribution is a small, disciplined vocabulary that keeps the two apart: population results (conditions of wellbeing for a whole population, like "children are born healthy") belong to no single agency and require many partners moving together; performance measures (how well a specific program serves its specific clients) belong to one agency and should be judged only against what that agency can actually influence. Friedman's "three performance questions" — how much did we do, how well did we do it, and is anyone better off? — is now embedded across US state and county human-services contracting and, via the RBA-aligned consultancy and toolkit Clear Impact, widely used in UK and Commonwealth local-government commissioning. The practical stakes are contractual: a housing program should not be defunded because the city's homelessness rate rose from macroeconomic causes outside its reach, but it absolutely should be defunded if its own clients aren't being housed.

The math

Population accountability (the "big picture" a community, region, or nation shares):
  Result       — a condition of wellbeing (e.g. "residents are economically secure")
  Indicator(s) — a measure of that condition (e.g. unemployment rate, median household income)
  → no single program owns the indicator; movement requires many contributors

Performance accountability (what one program is responsible for):
  How much did we do?      — activity volume (clients served, units delivered)
  How well did we do it?   — quality/efficiency (% completing program, cost per client)
  Is anyone better off?    — the outcome that matters (% in employment 6 months after
                              program, before/after or against a comparison group)

A program is judged on the third performance question, never on the population
indicator directly, unless its scale and design could plausibly move it alone.

Worked example

City-funded employment support program, 500 participants/year, contracted by a local authority under an RBA-style performance framework:

Population indicator (context, not the program's scorecard):
  City unemployment rate: 6.2% (up from 5.8% the prior year, driven by a factory closure
  outside the program's control)

Performance measures (the program's actual accountability):
  How much:   500 participants enrolled (target 480) — met
  How well:   78% completion rate; cost per completer = £340,000 / 390 completers ≈ £872
  Better off: of 390 completers, 260 in sustained employment at 6 months = 66.7%
              versus a matched comparison group's 41% (see counterfactual-analysis)

Under a population-accountability read, the program looks like it is failing — the city's unemployment rate went up on its watch. Under RBA's performance-accountability read, the program is succeeding: it hit its volume target, held quality steady, and produced an employment outcome 25.7 percentage points above a matched comparison group, while the population indicator moved for reasons (a factory closure) entirely outside the program's control.

Software engineering connection

RBA maps directly onto a familiar SRE distinction: population indicators are like business-level North Star metrics that no single engineering team owns end to end (company revenue, market share), while performance measures are like a team's own SLOs — the things that team's design decisions actually move. A dashboard that reports both without labeling which is which invites exactly the misattribution RBA was built to prevent: an on-call engineer getting blamed for a metric a dependency team controls. When commissioning or building reporting tools for outcomes contracts, build the "how much / how well / better off" triad as first-class, separately-filterable fields rather than a single blended KPI — it is the same discipline as separating leading and lagging indicators in public sector KPIs. RBA is also the accountability logic underneath payment by results and social impact bonds: a PbR contract can only fairly pay on the "better off" performance measure, never on the population indicator, unless the intervention is genuinely the dominant driver of it.

Pitfalls

  • Paying or penalizing a program against a population indicator it cannot control: this is the single mistake RBA exists to prevent; always trace whether the program is a major or minor contributor to the population result before attaching consequences to it.
  • Reporting "how much" as if it were "better off": activity counts (clients seen) are the easiest data to collect and the least informative; insist the "is anyone better off" question is answered with real outcome data, ideally against a counterfactual (see counterfactual-analysis).
  • Treating RBA indicators as fixed forever: Friedman's method is explicitly iterative — a "data, story, what works, action plan" cycle — not a one-off scorecard design exercise.
  • No comparison group for "better off": a before/after change with no counterfactual conflates program effect with the trend the population would have shown anyway.

Sources