03 · Indicator design and measurement
Indicator types and levels
An indicator is a variable that provides a simple, reliable means to measure achievement or change. Indicators sit at different levels of the results chain — input, activity, output, outcome, impact — and come in distinct types whose arithmetic differs. Confusing the levels, or summing values that cannot be summed, produces numbers that look precise and mean nothing.
Last updated · Reviewed against 4 cited sources
What an indicator is
The OECD glossary — the definition of record across development practice — defines an indicator as a “quantitative or qualitative factor or variable that provides a simple and reliable means to measure achievement, to reflect the changes connected to an intervention, or to help assess the performance of a development actor” [1]. Three parts of that sentence do real work. Simple: an indicator reduces a complex reality to something observable and repeatable. Reliable: two people applying the definition to the same reality should produce the same value. Means to measure — not the achievement itself: the indicator points at the result; it is not the result.
Where indicators live — in the cells of a logframe or a results framework — is covered on the Monival platform site’s explainers on the logical framework and the results framework; this page is about the properties of the indicator itself, whatever framework houses it.
The five levels
Results-based management arranges results in a chain, and indicators inherit their level from the result they measure [2, 3]. A single running example — a community health worker (CHW) programme in a Kenyan county — shows how the same programme reads at each level:
- Input — resources mobilised. Value of funding disbursed to the CHW programme (KES).
- Activity (process) — what the programme does. Number of CHW training sessions conducted.
- Output — the direct, countable product of activities, substantially within the programme’s control. Number of CHWs trained and certified to the national curriculum.
- Outcome — the change in behaviour, practice or condition among those the programme reaches. Proportion of children under five with fever assessed by a CHW within 24 hours of onset, in target sub-counties.
- Impact — the higher-order change to which the programme contributes. Under-five mortality rate in the target counties.
The gradient matters for what you may claim. Output values are largely attributable to the programme; outcome values are influenced by it alongside other causes; impact values are shaped by forces far beyond it. Kusek and Rist’s handbook is blunt that a monitoring system reports movement at every level but demonstrates causation at none — that is evaluation’s job, with a design suited to the claim [2].
Types that cut across the levels
Level is one axis; type is another. Any level can carry any type.
Quantitative vs qualitative. The OECD definition explicitly admits both [1]. A qualitative indicator is still an indicator — it needs a definition, a source and a repeatable judgement procedure (a rubric, a scored scale, a documented categorisation), otherwise it is an anecdote with a label.
Direct vs proxy. A direct indicator measures the result itself; a proxy measures something observable that stands in for a result that is unobservable, slow or costly to measure. Proxies carry their own validity obligations — treated fully in Proxy indicators and composite indices.
Arithmetic form. Counts, percentages, ratios and indices behave differently under aggregation. Percentages need their denominator reported alongside them, and cannot be averaged across sites without weighting by that denominator. Ratios invert awkwardly (a fall can mean the numerator fell or the denominator rose). Indices bundle several measurements into one number and inherit every construction choice made along the way [4].
The money-unit trap. Monetary indicators must state currency and unit in the definition — “value of loans disbursed” reported by one office in Kenyan shillings and another in thousands of shillings aggregates into nonsense that no downstream check will catch, because both figures are plausible. The unit of measure belongs in the indicator’s reference sheet, not in the reporting officer’s head.
Cumulative, milestone and snapshot: the reporting-period arithmetic
The most common indicator-table error is not a bad definition but bad arithmetic across periods. Three measurement behaviours must be distinguished at design time [2, 4]:
| Behaviour | Example | Quarterly values | Correct annual figure |
|---|---|---|---|
| Cumulative (flow) | New CHWs certified | 40, 25, 30, 25 | Sum: 120 — if each person is counted once |
| Milestone | Curriculum accredited (yes/no) | No, No, Yes, Yes | Status at year end: Yes |
| Snapshot (stock) | Facilities with tracer drugs in stock (%) | 72, 65, 80, 78 | Latest period (78) or stated average — never the sum |
Two traps follow. First, double counting in cumulative indicators: a person trained in Q1 and again in Q3 is one person trained, unless the indicator is explicitly “training places delivered”. The definition must say which. Second, summed snapshots: adding a stock percentage across four quarters produces a number over 100 that reporting software will happily accept. The aggregation rule is a property of the indicator, and belongs in its reference sheet.
Indicator, target, result: the grammar
A results statement has three grammatical parts, and conflating them corrupts all three [2, 3]:
- The indicator names what is measured and how: “Proportion of children under five with fever assessed by a CHW within 24 hours.” It contains no desired value and no direction of change.
- The target names the value sought by a date: “70% by December 2027, from a baseline of 41%.” Targets are treated fully in Baselines, targets and disaggregation.
- The result is the value observed: “63% at December 2027.”
Writing the target into the indicator (“increase assessment coverage to 70%”) is the commonest violation, and it has a practical cost: when the target is revised, the indicator’s identity changes with it, breaking the time series and the audit trail. Keep the measurement stable and let the target move around it.
Choosing the level to measure at
A workable indicator set spans levels deliberately rather than piling up where measurement is easiest. Output indicators are cheap and fast; a set made only of them tells you the machine is running, not that it is producing change. Outcome indicators are where most programme accountability properly sits. Impact indicators usually come from national systems (surveys, vital statistics) rather than programme measurement, and should be labelled as context the programme contributes to rather than results it claims [2, 3]. The quality tests that then apply to each candidate indicator — SMART, CREAM, SPICED and their competing logics — are the subject of the next page.
Sources
- Glossary of Key Terms in Evaluation and Results-Based Management for Sustainable Development (Second Edition) — OECD Publishing, 2023.The definition of record for indicator, output, outcome and impact in development evaluation.
- Ten Steps to a Results-Based Monitoring and Evaluation System: A Handbook for Development Practitioners — The World Bank, 2004.Kusek & Rist. Steps 3–5 treat indicator selection, baselines and targets as the load-bearing middle of an M&E system.
- Results-Based Management Handbook: Harmonizing RBM Concepts and Approaches for Improved Development Results at Country Level — United Nations Development Group (UNDG), 2011.The UN system's harmonised treatment of results levels and the grammar of results statements.
- DIME Wiki: practical impact-evaluation and measurement resource — World Bank Development Impact (DIME), n.d..Continuously updated; practical entries on indicator construction and measurement in field research.