# Impact Evaluation Indicator Matrix — illustrative template

**Illustrative structure. No client information. Actual scope and evidence requirements vary by programme.**

Published by Gurjas Evidence and Policy Analytics · gurjas.org
Template version 1.0 · 21 July 2026

---

## What this document is for

An indicator matrix forces an explicit decision, before fieldwork or
data collection begins, about what will actually be measured, how, and
what level of inference the design can support. Programmes commonly slip
into overclaiming not because anyone intends to mislead, but because the
indicator, its data source and the honest limitation of that source were
never written down together in one place.

## Before completing this matrix

Decide and record, separately from the table below:

- **Programme decision this evaluation informs** — what will actually change
  based on the result, and who makes that decision.
- **Feasible level of inference** — whether the design supports causal
  attribution, credible contribution analysis, or description only. State
  this honestly before selecting indicators; it constrains which indicators
  are worth collecting.
- **Baseline availability** — whether a genuine pre-programme baseline
  exists. If it does not, no indicator matrix can retrofit one; say so.

## Template

| Indicator | Definition (precise, single measure) | Data source | Baseline value | Target / comparison | Collection method | Frequency | Responsible party | Known limitations |
|---|---|---|---|---|---|---|---|---|
| | | | | | | | | |
| | | | | | | | | |
| | | | | | | | | |

**Column notes**

- **Definition** — write it so precisely that two different data collectors
  would record the same value for the same underlying fact. A vague
  definition is the single most common cause of indicator matrices that
  cannot be reproduced.
- **Data source** — administrative record, survey instrument, qualitative
  method or secondary dataset, named specifically.
- **Known limitations** — coverage gaps, self-report bias, attrition,
  seasonality, or anything that would make a reader over-trust the number
  without this note attached.

## Contribution versus attribution — state this explicitly

For each indicator, note whether the design can support a claim that the
programme *caused* the observed change (attribution — requires a credible
counterfactual) or only that the programme is a *plausible contributor*
among other factors (contribution — the more common, and often more
honest, claim for programme evaluations without a control group). Do not
let report language claim more than the weakest indicator in the matrix
can support.

---

*This is a structural template only. It contains no indicators, values or
information from any Gurjas engagement. Gurjas Evidence and Policy
Analytics · gurjas.org*