Home
AI AgentsAI Strategy & RoadmapData & AnalyticsInsights & DiagnosticsAI AcademyAnalytics HubICT Solutions All solutions
Baseline EvaluationsMidline EvaluationsEndline EvaluationsProcess EvaluationsNeeds AssessmentsImpact Assessments Overview
About DatalystPricingPartnershipContact Book a meeting
Start free trial Talk to Sales Sign in

info@datalystafrica.com
+263 773 806 868

Impact Assessments

Did it change because of you — or would it have changed anyway?

Impact assessment isolates the change your intervention actually caused, using a credible counterfactual, and measures whether it lasted after the funding stopped.

Request a proposal Book a scoping call
10–16wk
Typical timeline
1–5yr
After completion
2
Groups compared
4
Attribution methods
When to commission it

Attribution is the hardest claim in evaluation, and the most valuable

Incomes rose 18%. Excellent — but the harvest was also good, a road was tarred and a second organisation ran a similar programme nearby. How much of that 18% was you?

Impact assessment answers that with a counterfactual: what would have happened without the intervention. Depending on what the design allows, we use comparison groups, difference-in-differences, propensity score matching or rigorous contribution analysis — and we state honestly how strong the resulting claim is.

Request a proposal Book a scoping call
Impact Assessments 10–16 weeks
THE QUESTION IT ANSWERS

“What changed because of us, and not for other reasons?”

CORE OUTPUTS
  • Attributable impact with a credible counterfactual
  • Intended and unintended effects
  • Cost-effectiveness per outcome achieved
  • Sustainability of change over time
STAGE
1–5 years after
TYPICAL COST
US$22,000–65,000
Method

How we run it

1

Design the counterfactual

Establish the strongest available comparison given the programme design and existing data.

2

Collect from both groups

Survey participants and comparison population using instruments comparable to baseline.

3

Estimate the effect

Statistical estimation with controls, robustness checks and confidence intervals reported.

4

Assess sustainability

Whether change persisted after support ended, and what maintains or erodes it.

Deliverables

What you receive

Attributable impact estimate

The measured effect with confidence intervals and a plain statement of attribution strength.

Counterfactual analysis

The comparison basis, why it is credible, and what threatens its validity.

Unintended consequences

Effects nobody planned for — positive and negative, on participants and non-participants.

Cost-effectiveness

Cost per unit of outcome achieved, comparable to alternative interventions where benchmarks exist.

Sustainability findings

Whether the change held after funding ended, and the mechanisms keeping it in place.

Strategic implications

What to scale, what to redesign and what not to repeat, with the reasoning shown.

Methods

How we establish the counterfactual

Comparison group design

A similar population that did not receive the intervention, matched on observable characteristics.

Difference-in-differences

Change in the treated group compared to change in the comparison group over the same period.

Propensity score matching

Statistical matching where participants self-selected rather than being randomly assigned.

Contribution analysis

Where no counterfactual is possible, a rigorous causal narrative testing rival explanations.

Request a proposal
Other evaluation types

Often commissioned alongside this

Baseline Evaluations

Establish a defensible benchmark before implementation

Open

Midline Evaluations

Course-correct while there is still time to act

Open

Endline Evaluations

Prove outcomes, effectiveness and value for money

Open

Process Evaluations

Diagnose delivery quality and operational efficiency

Open

Needs Assessments

Locate the real gaps before you commit resources

Open
FAQ

Questions we get asked

Yes, though the claim is weaker and we will say so. Propensity score matching against a comparable population, or structured contribution analysis, can still produce credible findings. What we will not do is present correlation as proof of causation.
For behavioural or economic outcomes, one to three years is usually the informative window — long enough for effects to consolidate, recent enough for reliable recall and traceable participants. Institutional or generational change may need five.
No. An endline asks whether targets were met at closure. An impact assessment asks what portion of the observed change your intervention caused, and whether it lasted — which requires a counterfactual and typically happens years later.
Usually a substantial majority, provided baseline records included contact details. We plan for attrition, report the rate transparently and test whether those we lost differ systematically from those we found.
No, and any evaluator promising otherwise should be avoided. We have reported null and mixed results. A credible negative finding that redirects future funding is worth considerably more than a flattering one nobody believes.
Get started

Send us the terms of reference.

You will get a method, a named team, a timeline and a fixed price — and an honest opinion on whether this is the right evaluation for your question.

Request a proposal Book a scoping call

No obligation · Response within one business day · NDA on request

Chat with us