Definition

Outcome harvesting is an evaluation approach that collects — harvests — evidence of what has changed, and then works backwards to determine whether and how an intervention contributed to those changes.

It inverts the normal direction of monitoring. Conventional monitoring specifies results in advance and then measures whether they occurred. Outcome harvesting does not specify results in advance at all. It identifies changes that actually happened in the behaviour, relationships, actions, policies or practices of social actors — individuals, groups, organisations, institutions — and then asks, for each one, whether the intervention plausibly contributed.

The unit of analysis is the outcome, defined for this method as an observable change in a social actor, influenced by the intervention. Not a change in the intervention’s own outputs, and not an aggregate statistic — a specific, verifiable change in what someone or some institution does.

Origin

The approach was developed by Ricardo Wilson-Grau and Heather Britt and set out in their 2012 guide, written for the Ford Foundation’s Middle East and North Africa office. It emerged from evaluation practice with advocacy, networks and human rights organisations, where the standard question — did you achieve your pre-specified targets — was producing evaluations that missed almost everything the organisations had actually accomplished.

The method

Outcome harvesting is usually presented as six steps:

1. Design the harvest. Agree the useable questions with the primary users, the scope, the period, and who the informants will be. The harvest is designed around what someone will do with the findings.
2. Review documentation and draft outcome descriptions. The harvester reads reports, minutes, correspondence and media, and drafts candidate outcomes.
3. Engage with informants. Work with the change agents — staff, partners, participants — to refine, correct and add to the drafted outcomes. This is iterative and is where most of the outcomes are actually formulated.
4. Substantiate. Selected outcomes are verified with independent sources who are knowledgeable but not part of the intervention. Substantiation is what separates outcome harvesting from a collection of self-reported success stories, and skipping it is the single most damaging shortcut available.
5. Analyse and interpret. Classify, pattern, and answer the harvest questions. Contribution is assessed across the body of outcomes, not claimed for each one individually.
6. Support use of the findings. Deliberately built into the method, on the reasoning that a harvest that is not used was not worth commissioning.

The outcome statement

Each outcome is captured as a short structured record — conventionally rendered as six fields, though the labels vary between practitioners. Three of them carry the substance:

• Outcome description. Who changed, what changed, when and where. Written specifically enough that an independent person could check it. “Civil society engagement increased” is not an outcome description; “in March, the county assembly’s budget committee adopted the public participation schedule drafted by the coalition, and published it on the county website” is.
• Significance. Why this change matters — in relation to the harvest questions, the context, and the larger change the intervention is pursuing. This field is where the analytical work happens and is the one most often left thin.
• Contribution. What the intervention did that plausibly influenced the change, stated as contribution rather than attribution, and honest about the other actors involved.

The remaining fields are record-keeping that makes the harvest usable: an identifier, the date or period of the change, and the source of evidence — including the substantiation source where the outcome was verified.

When it fits — and when it does not

Outcome harvesting is the right instrument when results cannot honestly be specified in advance:

• Advocacy and policy influence, where the win depends on a legislative window nobody controls.
• Systems change and network strengthening, where the intended effect is that other actors start behaving differently.
• Complex, adaptive programmes that deliberately change course in response to what they learn.
• Retrospective evaluation of work that was never framed in results terms in the first place.

It is the wrong instrument for service delivery with countable outputs. If your programme vaccinates children, builds latrines or trains health workers, you should be setting targets and measuring against them. Harvesting there would be an expensive way to discover what a logframe would have told you for free.

It is also comparatively expensive. Substantiation in particular takes independent people’s time, and harvests that cut it to save budget produce something that looks like outcome harvesting and carries none of its credibility.

Artefacts it produces

• A harvest design document with the usable questions.
• The outcome database — the set of outcome statements, which is the primary artefact and is directly analysable.
• Substantiation records naming who verified which outcomes.
• The analysis answering the harvest questions, usually with outcomes classified by type, actor, or position on the change pathway.

How it relates to the other frameworks

• It is the natural counterpart to a theory of change. A common and strong pairing is to use the theory of change to frame what change is being sought, then harvest to find out what actually changed — including outcomes the pathway never anticipated, which is often the most valuable part of the finding.
• It is the alternative to a logframe when pre-specification is dishonest rather than merely difficult. The two are not compatible in the same reporting line: a funder cannot hold you to targets and simultaneously accept that results could not be named in advance. That conversation has to be had at contracting.
• Harvested outcomes feed an evaluation against the OECD-DAC criteria, particularly effectiveness and impact — the impact criterion’s interest in unintended effects is served by harvesting almost by construction.
• It sits uneasily with value for money assessment, because outcomes harvested retrospectively resist unit costing. Where both are required, the VfM assessment usually has to work at a different level of the portfolio.

Common mistakes

1. Skipping substantiation. Without independent verification a harvest is a collection of things staff believe they achieved. This is the mistake that destroys the method’s credibility fastest.
2. Harvesting outputs. “We held twelve workshops” is not an outcome. The outcome is what somebody did differently afterwards.
3. Claiming attribution. The contribution field asks what you did that plausibly influenced the change, alongside everything else that was going on. Harvests that read as sole-credit claims do not survive review.
4. Vague outcome descriptions. If an outcome cannot be checked by someone outside the organisation, it cannot be substantiated, which means it cannot be counted.
5. Using it to avoid targets. Outcome harvesting is not a way out of accountability; it is a different and in some ways more demanding form of it. Choosing it because pre-specification is uncomfortable rather than impossible is a misuse.
6. Leaving significance blank. A harvest of a hundred outcomes with no significance analysis is a database, not an evaluation.
7. Treating harvesting as continuous monitoring. It is a periodic exercise. Running it as a permanent reporting requirement exhausts everyone involved.

How Monival supports this

Directly and narrowly: Monival has no outcome harvesting module. Outcome mapping appears in the framework switcher as a not-yet-enabled option, and outcome harvesting is not in the product at all. We would rather say that than describe a form builder as a harvesting tool.

What is genuinely useful is that an outcome statement is a structured record, and Monival’s form builder handles structured records well. A harvest form with fields for the outcome description, its significance, the contribution claim, the social actor, the date and the evidence source is straightforward to build, and the harvest team can fill it in from the field — including offline, which matters when informants are interviewed away from connectivity. Submissions carry their own metadata and version, an approval workflow can route each drafted outcome for review before it enters the harvest, and the substantiation round can be run as a second form linked to the same outcome.

That is Monival being used as a disciplined collection instrument for a harvest, not as a harvesting methodology. The design of the harvest — the usable questions, the informant strategy, the substantiation approach, the analysis — is evaluation work, and Sibasi delivers it as consulting.