Developmental Evaluation

The niche: innovation that has no model to judge yet

Conventional evaluation assumes there is a thing to evaluate: a programme with a design, objectives and an implementation the evaluator can hold still long enough to measure. Developmental evaluation exists for the situations where that assumption fails — social innovation in complex and uncertain environments, where the team is still discovering what the intervention is, the context keeps shifting under it, and the honest answer to “what is your model?” is “we are finding out” [1][2].

Patton’s claim is not that such situations are exempt from evaluation. It is the opposite: they need evaluation more, and they need it in a different mode — continuous, embedded, decision-facing — because the alternative in practice is that adaptation happens anyway, driven by anecdote, charisma and the loudest voice in the room [1]. DE puts systematic data and evaluative reasoning inside the adaptation loop instead of after it.

The approach carries its lineage openly. Patton built DE on the use-driven foundation of utilization-focused evaluation: the primary intended users are the innovating team, and the intended use is the development of the innovation itself [1][3]. What changes is the tempo and the stance — findings are not delivered, they are worked with, week by week.

DE is not formative evaluation on a longer contract

The most common confusion, and the one worth resolving first. In the classic two-purpose scheme, formative evaluation improves a programme that is heading toward a stable model, and summative evaluation then judges the merit and worth of that stabilised model. Formative work is preparation for judgement [1].

Developmental evaluation refuses that destination. Where the environment is genuinely complex — where what works this year in this county may not work next year or in the next one — there may never be a fixed model to hand over to summative judgement. Development, not refinement toward stability, is the purpose; the evaluation supports an intervention that intends to keep adapting [1][2].

The distinction has a contractual edge. A formative evaluation that never ends is a summative evaluation being evaded; a developmental evaluation is a different engagement declared as such at the outset, with different deliverables and a different definition of success [1]. Naming the purpose honestly at contracting time is the first quality test.

Purpose, timing and evaluator stance across formative, summative and developmental evaluation

Dimension	Formative	Summative	Developmental
Core question	How can the model be improved?	Does the model merit continuation or scale?	What is being developed, and what should change next?
Assumes	A model taking shape toward stability	A stabilised, implemented model	No fixed model; complex, shifting conditions
Evaluator stance	Advisor to implementers	Independent judge	Embedded team member bringing evaluative thinking
Timing of findings	Periodic, feeding revisions	End of cycle	Continuous, timed to the team's decisions
End state	Model ready for judgement	A verdict	Ongoing adaptation; no terminal verdict

The complexity concepts DE actually uses

“Complexity” is invoked loosely across the sector; DE borrows a specific, limited set of concepts and puts each to work [1]:

• Emergence. In complex settings, what the intervention becomes arises from interactions that cannot be fully specified in advance. Evaluation therefore tracks what is emerging — including effects and framings nobody planned — rather than only auditing fidelity to a plan that is itself in motion.
• Nonlinearity. Small actions can produce outsized effects and large investments can produce nothing; dose–response intuitions mislead. DE watches for tipping points and disproportionate results instead of assuming steady incremental change.
• Dynamic adaptation. The environment responds to the intervention and the intervention responds back. Evaluation questions themselves are expected to change as the situation does — a fixed question set held for three years is evidence the engagement is not actually developmental.

The discipline these concepts impose runs in both directions. They justify abandoning fixed-endpoint designs where conditions genuinely are complex — and they forbid claiming complexity as cover where conditions are merely inconvenient. Direct service delivery with well-understood outputs is not complex in this sense, and evaluating it developmentally wastes the approach [1][2].

What the developmental evaluator does

The role is concrete, and it is a working position inside the team rather than a visit schedule [2]:

• Asks the evaluative questions in real time. What are we observing? What does it mean? What are the implications for the next iteration? The evaluator’s craft is keeping these questions systematic when the team’s instinct is to move on intuition.
• Brings data to decisions on the decisions’ timetable. Rapid feedback from field observation, monitoring data, short-cycle interviews and whatever methods the moment requires — DE, like its parent approach, is methods-agnostic [1][3]. Where results vary sharply by setting, DE pairs naturally with the context–mechanism thinking of realist evaluation.
• Documents the development path. Decisions, forks taken and not taken, and the evidence in play at each fork. This documentation is DE’s accountability answer: the record shows that and why the intervention changed, which is exactly what a later summative evaluation of a scaled model will need.
• Surfaces problems early. The embedded position exists so that inconvenient data reaches the team while it can still act, not in a report after the fact.

The independence tension is real and should be managed, not denied. An evaluator embedded long enough to be useful is close enough to be captured — invested in the innovation’s success, softened by the relationships that make the role work. The available safeguards are structural: explicit agreement at contracting that the evaluator’s job includes unwelcome findings; documentation habits that leave an auditable trail; and periodic external review of the DE record itself. Teams that want only encouragement should hire a coach, not an evaluator.

Principles-focused practice

Within the DE family, Patton has developed a sibling practice — principles-focused evaluation — for interventions that are guided by principles rather than by a fixed model: the evaluand becomes the guiding principles themselves, and the evaluation examines whether they are meaningful, actually followed, and producing the intended results when followed. In adaptive programmes, principles are often the only stable object available to evaluate, which is what makes this line of practice a natural extension of DE’s logic. It is noted here as orientation; its book-length treatment sits outside this page’s verified sources.

Contracting and rhythm

DE fails on conventional contracts before it fails on method. The engagement needs [2]:

• Continuous or retainer-style engagement, not a baseline–midline–endline visit pattern; the value is presence at decisions.
• Feedback products sized to the decision cycle — briefings, working sessions, short memos — with the long report demoted to an occasional consolidation, not the unit of delivery.
• Success criteria written for development: the contract should say the deliverable is timely evaluative input to the innovation’s decisions and a documented development path, so that neither party later reinterprets the engagement as a delayed summative study.
• An exit trigger. When the innovation stabilises into a model, the developmental engagement has succeeded and should hand over — to formative refinement, and eventually to summative judgement of the stabilised model.

In development and humanitarian organisations, this rhythm is the evaluation-side counterpart of the broader adaptive-management agenda — building organisations that can absorb evidence mid-course and change what they do [4]. The organisational disciplines that make such absorption possible are treated under learning and adaptive management.

When not to use DE

The approach’s own literature is candid that DE is not the solution to every situation [2], and the boundary cases are worth listing plainly:

• The model is known and the question is accountability. A stable intervention with defined outputs needs monitoring and, when the causal question arises, a design from the methods cluster — not an embedded evaluator.
• The funder’s real requirement is a verdict. If continuation depends on a summative judgement at month 36, contracting DE sets everyone up for conflict; the purposes must be negotiated honestly first.
• “Developmental” is being used as shelter. A team that cannot say what it is learning, show the data behind its last three adaptations, or produce its decision record is not doing developmental evaluation — it is avoiding evaluation while borrowing the vocabulary [1]. The documentation trail is the test.