世 光 · 工 具 箱
SEESALT TOOLBOX · STAKEHOLDER ENGAGEMENT
Consultation is not participation
Engagement is the part of development work that everyone claims to do and almost nobody audits. The failure is rarely a lack of goodwill; it is that the decisions were already made before the room was convened. This page collects the frameworks we actually use, the questions we ask before designing a process, and the ways we have seen it go wrong — including in our own work.
How engagement fails
Predictable, and therefore preventable
Bad engagement does not usually look like refusal. It looks like a well-attended meeting that changes nothing. Six failure modes account for most of what we have seen:
- Consultation after the decision. The budget, the site and the logframe are fixed; the meeting is about implementation. Participants can tell, and they respond by disengaging — which is then read as apathy.
- Consulting the accessible rather than the affected. Whoever speaks the donor's language, owns a phone, and can spare a weekday afternoon becomes "the community".
- One meeting, treated as a mandate. A single session is a snapshot of who could attend that day. It is not consent and it does not survive a change of plan.
- Extraction without return. Communities give time and knowledge and never learn what happened with it. The second round of engagement is always harder than the first, and this is why.
- Delegating engagement to the least senior person. If the people in the room cannot change the design, the room cannot change the design.
- Measuring attendance instead of influence. Counting participants tells you about logistics. It tells you nothing about whether anything moved.
The single diagnostic question. Ask of any engagement process: what specific decision could this process still change, and who has the authority to change it? If there is no honest answer, what is being run is communications, not engagement — and it should be described that way to the people attending.
Stakeholder analysis
Who is affected, who decides, and who is missing
Start by separating two questions that get merged: who is affected by this, and who can affect this. They are different lists, and the gap between them is usually where the injustice sits.
The power–interest grid
The standard first pass. Plot each party by how much influence they hold over the decision and how much the outcome matters to them.
Power–interest mapping is common to most planning traditions. Treat the axes as questions rather than measurements: power over what decision, interest measured how, and assessed by whom?
Three refinements worth the extra hour
- Legitimacy and urgency, not only power. Mitchell, Agle and Wood's salience model adds two axes to power: whether a claim is legitimate, and whether it is urgent. A party can be powerless, entirely legitimate and running out of time — that combination is invisible on a two-axis grid.
- Map internal difference. "The community" is not a stakeholder. Within any village there are people who gain and people who lose from the same intervention. If your map has one box labelled with a place name, it is not finished.
- Name who is structurally absent. Write the list of those who cannot attend: people doing unpaid care work at meeting time, those without transport, those whose participation would be observed by someone with power over them. This list is the most useful output of the exercise.
How much power are you actually sharing?
Arnstein's ladder, and what it is for
In 1969 Sherry Arnstein published an eight-rung ladder describing degrees of citizen power in public decision-making, grouped into three bands: non-participation, tokenism, and citizen power. It is the reason the phrase "consultation is not participation" exists, and it remains the fastest honesty test available for any engagement plan.
Most development "participation" sits on rungs 3 to 5. That is not automatically wrong — informing is a legitimate activity. It becomes wrong when rung 4 is described to participants as rung 6.
Using it honestly
- Pick the rung before the meeting, and say it out loud. "We are consulting; the decision rests with the ministry" is a defensible sentence. Letting people assume partnership and discovering otherwise is not.
- Different decisions can sit on different rungs. A project may inform on budget envelope, consult on timeline, and delegate on site selection. Map rungs per decision rather than per project.
- Note the direction of travel. A programme that begins at rung 3 and reaches rung 6 in year three is doing something different from one that stays at 3 for a decade. Where the ladder is used as an evaluation tool, the movement is the finding.
The IAP2 Spectrum of Public Participation — inform, consult, involve, collaborate, empower — is the practitioner-facing descendant of Arnstein's ladder, with the moral edge filed off. It is more usable in a workshop; the ladder is more useful when you need to be honest with yourself. We use both, for different rooms.
Designing a consultation
Twelve questions before anyone books a room
Most consultation design happens implicitly, through defaults — the meeting is held where the office is, when staff are free, in the language the report will be written in. Each of those defaults selects a subset of participants. These questions surface the defaults.
- What decision is genuinely open, and who can change it?
- Which rung of the ladder is this, and have we said so?
- Who is affected but has no lever, and what would it take for them to attend?
- Who is excluded by the time of day we have chosen?
- Whose presence would silence someone else in the room? (Employers, in-laws, local officials, the person who allocates aid.)
- In whose language does this happen, and who translates — a neutral party, or one of the negotiating sides?
- Who pays for the travel and the lost working hours? If the answer is "the participants", we have designed a process only the comfortable can join.
- What is the alternative channel for people who will not speak in a group?
- How will we record disagreement, and where will it appear afterwards?
- What are we going to do with what we hear, and by when?
- How will participants find out what happened? Through what channel, in what language, and who is accountable for sending it?
- If nothing changes as a result, will we say so plainly?
Where consent, not consultation, is required. Where a project affects Indigenous peoples' lands, territories or resources, the applicable standard is Free, Prior and Informed Consent — a right recognised in the UN Declaration on the Rights of Indigenous Peoples and operationalised in the policies of several UN agencies and financial institutions. FPIC includes the right to withhold or withdraw consent. A consultation framework is not a substitute for it, and calling an FPIC process "consultation" is a material misstatement, not a wording preference.
Engaging civil society
Partners, not implementers
Local organisations are routinely brought in at the point where a design needs delivering rather than at the point where it is decided. That arrangement transfers risk downward and authority nowhere.
| Question | What a weak answer looks like | What to aim for |
|---|---|---|
| When were they brought in? | After the proposal was submitted. | Before the problem statement was fixed. |
| What share of the budget do they hold? | A per-diem line and a workshop fee. | Core costs, including their own overheads and staff time. |
| Who owns the outputs? | The international partner, with acknowledgement. | Joint, with the local partner able to reuse them independently. |
| Who talks to the funder? | Only the international partner. | Both, and not always with the other present. |
| What happens when the grant ends? | The relationship ends with it. | Capability, data and relationships remain with them. |
Two things worth being explicit about
- Civil society is not homogeneous, and not automatically representative. An NGO may be the most accessible route to a community and still not speak for it. Ask who elected or appointed them, and to whom they report.
- Do no harm applies to engagement itself. In contested settings, visibly partnering with one organisation can expose it, or shift local power in ways you will not see and will not bear the cost of. Ask what association with you costs them.
Working across languages and cultures
Where meaning is lost, and where it is taken
Cross-cultural communication guidance often stops at etiquette. The harder problems are structural.
- The concept may not travel. Terms like participation, accountability, gender, resilience and even community arrive already loaded from elsewhere. Before translating a questionnaire, find out how people describe the same phenomenon unprompted — and expect the categories, not just the words, to differ.
- The interpreter is a participant. Whoever translates chooses register, softens conflict and decides what is worth relaying. If they are employed by one side, that is a structural bias, not a personal failing. Where a disagreement matters, brief the interpreter that literal rendering is wanted, including rudeness.
- Silence is not agreement, and it is not always deference. It can be a considered refusal to legitimise the meeting. Build a channel that does not require speaking in front of the group, and treat what comes through it as equally valid data.
- Write down what you could not translate. A short list of terms that did not carry is often the most useful analytical output of a field visit. It marks where the framework does not fit — which is information, not noise.
The Core Humanitarian Standard makes this an obligation rather than a courtesy: it requires communication in languages and formats that are accessible, understandable and contextually respectful, and participation that is meaningful to the people participating — not merely to the organisation recording it.
Feedback and accountability
The loop that almost never closes
Collecting feedback is easy and common. Acting on it, and telling people you did, is neither. A feedback mechanism that never visibly changes anything teaches the community that speaking is pointless — which is a worse outcome than not having asked.
A minimum viable mechanism
- More than one channel, including at least one that does not require literacy, airtime, or approaching a staff member in person.
- A named person accountable for response times, not a shared inbox.
- A separate, confidential route for complaints about staff conduct, which cannot be the same channel as the one staff manage.
- A published turnaround commitment, and honesty when it is missed.
- Visible closure. What was raised, what changed, what did not change and why — communicated back through the same channels people used, in the same languages.
The test we apply
Can you name a specific decision in the last twelve months that was changed because of something a community member said through a feedback channel? If not, the mechanism is collecting, not accounting — and that should be reported as a finding rather than left unexamined.
Grievance redress mechanisms are also a formal requirement under most international financial institutions' environmental and social standards. Meeting the requirement on paper and having a mechanism people actually use are different achievements, and only the second one protects anyone.
Was the engagement real?
Indicators that are hard to fake
Attendance counts, satisfaction scores and photographs of meetings are all easy to produce without sharing any power. These are harder:
- Documented design changes attributable to participants, with the specific input and the specific change recorded next to each other.
- The proportion of raised issues that received a traceable response, and the median time to that response.
- Participation by people who are structurally hardest to reach, measured against your own absence list from section 02 — not against the general population.
- Disagreement on the record. A consultation report with no documented disagreement describes a meeting where disagreement was not possible.
- Return participation. Whether people who attended once come back. This is the single best proxy for whether the first round was worth their time.
- What the community says without you in the room. Ideally gathered by someone who does not work for you.
An honest scorecard has failures in it. If your engagement indicators are all green, the most likely explanation is that they were chosen to be.
Sources, and what this is not
Where these frameworks come from
Almost nothing on this page is ours. It is a working synthesis of established practice, assembled because we needed it and kept because others might. The frameworks belong to the people who built them:
- Arnstein, S. R. (1969). A Ladder of Citizen Participation. Journal of the American Institute of Planners 35(4), 216–224. doi:10.1080/01944366908977225 · summary and commentary
- IAP2 Spectrum of Public Participation — inform, consult, involve, collaborate, empower; the practitioner descendant of Arnstein's ladder. a comparison of the two
- Mitchell, Agle & Wood (1997) — stakeholder salience through power, legitimacy and urgency.
- Free, Prior and Informed Consent — recognised in UNDRIP; practical guidance in the FAO manual. FAO · FPIC
- Core Humanitarian Standard on Quality and Accountability, 2024 edition — nine commitments on participation, communication and accountability to affected people. corehumanitarianstandard.org
- Do No Harm — Mary B. Anderson's framework on how assistance interacts with conflict dynamics.
What this page is not
It is not a standard, not a certification, and not evidence that we have applied all of it well. Several of these tests we have failed. The feedback-loop test in section 07 is one we set ourselves and have not yet passed for every programme, and we would rather publish the test than quietly drop it.
If you use these frameworks in practice and think something here is wrong, incomplete, or naive about how funding actually works — that correction is more useful to us than agreement.
SEESALT Toolbox
Corrections and additions — zhijun.he@yale.edu
Frameworks cited above belong to their authors. This synthesis is offered freely for reuse and adaptation.
播 种 光 明 ， 收 获 永 恒