Most companies meet the eight effectiveness criteria once. A consultant maps the new grievance mechanism against Principle 31 of the UN Guiding Principles and produces a table with eight green cells. The table goes into the design report and the lender file. Nobody looks at it again. Three years later the mechanism has drifted. The suggestion box is emptied when someone remembers. The timelines in the procedure are aspirations. The satisfaction score is an average nobody has disaggregated. The eight cells are still green in the file.
The criteria were never meant to be a design checklist used once. The commentary to Principle 31 says they provide a benchmark for designing, revising or assessing a mechanism. It then adds the sentence that should be pinned above every grievance officer’s desk. A mechanism can only serve its purpose if the people it is intended to serve know about it, trust it and are able to use it. That is a statement about how the mechanism performs in the field this quarter, not about how it was drawn.
This article turns the eight criteria into an operating audit, run quarterly, with evidence per criterion and consequences attached.
Eight criteria, and what each one asks in practice
The text of Principle 31 is short enough to read in a minute and demanding enough to fail most mechanisms I have looked at. Legitimate: the mechanism enables trust from the groups it serves and is accountable for the fair conduct of its processes. Accessible: those groups know it exists and can reach it, with help for anyone facing barriers. Predictable: a clear, known procedure with indicative time frames for each stage, and clarity about outcomes and how implementation is monitored. Equitable: aggrieved parties have reasonable access to the information, advice and expertise needed to engage on fair and informed terms. Transparent: parties are kept informed of progress, and enough information is provided about performance to build confidence. Rights-compatible: outcomes and remedies accord with internationally recognised human rights. A source of continuous learning: lessons are drawn to improve the mechanism and prevent future harm. And, for operational-level mechanisms, based on engagement and dialogue: affected groups are consulted on design and performance, and dialogue is the means of addressing and resolving grievances.
Read that way, the criteria describe behaviour, not documents. Every one of them can be true in the procedure and false on the ground. The audit’s job is to find out which.
The lender frame points the same way. IFC Performance Standard 1 requires a grievance mechanism that is culturally appropriate, accessible and free from retribution. It also requires the company to monitor and measure the effectiveness of its management programme, of which the mechanism is part. A green table from the design year is not monitoring. A quarterly finding against each criterion, with evidence and an action, is.
Where mechanisms pass on paper and fail in the field
Legitimacy fails quietly when the mechanism is run by the department whose conduct generates the complaints. The community sees the same faces at intake and at the fence line and draws the obvious conclusion. I have sat in villages where people could name the grievance officer and the security supervisor in the same breath, and meant it as a single observation. Accessibility fails when channels decay. A hotline nobody answers reports the same zero as a hotline nobody trusts, and the design report cannot tell them apart. Predictability fails when the procedure promises an acknowledgment within five working days and the register shows three weeks, or when extensions happen without anyone telling the complainant.
Equity is the criterion most mechanisms fail without noticing, because the imbalance is structural. Picture a farmer negotiating over a cracked wall with a company engineer, a company lawyer and a company valuer. He has no access to the information, advice or expertise the criterion requires, and no amount of politeness fixes that. Transparency fails when the mechanism reports up but never back: the lender gets a quarterly table and the village gets silence. Rights-compatibility fails when settlements include waivers of legal recourse or when a remedy for a livelihood loss is a one-off payment that restores nothing.
The last two criteria fail together. A mechanism that logs cases and closes them, without asking why the same category recurs, is not learning. And a mechanism that was designed for the community rather than with it, and has never asked the community how it is performing, is not based on engagement and dialogue. It does not matter how many meetings it holds. The point I made in You Have a Grievance Mechanism. Now What? still holds: the existence of a process is not the same as its effectiveness, and the criteria are the difference.
Evidence per criterion
An operating audit needs evidence, not opinion, and each criterion points to its own. For legitimacy: who governs the mechanism, whether community representatives sit in that governance, and whether any case has been decided against the company’s operational interest in the last year. For accessibility: a channel test, done unannounced, that files a test grievance through every advertised route and records what happens. Add a perception check in the settlements furthest from the gate.
For predictability: the register itself. Acknowledgment-on-time rate, cases past their timeline, and whether every extension was communicated before the original date passed. For equity: a review of a sample of closed cases. Did the complainant have independent advice, have the technical evidence explained, and have time to consider the offer away from the room? For transparency: whether complainants were told what was happening at each stage, and whether the community has seen any aggregate performance data in a form it can read.
For rights-compatibility: a review of remedies against the harm, with particular attention to livelihood cases, waivers and anything involving security providers or gender-based harm. For continuous learning: whether any operational change in the last two quarters traces back to grievance data, and whether recurring categories have a root-cause file. For engagement and dialogue: when the community was last asked how the mechanism performs, what it said, and what changed. The relationship measures I set out in Social Performance Indicator Tracking sit naturally alongside this evidence set.
Reading silence and averages correctly
Two misreadings undo more audits than any missing document. The first is treating low intake as success. Consider a mechanism logging a handful of grievances a quarter beside an operation that blasts, hauls and draws down the water table. That is almost always a mechanism people have decided not to use. In my experience the quiet mechanisms deserve more worry than the busy ones. The diagnostic runs in order. Check the channels are alive. Then ask people whether they know the mechanism and would use it. Then re-run the barrier analysis for the current project phase. Expect the cure to look like deterioration, because a mechanism that starts earning trust sees intake rise before it falls.
The second misreading is the average. A satisfaction score of eighty percent can sit on top of a gap between men and women, or between the village by the gate and the village behind the hill. The average was built to hide it. Disaggregate every outcome measure by gender, location and grievance category before drawing a conclusion. A mechanism that rates men’s land claims as serious and women’s water claims as minor will produce an excellent average and a failing equity score. Only the disaggregated view shows it. I have seen registers in which cases were marked closed on the day the agreement was signed, with no evidence anywhere that the remedy was delivered. On the dashboard those cases counted as successes. In the village they counted as broken promises, and the village’s arithmetic is the one that decides whether the next grievance is filed or shouted.
Running the audit each quarter
The audit should be small enough to happen and independent enough to be believed. A half-day each quarter, led by someone outside the community relations function, with a community representative or an external practitioner in the room, is enough. It works through the eight criteria against the evidence above, records a finding per criterion, and ends with named actions and owners. Once a year, the same exercise is done in the community, in the community’s terms: what do people know, what have they used, what happened, would they use it again.
The output is one page. A single line can note that a large share of the quarter’s cases trace to one contractor. Or that women’s satisfaction is half of men’s. Or that no case has reached the mediated tier in a year. That line will move a general manager faster than a forty-page report. The design principles in Designing Grievance Mechanisms That Actually Resolve Conflict give the mechanism its shape; the quarterly audit is what keeps that shape from drifting.
Turn the criteria into a quarterly instrument
The UNGP 31 Effectiveness Audit Matrix is a framework in four sections covering the eight criteria in pairs. The pairs are legitimacy and accessibility, predictability and equity, transparency and rights-compatibility, and continuous learning and engagement. For each criterion it gives the practical test the audit team applies and the evidence to collect, sixteen entries in all. Use it by working across the two columns for each criterion: read the test, gather the evidence named beside it, and record a finding and an action. There is no score to total. The point is a written finding per criterion each quarter, so that drift is visible the quarter it starts rather than the year the lender notices. Download the UNGP 31 Effectiveness Audit Matrix
Dialogue is a criterion, not a courtesy
The eighth criterion is the one that holds the other seven together. A mechanism based on engagement and dialogue is one the community helped design and is asked to assess. It is one where grievances are resolved by talking rather than by decision. That is also the description of a mechanism people trust, know about and can use, which is what the commentary says effectiveness means. An operating audit that finds dialogue missing has found the reason the other criteria are slipping.
It is also why the mechanism needs a mediated tier for the cases negotiation cannot settle, staffed by someone independent of both the company and the complainant. That tier is where equity is restored for the farmer facing three company specialists. It is where rights-compatible outcomes are tested rather than imposed, and where the learning that prevents the next case actually happens. The Social Accord Architecture builds the grievance mechanism into a wider structure for exactly this reason. Its Feedback Loop carries the quarterly audit findings to a council where both company and community sit. Its mediation pathway, GROUNDS, is the escalation route the mechanism’s own procedure names. Dual accountability makes the audit honest. The company is accountable for acting on findings, and the community for using the mechanism it helped design rather than the road outside the gate. Livelihoods and investments are protected by the same instrument.
One thing to do before the quarter ends. Take the eight criteria, sit with your grievance register and your community liaison officer, and write one sentence of evidence against each. Where you cannot write the sentence, you have found the criterion that is failing. To design or audit a grievance mechanism against the criteria, or to set up its mediated tier, write to me at thomas@thomasgaultier.com.