The completion audit: what it actually tests and why projects fail it

1. What the audit is for

A completion audit exists because resettlement obligations do not end at payment. The standards require that displaced people's livelihoods be restored, and restoration is an outcome observable only after time has passed. [1][5] An audit conducted at the end of implementation asks whether that outcome was achieved, and whether the commitments made along the way were delivered.

It is normally required to be independent of the party that implemented the programme, for the same reason financial audit is independent [1] — the completion audit itself is a PS-5 requirement: the implementer's assessment of its own performance is not evidence, however honest. Independence here means more than a different firm — it means a party with no stake in the finding, which is a condition that sub-contracting to a consultant who hopes for further work does not satisfy.

The audit's conclusion typically identifies outstanding obligations and, where livelihoods have not been restored, what further measures are required. In lender practice this can hold open a project's environmental and social closure long after physical works are complete, which is the mechanism giving the requirement its force.

2. Sampling, not appraisal

Auditors do not read a programme and form a view of its quality. They select cases and test them. A sample of households is drawn, and for each the auditor traces what the instrument entitled them to, what the records say they received, and what the household says happened. Discrepancies between the three are the audit's raw material.

This method explains findings that project teams experience as perverse. A programme that treated most households well but cannot produce records for the sampled ones will fail; a programme with mediocre substance and complete records will pass more comfortably. The auditor is not indifferent to substance — the household interview is precisely where substance is tested — but they cannot generalise from cases they cannot examine.

It also explains why the hard cases dominate the finding. Guidance Note 5 sets out the categories an audit is expected to test with particular care. [2] Auditors deliberately oversample the difficult categories: absentee owners, disputed parcels, female-headed households, tenants, people who refused an offer. These are where entitlement frameworks are stressed and where a programme's real quality shows, and a project that reports only aggregate completion has not looked where the auditor will look.

3. The baseline problem, again

The audit's central question — were livelihoods restored to at least pre-project levels — requires a pre-project measurement. Where the socioeconomic baseline was inadequate, no audit can answer the question, and the finding will record that restoration could not be assessed.

This is the single most consequential determinant of audit outcome and it is fixed years earlier, usually by people who will not be present for the audit. A baseline that recorded income in broad bands, that sampled rather than enumerated, that omitted the seasonal composition of livelihoods, or that was conducted after acquisition had already disturbed the economy, forecloses the comparison permanently.

Projects sometimes attempt retrospective reconstruction — asking households to recall pre-project income. Auditors treat recall data with appropriate scepticism, particularly where the respondent has an interest in the answer, and its use tends to weaken rather than strengthen a project's position.

4. Why anticipation is the whole game

Everything an audit requires — an entitlement framework applied consistently, individuated records of what each household received, contemporaneous evidence of engagement and disclosure, a baseline capable of supporting comparison, monitoring conducted over a meaningful period [3][4] — must exist before the audit begins and cannot be created afterwards. [6]

This makes the audit unusual among compliance instruments in that preparation for it is indistinguishable from doing the work properly. There is no separate audit-preparation activity that a well-run programme needs; a well-run programme is already audit-ready, and a badly-run one cannot become so.

The practical implication is that the audit's disciplinary value operates at the start of a programme, not the end. Teams that ask at design stage what the completion audit will require make different decisions about census instruments, record systems and monitoring design. Teams that first consider the audit when it is scheduled discover a set of requirements that are, by then, unmeetable.

5. Conclusion

The completion audit tests whether a project can demonstrate what it did, to whom, and with what result. It is not a review of intentions or of effort, and projects that were well-intentioned and hard-working still fail it when their records cannot carry the weight the method places on them.

Its real function is prospective. An audit that everyone knows is coming changes what gets recorded while it can still be recorded, which is worth more than the finding it eventually reaches.