Choosing a data system for a RAP: what to test before you commit

Nearly every resettlement programme starts in a spreadsheet, and for a small programme that is a defensible choice. The point at which it stops being defensible is identifiable in advance: it is the moment the file is edited by more than one person, or the moment a figure in it has to be explained a year after it was entered.

What a resettlement record actually has to do

The requirement is not data storage. It is the ability to answer, years later and to someone who was not there, a question of the form: what was this household entitled to, on what basis, what did it receive, when, and who decided?

A completion audit is that question asked repeatedly against a sample. A grievance is that question asked by the household. An accountability mechanism investigation is the same question asked adversarially.[4] A system that cannot answer it is not a resettlement system, whatever else it does well.

The capabilities that matter

1. Attribution and time-stamping on every change. Who changed what, when, from what value to what value. Without this the register is a current-state snapshot and cannot evidence anything about the past.
2. A data model with multiple claims per parcel. One owner per parcel makes tenants, sharecroppers and customary co-holders unrepresentable.
3. Versioned entitlement rules. Rates change; the system must record which version applied to each calculated entitlement.
4. Payment reconciliation to confirmed receipt, not to instruction.
5. Offline capture with explicit conflict handling, since enumeration happens without connectivity.
6. Grievances linked to the household record, so a complaint about a valuation sits against the valuation it concerns.
7. Documents and photographs attached to records, not stored in a parallel folder tree.
8. Export of everything, in an open format, without vendor assistance.

If the system cannot show what a record looked like on the date a payment was made, it cannot support the claim that the payment matched the entitlement.

Questions that expose a system quickly

Ask for a demonstration rather than a specification, and ask it to do these things with data in front of you:

• Show every change made to one household record, with who and when.
• Show the entitlement calculation for one household, broken into its components.
• Add a second claimant to a parcel with a different claim type.
• Change a crop rate and show which existing entitlements were affected and which were not.
• Produce a list of households paid less than they were entitled to.
• Produce a list of grievances open more than thirty days, by category.
• Capture a record with the device in airplane mode and then sync it.
• Export the whole dataset, including attachments and the change history.

The last one is the most revealing. A system that cannot export its audit trail has an audit trail that exists for the vendor's benefit rather than the project's.[2]

Where the spreadsheet breaks

Not gradually. The specific failures are: no record of who changed a figure or when; no way to reconstruct a prior state after a rate revision; silent formula breakage on row insertion; several files in circulation with divergent contents and no authoritative one; and personal data of thousands of households distributed by email attachment.

Any of these is enough to make a completion audit an exercise in reconstruction, which is the situation audits exist to detect.[1]

Ownership and continuity

A consideration that outranks features: who holds the data when the consultant demobilises. Where the register lives in a contractor's system and is handed over as a final deliverable, the implementing agency inherits a dataset it did not build, in a shape it cannot query, without the people who know its exceptions.[3]

The obligations outlive the contracts — livelihood restoration, monitoring and grievances all run past demobilisation — so the system has to be the agency's, with the consultant working inside it. That is a procurement decision made in the terms of reference, long before any software is chosen.

Data protection

A resettlement register holds names, locations, household composition, income, land claims and payment amounts for thousands of people. Most jurisdictions in the region now have data protection legislation that applies to it squarely, and the obligations — lawful basis, retention limits, access control, breach notification — sit with the implementing agency whether or not anyone assigned them.

At minimum: role-based access so enumerators cannot read payment data, no bulk export to personal devices, and a retention position that survives the end of the project.