Custody, Not Capture: Data Stewardship Terms for U.S.–Madagascar Partnerships

A household survey can appear to be a neutral instrument: a set of questions, a tablet, a schedule. In practice, every survey creates a record of names, incomes, land use, health conditions, and daily routines. Those records outlast the grant cycle. They travel through email threads, cloud folders, and donor reporting systems. For U.S. funders working with Malagasy community partners, the terms that govern that record are too often settled after collection has started, if they are settled at all. A data stewardship addendum, agreed before the first interview, is a more reliable approach.

Custody Is a Better Starting Point Than Ownership

Ownership language is legally convenient but socially blunt. A funder may ask who owns the data, expecting a single answer. In many community-led settings, knowledge is held collectively and responsively: a village committee may speak for a community on some questions, a cooperative on others, and individual households on matters that concern them alone. A clause that assigns all ownership to one party can silence those distinctions.

A clearer structure separates three things: custody, control, and benefit. Custody describes who holds the records day to day. Control describes who may approve access, transfer, or secondary use. Benefit describes what the community receives when data produces value. The U.S. funder may reasonably require a license to use data for reporting, evaluation, and audit. That is not the same as taking title. The addendum should say so in plain language.

Name a Local Data Steward Before the First Interview

A data steward is a named person with defined authority, not a generic office. The steward should be based with the Malagasy partner organization and should have a written role in the grant budget. Responsibilities may include receiving collected files, maintaining the access log, reviewing requests from the funder or third parties, arranging translation of summaries, and deciding when a dataset is too sensitive to share in raw form.

This role should not be treated as a mailbox. The steward needs enough standing to say no, and enough time to say yes carefully. If the U.S. funder expects weekly data transfers, the steward should agree to that schedule in advance and be compensated for the labor. The steward should also review the survey instrument before launch, because some questions create obligations that no storage clause can repair later.

Decide Where Data Lives and Who Can Reach It

Storage terms should be specific. Will data be held on local servers, in a U.S. cloud account, or both? Who holds the encryption keys? What happens when internet access is intermittent? Many household surveys in Madagascar are collected offline and synchronized later, which creates temporary copies on multiple devices. The addendum should identify those copies and require deletion after synchronization.

Access should be tiered rather than binary. A funder program officer may need aggregate results. An external evaluator may need de-identified records. A community committee may need household-level information to verify targeting. These are different requests, and they should not all be answered by a single shared folder. The agreement should require an access log, a retention schedule, and a breach notification process. It should also state which data categories are never to be transferred without renewed consent, such as precise locations of vulnerable households or records involving minors.

Build Return of Findings Into the Schedule

A common failure is not the loss of data but the loss of meaning. Findings return to the funder as a report and to the community as a distant memory of being surveyed. The addendum should require return in usable formats: translated summaries, data dictionaries, community briefings, and where appropriate, visual dashboards. If the funder receives a cleaned dataset, the partner should receive a copy in a format it can open without proprietary software.

Return is not only a courtesy. It is a condition of legitimate partnership. The schedule should name deadlines for interim findings, not only final reports, and should identify who presents results back to participating households. Budget lines for translation, validation meetings, and local review are part of the data terms, not separate communications costs.

Consent for Today and Reasonable Tomorrow

Consent forms should anticipate future use without demanding blanket permission. A household may agree to participate in a livelihood survey and may not agree to have its responses used for commercial product research. Tiered consent, explained in Malagasy and in plain language, gives people a genuine choice. The addendum should describe how secondary research use will be requested, how commercial use will be handled, and whether benefit-sharing applies if data contributes to a product or service.

Withdrawal deserves candor. Once data are aggregated, withdrawal may no longer be technically possible. The agreement should say that clearly at the point of consent, rather than implying a right that cannot be honored. It should also state how requests for deletion will be handled for identifiable records and how anonymization will be verified.

Photos, Coordinates, and the Precision Problem

Photographs and GPS points are data, and they carry risks that a questionnaire does not. A geotagged image can reveal the location of a sacred site, a land claim, or a household receiving support. Precision should be a deliberate choice. The addendum can require aggregation of coordinates, buffering of sensitive locations, and separate consent for identifiable images. Metadata should be scrubbed before public release. These are not technical niceties; they are terms of respect.

Put It in an Addendum, Not a Footnote

A workable addendum need not be long. It should define data categories, name the steward, set storage and access rules, require return of findings, describe consent for future use, and address photos and coordinates. It should be reviewed by counsel familiar with both U.S. grant requirements and Malagasy law, and it should include a plain-language summary for community participants. Most importantly, it should be signed alongside the grant agreement, not appended after the survey has already begun.

Data terms are partnership terms. If a U.S. funder cannot say who decides, where the records live, and how findings return, the survey is not ready. Custody is not a bureaucratic detail. It is the difference between evidence gathered with a community and knowledge taken from it.