An NGO can run a strong programme and still have weak programme data. Field staff collect what the form asks for, area offices aggregate in spreadsheets, and the donor report is assembled at the deadline from three versions of the truth. The programme is real; the evidence is not reliable enough to manage it or defend it.
This troubleshooting guide covers the problems that recur across programme data systems in Bangladesh, the causes behind them and the fixes that hold when field teams are busy and reporting deadlines do not move.
Symptoms worth investigating
- The same beneficiary appears in more than one project register.
- Two reports for the same period show different totals.
- Field staff fill long forms that nobody uses for decisions.
- Consent is recorded once, in general terms, and never revisited.
- The donor report takes days of manual reconciliation.
- Ex-staff still have access to beneficiary files.
- Photos and case stories are shared without a clear permission trail.
These are not separate failures. They usually trace back to unclear ownership and a data model designed project by project.
Problem 1: parallel registers and duplicate beneficiaries
Cause: each project creates its own register, often in Excel, with different identifiers. A person enrolled in two projects is counted twice, and referral follow-up is impossible.
Fix: establish one beneficiary identity within the organisation, with a unique identifier and a controlled set of matching rules. Match on a combination of name, date of birth, guardian name, mobile number and address, with human review for uncertain matches. Keep project enrolment separate from the person record so one person can participate in several programmes without being duplicated.
Evidence: a duplicate report, a merge log and a sample of verified records.
Problem 2: indicators mean different things
Cause: each donor and project defines an indicator slightly differently. "Households reached" may mean registered, attended one session or completed the full cycle. The field team guesses, and the aggregate becomes unusable.
Fix: publish an indicator dictionary with a plain-language definition, the numerator, the denominator, the collection method, the frequency and the owner. Map donor indicators to the internal definition instead of changing the field form for every proposal. Where a donor definition genuinely differs, record the mapping and the adjustment.
Evidence: the dictionary, a mapping table and a reconciled report for one project.
Problem 3: consent and data sharing are unclear
Cause: consent is captured as a signature on an enrolment form, without explaining purpose, sharing or retention. Field staff share lists with partners by WhatsApp because it is fast. Nobody owns the decision.
Fix: write a short consent notice in the local language, separate consent for programme delivery, referrals, photography and communications, and record what was agreed. Define who may share which fields with which partner, and use a controlled transfer method rather than personal messaging. Build a process for withdrawal and deletion.
Evidence: the notice, consent records, a data-sharing register and a tested withdrawal process.
Problem 4: field data quality is poor
Cause: forms are designed at head office without field input, collect too many fields and allow free text where a list would do. Offline forms create conflicting copies, and supervision focuses on submission, not accuracy.
Fix: shorten the form to the fields used in decisions, add validation at entry, use controlled lists and test the form with field staff before rollout. For offline collection, define how records merge and who resolves conflicts. Add a short verification step: call a sample of households, check a sample of sites and record the result.
Evidence: form review notes, validation rules, conflict logs and a verification sample.
Problem 5: reporting is a monthly reconstruction
Cause: data lives in project files, partner reports and personal inboxes. The reporting officer collects numbers rather than querying a system. Definitions change mid-project, and late partner reports arrive after the donor deadline.
Fix: agree a reporting calendar with data cut-off dates, publish a single reporting dataset and reconcile partner submissions against it. Lock the indicator definitions for the reporting period. Automate the standard tables and keep the narrative separate. Where a partner reports late, record the gap rather than estimating silently.
Evidence: the calendar, a reconciled dataset and a report that can be traced back to source records.
Problem 6: access and retention are unmanaged
Cause: shared drives, generic logins and downloaded beneficiary lists on personal laptops. No one removes access when a contract ends, and old project data is kept indefinitely because deletion feels risky.
Fix: define roles by task, use named accounts and require multi-factor authentication for remote access. Keep sensitive extracts off personal devices, encrypt laptops and phones, and review access quarterly. Set retention by data type and funder requirement, then delete or archive with a record of the decision.
Evidence: an access matrix, review notes, a retention schedule and a deletion log.
Diagnostic sequence
- Pick one programme and trace a beneficiary from enrolment to the latest donor report.
- Count the registers and spreadsheets that hold the same person.
- Compare the indicator definitions used by the field, the monitoring and evaluation team and the donor.
- Review consent forms and ask three field staff to explain what a participant agreed to.
- Test one offline form merge and inspect the conflicts.
- List who can access the beneficiary data today, including former staff and partners.
- Time how long the last report took and where the delays occurred.
A fix plan that fits a programme calendar
First month: agree the beneficiary identifier, publish the indicator dictionary and stop creating new project-specific registers.
Second month: clean and merge one programme's data, test consent and sharing rules, and shorten the field form.
Third month: run a report from the cleaned dataset, compare it with the old process and fix the gaps. Then extend the model to the next programme rather than attempting an organisation-wide redesign at once.
Metrics
- Duplicate rate found in the beneficiary register.
- Completeness of required fields at enrolment.
- Percentage of reports submitted by the agreed cut-off date.
- Verification coverage and the error rate found in sampled records.
- Number of systems holding the same beneficiary data.
- Time from data request to a reconciled response.
Where this fits in our work
If the programme needs a data model, field workflow and reporting layer that agree with each other, start with Custom Software and Integrations. The pricing page explains how discovery and phased delivery are scoped. For the foundation work, read how AI-ready data foundations develop over time.
Sources
- International Aid Transparency Initiative, IATI Standard, accessed 17 November 2026.
- NGO Affairs Bureau, Government of Bangladesh, accessed 17 November 2026.
- NIST Privacy Framework, accessed 17 November 2026.
- DataGuidance, Bangladesh jurisdiction overview, accessed 17 November 2026.
- UN OCHA, Humanitarian Data Exchange, accessed 17 November 2026.
Responses (0)
No comments yet. Add the first one below.
Leave a response
Your email address will not be published. Required fields are marked *.