Plan the Answer Before the Questions
An analysis plan specifies how evidence will answer an operational question. An analytical framework organizes the subjects and relationships; an analysis plan makes them usable by naming the variables, sources, comparisons and outputs. A questionnaire is only one possible collection instrument. Starting with a long form can produce detailed information that does not resolve the decision.
The WFP analysis-plan guidance, updated October 10, 2025, identifies objectives, questions, sources, methods, disaggregation and reporting outputs. Its linked food-security materials need their own context and skills. The table below is an original general teaching example, not WFP's required template.
Define the Decision Boundary
Fictional case: after flooding, the Reach response lead must decide where to request a technical water-service check. The lead has reports from two settlements, an older population list and a service operator's update. The task is to prioritize verification, not certify water quality, select individual recipients or estimate disease prevalence.
Write the geographic boundary, reference period, affected groups and decision deadline. State whether the unit is a water point, a visit, a household or a person. Do not switch units when combining reports. Record the distinction between planned capacity, reported functioning and actual safe access.
Fill the Analysis Plan
| Decision Question | Information and Unit | Source and Method | Intended Analysis and Output | Missing-Data Rule |
|---|---|---|---|---|
| Which reported service disruption needs technical verification? | Operational state of each named water point at a stated time; date and basis of report | Check operator update against permitted observation notes and coordination information | Point-by-point status table with source confidence and disagreements | Unknown stays unknown; an absent entry is not a functioning point |
| Who may be excluded despite a point being open? | Reported barriers by relevant location or access need, without identifiable case details | Accessible voluntary discussions and existing aggregate information, subject to protection review | Separate thematic account of barriers; do not calculate a population rate from convenience interviews | Record who could not safely participate and what that prevents the team concluding |
| What can the response lead ask for now? | Verified disruption, plausible options, technical/protection checks and available capacity | Technical lead and service owner review the evidence and unresolved constraints | Provisional verification request with owner, purpose and review trigger | If quality or safety is unverified, request a qualified check rather than announce that supply is safe |
Specify Comparisons Before Calculating
For a numerical comparison, write the numerator, denominator, time window and exclusions in advance. A count of visits cannot be divided by a household register to obtain a percentage of people using the service. If a source changes definition or coverage, show separate series or reconcile them with a documented basis.
Disaggregate only when it serves a decision and can be done safely and meaningfully. Name the categories, explain why they matter, check whether groups are missing and consider small-cell identification risk. An empty cell can mean uncollected, withheld or inapplicable information; it is not automatically zero. Combine or restrict outputs when the responsible data/protection roles judge detailed release unsafe.
Design Collection Only for the Remaining Gap
The UNHCR Needs Assessment Handbook, Part 1, printed pages 41-43, links objectives, analysis plans and a review of existing data before additional collection. Its institutional roles apply to its operating context. Here, the responsible assessment lead must decide whether a gap justifies collection, whether a suitable method is feasible and what safe alternative exists.
Test the instrument and the analysis together. Can a colleague build the proposed output from the proposed fields? Does the question measure the intended concept? Would an answer change the operational choice? If not, remove or revise it. Budget for translation, accessible formats, staff preparation, secure handling and analysis, rather than treating collection as the whole task.
Keep a No-Data Outcome Usable
In the fictional case, the operator cannot confirm the second settlement's current status and safe access prevents a visit. The output reads: "Status at the cutoff is unverified. The response lead will seek a permitted technical update through the service owner. No conclusion about safe water or the relative severity of needs is made for that settlement." The planning map shows the gap visibly; it does not color it as low need.
Version the plan when questions, sources or conditions change. Preserve why the change was made and which earlier comparisons it affects. Continue with the secondary-data review or turn findings into response options.