Choosing Data Collection Methods

The Method Selection Principle

Many M&E plans choose data collection methods in the wrong order. They start with "we should do a survey" and then work backward to find questions to fill it. As an illustration, a plan might end up with a very long questionnaire, a long analysis timeline, and a dataset that arrives after the decisions it was meant to inform have already been made.

The correct order is the reverse. Start with the decision to be made. Identify what you need to know to make it. Decide the minimum data quality that decision requires. Then select the method that delivers that quality at the lowest burden.

Two principles guide every method choice.

Methods follow decisions, not preferences. Suppose the decision is whether to extend a mobile clinic to more villages. That needs coverage data, meaning the share of people reached, and rates of use. A household survey can provide both. Routine records can show a trend in services delivered, but not coverage, because they have no count of the people who were not reached. Coverage needs a survey or a census to supply that count. A focus group cannot give the precision this decision requires, however comfortable the team is with qualitative methods.

Burden is a real cost. Every method takes time and energy from someone: community members, caregivers, community health workers (CHWs), field coordinators or data entry staff. That cost must be proportional to the value of the information. A 60-minute household survey of 600 households once a year is justifiable if it measures critical outcome indicators. The same survey every quarter is not.

Write down three things before choosing a method: the specific decision the data will inform, who will make it, and when they need the information. If you cannot complete all three, you are collecting data without a decision purpose. Stop and ask whether the collection is necessary.

The Main Data Collection Methods

Not every method suits every indicator. The table covers the eight methods most common in development M&E. Costs and times are typical ranges and vary by setting. Terms used in the table:

• Enumerators: trained field staff who collect survey data.
• Recall bias: faulty memory of past events.
• Social desirability bias: answering as others expect.
• KAP: knowledge, attitudes and practices.

Typical interview times: a household survey interview 30 to 60 minutes, a key informant interview 45 to 60 minutes, a focus group 60 to 90 minutes. Where connectivity is poor, use paper forms or phone forms that work offline and sync later. Whatever the method, get informed consent first and pretest the tools with a few respondents.

One pattern stands out. The methods best suited to accountability reporting (household surveys, facility assessments) are the most expensive and slowest. The methods best suited to routine monitoring (administrative records, mobile data collection) are the cheapest and fastest. An M&E system that blends both is more efficient than one that relies on a single method for everything.

Choose methods after the analysis plan, not before. If you need a stunting prevalence with a 95% confidence interval, you need a probability sample survey, where every household has a known chance of being selected. If you only need a directional trend, such as whether coverage is improving or declining, routine monitoring records are enough. Starting with the method and working backward to the analysis almost always leads to collecting too much or the wrong thing.

Before and After

The difference between hasty and thoughtful method selection is easy to see side by side.

Before (method chosen by habit):

"We'll do a survey to measure all our indicators."

After (methods matched to indicators):

"Indicators on prevalence and knowledge will be measured through a household survey with an anthropometric component (measuring height and weight) and a KAP module. Indicators on service delivery and quality will be tracked through program records and quarterly facility observation. Indicators on community perceptions will be explored through FGDs with caregivers. Indicators on system-level change, including referral completion, will be assessed through KIIs with health staff and a review of facility documents. The survey and the FGDs will run together at baseline, midline and endline as a convergent design (both methods run at the same time and the results are compared). Field coordinators will review program records monthly and consolidate them quarterly."

The second version is not more complicated. It is more honest about what each indicator needs.

Worked Example: Healthy Futures

Healthy Futures is a fictional child nutrition program in Kasenda District, funded by the Global Health Partnership Fund. It has 18 indicators. The M&E officer does not ask whether the household survey can measure all of them. For each indicator, the officer looks for the lowest-burden method that gives the quality the decision needs. Three indicators show the pattern.

Indicator 3 is the share of children aged 6 to 23 months who eat at least 5 of 8 food groups.

The baseline survey covers 600 households, 40 per community. The M&E officer manages it. 12 enumerators and 3 supervisors collect the data, and the zone leads support data entry. Nine FGDs, 3 in each zone, own no indicators. They explain the survey results, for example what makes nutritious foods easy or hard to reach.

The timing check finds that running the survey, the FGDs and the CHW assessment in the same weeks would overload the three zone leads. The M&E officer staggers the CHW assessment to follow the survey fieldwork.