Indicators, sampling, questionnaire, fieldwork and the baseline report
This guide shows how to conduct baseline survey work for an NGO, CSR or government programme before activities begin. You will define indicators, draw a sample that includes a comparison group, build and pilot a questionnaire, run verified fieldwork and write a short baseline report. Done well, it gives your endline a fair number to compare against.
- Nine steps from indicators to report
- A questionnaire outline you can adapt
- Sample size arithmetic in plain language
- The mistakes that ruin endline comparisons
Hello, I am Ram Kumar. This survey is run by Sample Organisation. Purpose: to learn about amenities in your area. It is voluntary, and you can say no at any time.
- Comes regularly
- Comes sometimes
- Rarely / never
- Don't know
A baseline survey measures the situation of the people or places a programme will serve before the programme starts. It records the same indicators you will measure again at midline or endline, ideally for both participants and a similar comparison group, so later change can be attributed with some confidence.
Timing matters more than most teams expect. If the baseline happens after enrolment camps, distributions or trainings have begun, you have measured an early midline and you will understate impact. Schedule fieldwork in the weeks between final site selection and the first activity, and avoid festival periods and peak harvest when households are away or busy.
Season also affects the numbers. Income, migration, water availability and school attendance all swing across the year. Whatever month you pick for the baseline, plan the endline for the same month.
Before you start, have
- Final list of programme villages, wards or schools
- Logframe or theory of change
- Budget and field window
- Ethics or consent approach
- A named owner for the data
These steps apply to household, farmer, student and facility baselines alike. Each ends with something written down, which is what your endline team will need years later.
Pick 8 to 15 outcome indicators
Take them from the logframe. Each indicator needs a definition, a unit and a recall period, for example 'households with year-round drinking water within 500 metres, last 12 months'.
- Outcome over output
- Clear numerator and denominator
- Measurable again at endline
Decide the comparison
Choose non-programme villages or schools that resemble programme sites on size, access and social profile. Without them you can only show before-and-after change, which also captures everything else that happened.
- Matched comparison sites
- Same district where possible
- Record why each was chosen
Calculate sample size
Work from the smallest change you need to detect, the baseline value, the number of clusters and the expected dropout by endline. Write a one-page sample size note.
- Minimum detectable change
- Design effect for clusters
- 10–20% buffer for attrition
Draw the sample
List households in each selected cluster, then pick them randomly with a fixed rule. Keep the list: it is how you find the same homes later.
- Fresh household listing
- Random start and interval
- Replacement rules agreed
Build the questionnaire
Map every question to an indicator or to a variable you will use for analysis. Cut questions that serve neither.
- Skip logic and range checks
- Local-language wording
- Consent at the start
Translate, back-translate and pilot
A second translator converts the local version back to English, and differences are resolved. Pilot with 20 to 30 households outside the sample.
- Timing per interview
- Confusing words replaced
- Final version frozen
Train enumerators and supervisors
Three to four days is common: concepts, consent, practice interviews, device use and a field day. Test enumerators before deployment.
- Mock interviews
- Neutral probing
- Written test
Collect and verify data
Use a mobile app that records GPS, time and duration. Supervisors back-check a fixed share of households with a short form and flag mismatches daily.
- Daily sync and review
- Back-check 5–10% of interviews
- Out-of-area records held
Clean, analyse and write the baseline report
Report each indicator with its denominator and confidence interval for programme and comparison groups, and store the tools and data for the endline team.
- Codebook and data dictionary
- Indicator table
- Limitations stated
A household baseline questionnaire usually runs 35 to 50 minutes and has eight modules: identification and consent, household roster, housing and assets, livelihoods and income, the programme's core outcome module, access to services, knowledge and practices, and interviewer observations. Adapt module order so sensitive questions come after rapport is built.
Module outline you can adapt
Each module lists sample question types. Replace the outcome module with your sector's indicators.
- Identification: cluster code, household ID, GPS, date, interviewer ID, consent recorded.
- Roster: members, age, sex, relation to head, education, main activity.
- Housing and assets: roof and wall material, toilet, cooking fuel, listed assets.
- Livelihoods: land operated, crops, livestock, wage days, migration last 12 months.
- Core outcome module: your 8 to 15 indicators asked exactly as defined.
- Services: distance and use of school, health centre, bank, ration shop.
- Knowledge and practices: short, specific items, not agree-disagree statements.
- Observation: interviewer notes on water source, toilet use, respondent comfort.
Writing questions that survive to endline
Ask about behaviour in a fixed recall period rather than general habits. 'In the last seven days, on how many days did your child attend school?' is comparable over time; 'Does your child go to school regularly?' is not. Avoid questions that name the programme at baseline, because comparison households will not recognise it and programme households will start giving expected answers.
Keep out of the form
- Names in the analysis file
- Aadhaar or bank numbers
- Questions no indicator needs
- Double-barrelled questions
- Leading wording about the programme
Our view on baselines
Most Indian baselines now use tablets or phones because range checks, skip logic and GPS catch errors while the enumerator is still in the house. Paper still works for very small studies, but data entry adds weeks and a new error layer.
Paper questionnaire
Low setup, high cleanup
- Printing and transport of forms
- Skips and ranges checked only after entry
- Double data entry needed
- No proof of location or time
- Weeks between fieldwork and first tables
Offline mobile app (CAPI)
More setup, cleaner data
- Form built once, updated centrally
- Checks fire during the interview
- GPS, time and duration on every record
- Works offline, syncs later
- Daily progress and quality dashboards
Most failed impact comparisons trace back to decisions made at baseline, not at endline.
Design mistakes
- No comparison group at all
- Indicators changed between rounds
- Sample sized for the total, not sub-groups
- Baseline after activities started
Field mistakes
- No household listing kept
- Replacement households chosen by enumerators
- Different season at endline
- Untrained translators in the field
Data mistakes
- No stable household IDs
- Codebook never written
- Raw data held only by the vendor
- Limitations left out of the report
Not the right fit if
- Programmes already running for a year or more
- Very small pilots where a rapid assessment will do
- Questions only administrative records can answer
- Sites where a comparison group is impossible and no one will fund an endline
A good fit if
- New programmes with a planned endline
- Funders asking for attributable change
- Multi-district rollouts with phased entry
- CSR projects that will need a Rule 8(3) assessment later
Baseline and endline survey service
Sampling, tools and verified fieldwork for programme baselines.
Learn moreCSR impact assessment rules
When a company must commission an independent impact study.
Learn moreCAPI, CATI and CAWI explained
Choosing a data collection mode for your study.
Learn moreNGO survey solutions
Field data collection and evaluation for development programmes.
Learn moreSurvey cost in India
What drives the budget for a household survey.
Learn moreStraight answers to the questions people ask most about this topic.
How long does a baseline survey take?
For a few districts, expect roughly two to three weeks for design and tools, one week for translation, pilot and training, then fieldwork that depends on sample and travel, followed by two to three weeks for cleaning and reporting. Large multi-state baselines take longer. Fix the field window around seasons first.
What is the difference between baseline and endline surveys?
A baseline measures indicators before a programme starts; an endline measures the same indicators, with the same definitions and ideally the same households or clusters, after it ends. The difference between them, compared with the change in a comparison group, is the estimate of programme effect.
Do we need a comparison group?
If you want to claim that the programme caused a change, yes. Without one, a before-and-after difference also includes rainfall, prices, other schemes and general trends. Where a strict comparison is impossible, say so in the report and describe change rather than impact.
How many households should a baseline cover?
It depends on the smallest change you need to detect, the baseline level of the indicator, clustering and expected attrition. Two groups and many sub-groups raise the number quickly. Write the calculation in a sample size note so the endline team can check and repeat it.
Share the logframe, districts and timeline. We will propose indicators, a sample design and a field plan for the baseline.