“How many people should we interview?” is one of the first questions in any research project. Too few, and the results swing by chance. Too many, and you spend more than you need. Here is a plain-language way to decide.
Start with the margin of error
The margin of error tells you how close your survey result is likely to be to the true figure. For a simple random sample at 95% confidence, the approximate margins are:
| Completed interviews | Approx. margin of error |
|---|---|
| 100 | ±10% |
| 400 | ±5% |
| 1,000 | ±3% |
| 2,500 | ±2% |
| 10,000 | ±1% |
Notice that accuracy improves quickly at first and then slowly. Going from 100 to 400 interviews halves the error; going from 2,500 to 10,000 only improves it by one point.
Then think about sub-groups
If you want to compare cities, age groups or customer types, each group needs enough interviews on its own — usually at least 100, and ideally 200 or more. A study comparing five cities therefore needs 500–1,000 interviews, not 400.
When very large samples make sense
Tens of thousands of interviews are justified when results are needed at ward, block or district level — for example utility surveys, beneficiary verification of government schemes, or state-wide programme evaluations. ARS can complete over 1,00,000 interviews in 15 days for such studies.
A note on real-world samples
The figures above assume random sampling. Field surveys that use clusters or quotas usually need somewhat larger samples for the same accuracy. A good research partner will explain this in the sampling plan.
Need help sizing your study? Send us your requirement or call +91 97899 76662 — our research consultancy starts from ₹5,000.