Attrition occurs when participants or units are lost during a study. Someone may stop answering a diary, miss a follow-up, or leave before an outcome is recorded. The rate describes the amount of loss; the reasons and pattern determine how much concern it raises for the findings.

If 20 of 100 enrolled participants are unavailable at a defined follow-up, attrition at that point is 20%. State the starting group and time point, since a rate from the previous wave answers a different question from cumulative loss since enrolment.

Examine who is missing

In an illustrative study of a new setup process, participants who found it most difficult might be least likely to complete the follow-up survey. Analysing only respondents could then present an unduly favourable account.

J-PAL’s implementation-monitoring guidance identifies attrition as a threat to evaluation. In an experiment, compare loss across conditions and examine available characteristics associated with it. Equal attrition rates do not prove that the same kinds of people are missing from each group.

A smaller sample also reduces precision, even when the loss does not create systematic bias. More initial recruitment may protect numbers, but it does not resolve selective loss by itself.

Reduce avoidable loss without pressuring participation

Set realistic demands, provide clear reminders, test the technology, and make incentives and contact arrangements dependable. A longitudinal study needs these arrangements across the full period, rather than only at enrolment.

Participants must remain free to stop. Investigate practical barriers and offer appropriate support without making continued participation feel compulsory. Record reasons when people freely provide them, but do not treat an unknown reason as evidence of a particular attitude.

Make the analysis assumptions visible

Report the flow of participants and the availability of each outcome. Consider suitable missing-data methods or sensitivity analyses with methodological support; each approach has assumptions, and filling gaps does not recreate observations that were never collected.

Keep attrition distinct from customer churn. A customer may leave the product and still complete the research, or remain a customer while dropping out of the study. Preserving that distinction helps the team understand what the missing evidence means for the decision.