Community context and needs diagnosis · Lesson 3

Triangulating evidence and counting units

4 min read

Lesson objectives

  • Triangulate interviews, observation and service records with compatible denominators; output an evidence matrix and uncertainty statement.
Reading tools

Triangulation checks the same question through evidence with different strengths and blind spots; it is not an average of three numbers. Service records may show a backlog, interviews may reveal that non-users do not understand registration, and observation may show an obscured help sign. These sources do not measure the same thing. A register counts transactions; an interview describes an experience; an observation records behaviour at a particular place and time. Before comparing them, give each source a row with its counting unit, population covered, period, collection method, supported claim and unsupported claim. If the unit or period differs, do not combine percentages. Use the difference to formulate a verification question.

Name the intended claim precisely. “Many residents cannot get in” requires evidence about residents and non-users. “Uncompleted requests rose in the centre's register” requires a register with a stable definition across two periods. Check completeness and duplication. One person may submit three requests, and a request may be closed and reopened. Fifteen late requests out of 60 means 25% of **requests** were late; it does not mean 25% of people experienced delay. Before comparing with another month, confirm that “late” and the logging process mean the same things. A changed data entry system can create an apparent increase without a service change.

Qualitative interviews should not usually be used to estimate population prevalence unless a sampling design suitable for that question has been used. Here their value is explaining how a barrier works and how experiences differ. Ask for a specific episode: “What happened the last time you tried to book?” is more useful than the leading question, “Isn't registration confusing?” Record cases that do not fit the emerging explanation. They may show that non-users face a different obstacle. When sources conflict, do not automatically choose the most official or vivid source. List plausible explanations and a way to test each. This is a decision method for this lesson, not a numerical rule attributed to an external authority.

Lesson summary

Community context and needs diagnosis: Triangulate interviews, observation and service records with compatible denominators; output an evidence matrix and uncertainty statement.