Reading Interview Findings: Counts, Disagreement and What Is Still Unknown

Short answer

Read interview findings the way a sceptical committee would. For each finding:

• check how many people said it, out of how many were asked;
• read what the people who disagreed said;
• look for what nobody could answer;
• open at least two of the conversations behind it.

A finding that survives those four checks is worth acting on. A finding that rests on a single vivid quote, hides its disagreement or never mentions an unknown is a hypothesis, not a finding.

Four checks for every finding

Check the count

A count tells you how widespread a point is among the people asked. "7 of 10 people describe it" is strong within that group. It says nothing about the people who were not asked, so check who they were. Ten day-shift pickers agreeing is not evidence about nights. The guide to how many interviews are enough covers reporting small numbers.

Check, too, whether everyone was asked about that topic. Suppose a point came up unprompted in four conversations and was never raised in the other six. "4 of 10" then understates it. The accurate reading is "4 of the 4 who discussed it; the other six were not asked".

Read the disagreement

Disagreement is information. The two people who disagree may work at the site that does things differently, have a different contract, or have been there longer. Their account may point to the solution. A finding with no disagreement on a contested subject deserves a second look: either the question led, or the people asked were too alike.

Look for what is still unknown

Good findings say what they could not establish. "Nobody could explain why the recount takes forty minutes" is a useful unknown, because it tells you who to ask next. The Magenta Book's quality framework for qualitative evaluation asks evaluators to be transparent about the limits of their evidence. A findings note with no unknowns is either unusually complete or not being honest about its gaps.

Trace the source

Open the conversations behind a finding and read the quotes in context. Check that:

• the quote says what the finding claims it says;
• the person was describing their own experience, not repeating something they had heard;
• the follow-up questions did not lead them to the answer;
• the example is recent and specific enough to rely on.

The UK government's Uncertainty Toolkit is written for analysts, but its central point applies here: say how certain a conclusion is, so readers do not treat it as more certain than it is.

Separate the observation from the interpretation

"7 of 10 pickers describe empty bins where the system showed stock" is an observation. "Inventory accuracy causes late orders" is an interpretation. You can check the first against the conversations. The second needs other evidence, such as the system's own discrepancy records, before it becomes a reason to act. Keep them on separate lines of the findings note.

Question the findings

A findings note is where analysis starts. Useful questions to put to it:

A fictional example: reading a finding

A fictional retailer's findings note says: "Store managers ignore head-office priority lists because they arrive too late." The regional director checks it:

• Count: 6 of 11 store managers describe receiving the list after opening.
• Disagreement: 3 say they read it but find it irrelevant to their store; 2 did not mention it.
• Unknowns: nobody could say when the list is sent, only when it arrives.
• Source: the conversations show that four of the six who receive it late work in stores that open at 7 am, and the list goes out at 8 am.

The finding becomes two. First, the list arrives after opening for early-opening stores (6 of 11). Second, some managers find its content irrelevant (3 of 11). The next step is to check the send time, and to ask the three who called it irrelevant what they would want instead.

Where Lontra fits

Lontra writes findings to pass these checks. Each finding shows how many people said it, where they disagree and what is still unknown, and every point links back to the conversations behind it. Every quote is checked against what people actually said, and Lontra tells you when the evidence is not enough. You can put follow-up questions to your results, such as whether a finding differs by role, and get answers grounded in the conversations. Exporting is always free.

The interpretation and the decision remain yours. Lontra never decides anything about a person.

Act on what survives

Read every finding as a claim to test. Act on the ones that survive the count, the disagreement, the unknowns and the source. The rest tell you what to ask next.