Vyxora
Vyxora Blog

Research notes

How to Compare Customer Interviews and Write Findings

Compare interview evidence in a reusable matrix, write findings you can trace to participants, and keep conflicting feedback visible.

Three fictional interview accounts retain their different conditions before a qualified finding is written
Fictional interview exercise. Similar comments can have different conditions; a finding must preserve those differences.

Compare interviews around one question, keep each observation attached to its participant and source, and write a finding that says when it holds. Use the matrix and memo below to separate what people described from your interpretation and proposed next steps.

Download the three transcripts, blank matrix, and filled matrix. The worksheet and completed memo are available without Vyxora.

Start with the finding you need to explain

All three interviews are fictional. Maya, Arun, and Lena were written for this exercise, not interviewed as customers. The records demonstrate a practical comparison method; they are not evidence of market demand or population prevalence.

The question is: What information do people need when returning to work discussed in a client call? Here is a finished memo you can trace through the records below.

Customer follow-up findings memo

Question and scope

What information do people need when returning to work discussed in a client call? Reviewed three fictional sessions: P01 Maya, independent consultant, November 3; P02 Arun, studio delivery lead, November 4; P03 Lena, operations coordinator, November 5, 2026.

Finding: dates need their conditions

Maya checks the conditions around dates on new projects. Arun describes a conditional deadline being read as a promise when its approval condition disappears.

Support: P01.3–P01.4 and P02.5.

Qualification: Maya's ten-minute example describes one familiar weekly check-in. Lena was not asked directly about date conditions; her silence is not agreement or disagreement.

Finding: proposed ownership needs confirmation

Arun describes a recap that did not confirm who had accepted an action, then qualifies his initial request for automatic assignment. Lena also checks accepted ownership in external partner handoffs.

Support: P02.2–P02.4 and P03.4.

Qualification: Lena's existing internal task board works for weekly meetings. Maya uses a personal list and did not describe problems confirming who accepted shared work. Keep those contexts distinct.

Interpretation and next questions

Preserving date dependencies and confirming shared ownership are useful things to test in these workflows. Ask Maya to show how she resolves a conditional date, Arun to demonstrate owner confirmation, and Lena to compare an internal meeting with a partner handoff.

These are proposed follow-ups. The exercise does not establish frequency, willingness to pay, or whether a generated summary solves the problem.

Choose a question before building the matrix

A question gives you a reason to compare a passage. Without it, every vivid comment can become a separate theme, and unrelated workflows can be flattened into one supposed need.

For this exercise, P01 concerns one person's client follow-up. P02 concerns shared delivery. P03 distinguishes internal work from external handoffs. That context matters before you compare their tools or requests.

If you are still preparing a single interview, use the earlier transcript-analysis walkthrough. This article starts after you have records from several sessions and need to explain their similarities and differences.

Use only records you have permission to process. The UK government's guidance on recording user research advises keeping notes tied to observations, using participant and session labels, and protecting personal information. Your consent, access, and retention obligations still apply when software helps with analysis.

Put one observation on each row

Open the CSV in Excel, Google Sheets, or another spreadsheet app. Keep the source locator on the same row when sorting or filtering. Use the Markdown worksheet if you prefer a document. Add rows rather than compressing several different observations into one cell.

The full matrix includes separate fields for observation, analyst interpretation, qualification, and follow-up. This shorter view shows the comparison:

Participant and contextSourceObservationQualification
Maya, familiar weekly check-inP01.2–P01.3One follow-up took about ten minutesNot a claim about all client follow-ups
Maya, new projectsP01.3Replays the conditions around a dateA different situation from the weekly check-in
Maya, personal task managementP01.4Keeps her own tasks in a short listDoes not describe problems confirming who accepted shared work
Arun, shared studio deliveryP02.2Recap did not confirm who had accepted an actionDoes not establish that the action was missed
Arun, initial feature requestP02.3Suggests automatic assignmentQualified by P02.4
Arun, ownership confirmationP02.4Rejects assignment based only on a mentionWants the proposed owner to confirm
Arun, client approvalP02.5An omitted condition makes a date read as a promiseHis account, not a measured failure rate
Lena, internal meetingsP03.2Existing task board worksDoes not request another task list
Lena, decision historyP03.3Keeps a note explaining why a decision changedCurrent task and decision rationale serve different needs
Lena, partner handoffsP03.4Asks who accepted an external actionP03.5 separates this from the internal process

A short quotation is useful when its wording matters, but read the surrounding passage before lifting it. P02.3 says the software should assign tasks automatically. P02.4 narrows that request: being mentioned should not create an assignment. A matrix containing only the earlier sentence would misrepresent this participant.

Compare the conditions, not just repeated words

Maya and Arun both discuss dates, but their circumstances differ. The supported comparison is about preserving a dependency when returning to a discussion. It is not “all customers struggle with follow-up.”

Arun and Lena both discuss accepted ownership, but Lena's example concerns an external partner. Her internal meetings already update a task board. Keep the internal row even if it weakens an attractive story about replacing task tools.

Do not treat an empty cell as “no problem.” Lena was not asked directly about date conditions. Write not asked. If a transcript is incomplete, write not recorded and identify the missing material. The absence of a statement is not a negative response.

Write the finding, support, and boundary together

Use this structure in a research memo or a section of your product-planning document. Replace every bracketed field with your own evidence, interpretation, and scope. Add another finding only when you can trace it just as clearly.

Customer interview findings template

Question and records

[Research question]. Reviewed [participant/session IDs and dates]. Missing material: [Any gaps]. Relevant limits: [Contexts and sample limits].

Finding

[A specific claim no broader than the evidence].

Support

[Participant IDs, source links, and passages or timestamps].

Qualification

[Conditions, corrections, differing accounts, or questions not asked].

Interpretation

[What you think the comparison suggests, separately from participants' words].

Next question or action

[What to investigate or test next. Do not present a proposed step as an agreed commitment].

Trace each interpretation in the memo to the interviews you reviewed. Repeated mentions can help you notice a question worth investigating; three fictional accounts cannot tell you how common a need is or whether anyone would buy a solution. Formal qualitative research also requires a method appropriate to its question, ethics, and claims.

Use software to retrieve the passages you need

If you use Vyxora, keep the interviews as distinct Notes so each account remains identifiable. A Workstream can bring related Notes and files together for source-backed questions. Ask which passages support a proposed finding, then open them and check the conditions and differing accounts yourself.

Full notes on an individual interview Note can provide additional explanation, but generated prose is not a new participant statement or independent evidence. Preserve the original records and distinguish your analysis from their wording. Full notes source boundaries, checked November 17, 2026.

Reuse the comparison resources

The original fictional transcripts, comparison resources, and memo are licensed under CC BY 4.0. Attribute the relevant resource title to Vyxora Editorial, link to this article and the license, and state whether changes were made. The license excludes the Vyxora mark, product UI, software, and third-party material. Keep that attribution beside copied CSV data or in the document where you reuse it.

Before sharing real findings, check each source reference, review the accounts that differ, and remove identifying material the recipient should not receive. Finish by naming the next question the evidence leaves open.

How this was made: AI assisted with research and drafting. Vyxora Editorial reviewed the sources, product claims, and final wording.

Sources and product details checked November 17, 2026.