Complaint email task
A diagnostic issue-identification technique used to uncover pain points, unmet needs, and improvement priorities. It asks respondents to write as if emailing the company’s complaints or feedback team — encouraging direct, detailed, and honest descriptions of what’s wrong, what’s missing, and what needs fixing.
The qualitative theory
Section titled “The qualitative theory”The complaint email is a well-established projective technique in qualitative research. Rather than asking customers “what problems have you had?”, it asks them to adopt the role of someone writing a formal complaint — which changes both what they say and how they say it.
The framing does several things at once. Writing to a complaints team legitimises the complaint: participants are not criticising the brand out of nowhere, they are doing exactly what the situation calls for. This lowers the social inhibition that often leads dissatisfied customers to soften their feedback or qualify their criticisms. It also encourages specificity — a real complaint email would be detailed, concrete, and include everything relevant — which produces more actionable data than abstract evaluations like “the service was poor”.
“What is the main issue you would want them to understand and fix?” focuses this energy. Rather than cataloguing every grievance, participants are asked to single out the issue that matters most — a built-in prioritisation that concentrates the conversation on the problem with the greatest weight for them.
The laddering sequence that follows is deliberately short and focused. The opening question establishes the priority issue; the two laddering questions then contextualise it (why does it matter?) and close with the solution (what should the company do?). Together, these produce not just a catalogue of complaints but a prioritised account of what needs to change and why.
The “friendly and empathetic” tone is not incidental. Complaint research requires a particular kind of psychological safety — participants need to feel that their frustration is being heard and taken seriously, not evaluated or justified. An empathetic register signals that the conversation is on the participant’s side, which produces more candid and more complete disclosure.
Configured settings
Section titled “Configured settings”This section documents every setting in the methodology so you can understand the intent behind each choice.
Research framing
Section titled “Research framing”| Field | Value |
|---|---|
| Context & Background | A diagnostic issue-identification technique used to uncover pain points, unmet needs and improvement priorities. Use for NPS Detractors, dissatisfied customers or when a product/service is known or suspected to have issues. It asks respondents to write as if emailing the company’s complaints or feedback team, encouraging direct, detailed and honest descriptions of what’s wrong, what’s missing and what needs fixing. |
| Research Objectives | Identify the main problems customers experience, understand why they matter and reveal where the product/service is falling short of expectations. The exercise also helps pinpoint which issues feel most urgent and what customers want fixed or improved. |
| AI Moderator Tone | Friendly and empathetic |
Friendly and empathetic is the only tone in this methodology library that explicitly names empathy. In a complaint context, participants are often frustrated, let down, or emotionally invested in what went wrong. An empathetic register signals that those feelings are valid and that the conversation is genuinely on their side — which is what makes honest, detailed complaint disclosure possible.
Conversation introduction
Section titled “Conversation introduction”Hi, I’m
{{companionName}}and I’m your survey companion today.Imagine you are writing an email to the complaints team about our products and services. What is the main issue you would want them to understand and fix?
The introduction establishes the complaint email frame immediately after the companion introduction. It is written generically — “the complaints team” and “our products and services” — so the methodology works for any company without editing. Two deliberate signals do the work:
“Imagine you are writing an email to the complaints team” — adopts the projective register. Participants are not being asked to evaluate the brand; they are performing a task that naturally elicits the specificity and directness of a real complaint.
“What is the main issue you would want them to understand and fix?” — asks participants to single out the one issue that matters most, building a prioritisation into the opening and setting up the two laddering questions that follow.
Laddering questions
Section titled “Laddering questions”Two questions, with Randomise off and Paraphrase off:
| # | Question | Purpose |
|---|---|---|
| 1 | What makes that issue most important to you? | Contextualises — surfaces the meaning behind the priority, connecting the specific problem to a personal need or expectation |
| 2 | What do you think the company should do to fix this? | Closes with the solution — surfaces what the participant actually wants, which is the most actionable output of the conversation |
Randomise is off because the two questions follow a deliberate order: the first explains why the priority issue matters, and the second builds on that to surface the resolution the participant wants. Asking what the company should do (Q2) only makes sense once the participant has explained why the issue matters (Q1).
Paraphrase is off for the same reason. The specific phrasing of each question is matched to its analytical purpose — particularly Q2 (“what do you think the company should do to fix this?”), which asks for a concrete resolution rather than further description.
Skip remaining questions on non-response is on. If a participant does not engage with a question, the AI Moderator will not continue pressing through the remaining questions — keeping the conversation natural and avoiding the sense of an interrogation when someone has little more to add.
Default probe
Section titled “Default probe”Tell me more about
{{topic}}?
Paraphrase on. Open and non-directive — gives the AI Moderator room to follow up on whichever complaint dimension the participant has raised without steering the content of the response.
Default prompt
Section titled “Default prompt”What about
{{topic}}, please tell me about that?
Paraphrase on. Kept neutral and open. Note that prompting is currently disabled across all topics — the default prompt is configured but will not fire unless you enable prompting on specific topics.
Probe & prompt limit
Section titled “Probe & prompt limit”2 — a focused limit that keeps the conversation short and centred on the single issue the participant chose to raise. Rather than casting wide across many topics, this methodology goes deep on what matters most, and a limit of 2 gives the AI Moderator enough headroom to probe that issue meaningfully without extending the conversation unnecessarily. This is a conversation-wide cap: even though individual topics carry their own probe allocations, no more than two probes will fire across the whole conversation.
Topics
Section titled “Topics”| Topic | Description | Probes | Prompts |
|---|---|---|---|
| Expectations not met | Probe for what the respondent expected instead and how the experience fell short of that expectation | Off | Off |
| Consequences | Explore what happens if the issue continues, to reveal risk to loyalty, trust, or future use | Off | Off |
| Features | Explore further any problems with specific product or service attributes, functionality, or design | 2 | Off |
| Behaviour and actions | Explore the behaviours or actions that are problematic or issues | 1 | Off |
| Usage experiences | Explore more about how the product is used that leads to issues or problems | 1 | Off |
| Loyalty/advocacy effect | Probe whether the issue influences future use, trust, retention, or likelihood to recommend | 1 | Off |
| Emotions | Explore how this makes customers feel | 1 | Off |
| Communication issues | Problems with how the company conveys information, responds to queries, or manages expectations | Off | Off |
| Resolution process | Feedback on the effectiveness, efficiency, or fairness of the complaint or resolution process | Off | Off |
| Value for money | Concerns about whether the product or service provides adequate quality or benefit relative to its cost | 1 | Off |
Randomise probe order: on. Randomise prompt order: on.
You choose which topics are prompted if not raised unprompted: Prompting is currently disabled by default for all topics, so you can decide which prompts are relevant to your specific survey or subject. Review the topic list before publishing and enable prompting only for the topics you want to ensure are covered if respondents do not raise them spontaneously.
Features carries two probes — the highest allocation. Product and service features are typically the most granular and multidimensional complaint dimension, and two probes give the AI Moderator room to explore both the specific attribute that failed and the nature of the failure.
Expectations not met, Consequences, Communication issues, and Resolution process all have probing currently disabled. Turn these on if you want to understand more about how an issue negatively impacts on customer experience.
Both randomise probe order and randomise prompt order are on — complaint conversations are unpredictable in sequence. Participants raise issues in whatever order feels natural to them, and randomising ensures no single topic dimension is systematically followed up before others across the sample.
Conversation ending
Section titled “Conversation ending”- Closing question: None
- Dynamic ending: On
There is no closing question. The final laddering question (“What do you think the company should do to fix this?”) already provides a natural, solution-oriented close to the structured phase of the conversation. A further closing question would add length without adding insight. Dynamic ending allows the AI Moderator to wrap up in a way that feels responsive to what was discussed.
Adapting this methodology
Section titled “Adapting this methodology”When you apply this methodology to a new conversation, these are the fields you’ll typically need to change:
- Conversation introduction — the introduction is written generically and can be published as-is. Optionally tailor it by naming your company, brand, or a specific product or service if you want it to feel more concrete to participants.
- Context & Background — describe the specific product, service, and participant group. If you are targeting NPS Detractors specifically, note the score range.
- Research Objectives — update to reflect the specific complaint dimensions most relevant to your study. If the research is focused on a known issue, name it in the objectives so the AI Moderator has the right context.
- Topics — enable probing on additional topics if you need systematic coverage of specific complaint dimensions. Consider enabling prompting selectively if you need guaranteed coverage of a dimension (such as Value for money or Consequences) that might not surface naturally.
- Probe & Prompt Limit — the default of 2 keeps the conversation focused and short. Scale up if you enable more active topics or want the AI Moderator to explore several issues in depth.
The complaint email framing in the introduction and the two-question laddering arc are the structural core of this methodology and should be left unchanged wherever possible.
Frequently asked questions
Section titled “Frequently asked questions”Why ask people to imagine writing an email rather than just asking what went wrong?
The complaint email frame does several things that a direct question cannot. It legitimises the complaint — writing to a complaints team is exactly what the situation calls for, which removes the social inhibition that leads dissatisfied customers to soften their feedback. It encourages specificity and completeness — a real email would be detailed and cover everything. And it removes the pressure to be “balanced” — you would not moderate your criticism when writing to a complaints team. The result is richer, more direct, and more actionable complaint data than open-ended evaluation questions typically produce.
Why are prompts disabled on all topics?
Prompts are disabled by default so you can choose which topics are relevant to actively prompt in your specific study, if any. If you need guaranteed coverage of specific dimensions across the sample, enable prompting selectively.
Why is the tone “friendly and empathetic” rather than the more common “friendly and casual”?
Complaint research is emotionally loaded. Participants are frustrated, let down, or genuinely upset — and they need to feel that the conversation is on their side before they will disclose honestly. Empathetic signals that the AI Moderator understands this context and takes it seriously. A casual tone in a complaint setting can feel dismissive; a clinical or neutral tone can feel like an audit. Empathy creates the psychological safety that makes full and honest complaint disclosure possible.
Why does the final question ask what the company should do to fix the problem?
Complaint research that stops at identifying and prioritising problems misses its most actionable output: what customers actually want done. The final laddering question is not a creative ideation question — it is asking participants to articulate the resolution they feel is owed to them. These responses often reveal whether the expected fix is operational (a refund, a repair), systemic (a policy change, a process improvement), or relational (an apology, better communication). That distinction is critical for deciding what to do with the insight.
Can this methodology be used proactively, before complaints arise?
Yes, with some adaptation. The complaint email frame works best when participants have genuine grievances to raise — it is most powerful with NPS Detractors or known dissatisfied groups. For a broader audience that includes satisfied customers, the framing can be softened (“imagine you were writing to the feedback team about anything you’d like to see improved”) — but this changes the data you get. The more the framing invites complaint, the more honest and complete the output will be.