Signal & Noise

How many AI-moderated interviews are enough? Saturation evidence, 4 rules for sizing AI interview studies, and when 200 interviews beat 20 for market research.

Fake respondents now reach qualitative studies too. 8 warning signs of bots and fraudsters in AI-moderated interviews, and how to screen, detect and clean them.

A 7-point quality check for AI-generated research insights: verify quotes, coverage, counter-evidence and claim scope before findings reach the client.

How to run laddering interviews and analyse them: means-end chain theory, interview probes, coding attributes, consequences and values, and drawing a value map.

How to analyse concept test interviews: a 5-step method for coding reactions, comparing concepts and segments, and writing a clear go/no-go recommendation.

A qualitative topline report template and 24-hour plan: what to include, how to structure findings and quotes, and how to deliver it the day after fieldwork.

How to analyse brand tracker open-ended responses wave after wave: a stable codebook, consistent coding, theme trends over time and the why behind KPIs.

How to run and analyse willingness-to-pay interviews: value drivers, reference prices, objections, and how qual explains Van Westendorp and conjoint results.

Qualitative research for CPG and FMCG brands: 6 core studies from shopper interviews to post-launch reviews, and how AI interviewing and analysis change each.

New 2026 research confirms AI 'digital twins' fail at individual-level prediction and segment targeting. Here's what synthetic respondents can and can't do.

Customer insights research explains why customers behave, switch or churn. See how it differs from market research and what data and analysis it takes.

Most open text analysis tools stop at sentiment or category buckets. Here's how inductive theme discovery finds the patterns predefined categories miss.

How to analyse in-depth interviews (IDIs): within-case and cross-case coding, sample size, quote selection, and a worked example from a churn study.

Focus group transcription is harder than 1:1 interviews: crosstalk, speaker attribution, accuracy rates. How to transcribe, format and check a group transcript.

Most commercial qualitative research is built top-down: skim for themes, find quotes to fit. Academic research does it bottom-up. AI now makes that affordable at deadline.

Selling AI-assisted research on speed alone leads to shorter projects and thinner margins. The better pitch is what teams do with the time AI frees up.

Nobody accepts a black-box result from a spreadsheet. Why do we accept one from AI-assisted qualitative analysis? A case for showing the steps, not just the summary.

AI tools can show you the quote behind a finding. They can't show you what they never looked at. Here's why that gap decides whether a client trusts your research.

Clients expect AI to make research cheaper and faster. Agencies that compete on that basis commoditise themselves. Here's the alternative: compete on depth instead.

Fragmented VoC data produces contradictory findings, slows decisions, and lets the loudest team win instead of the best evidence. Here's how to size the real cost.

A vendor-neutral evaluation framework for qualitative analysis software: 7 scoring domains, the questions to ask, a pilot protocol, and the procurement annex.

Eight small research projects a customer insights team can run in days, using data you already have or can collect quickly, and publish as thought leadership.

Responsible AI in qualitative market research: 7 rules covering where AI helps, why synthetic respondents fail, and what human-in-the-loop really requires.

A line-by-line cost model for a qualitative study: recruitment, incentives, moderation, transcription and analysis, with 2026 rates and a worked example.