How to summarize interviews - 5 steps from expert call notes to client-ready insights

You've just finished 30 expert network calls through GLG, Inex One or AlphaSights. Or conducted 40 client interviews for a market study. Or had 30 calls with employees from your client's organization across the world. The transcripts are sitting in a folder. The synthesis deck is due Monday.

How do you turn 600 pages of interview notes into actionable insights without drowning in the details?

After conducting over 1,000 business interviews during my 18 years at McKinsey, I've developed a systematic approach that works whether you're doing due diligence, market research, or strategic analysis. The key is treating interview synthesis like mining: extract the obvious gold first, then systematically process everything else.

Here are the five steps that I would recommend

Step 1: Extract the gold nuggets first

Don't start by reading everything sequentially. Start by capturing the obvious, critical insights that jumped out during the interviews themselves.

Immediately after each interview, while still fresh, write down 3-5 synthesis statements that matter. Use the Synthesis→Evidence structure:

Synthesis:

"Industry expert sees 2026 will be a bloodbath in the AI wrapper space"

Evidence:

• "Everyone is burning cash on AI tokens and unit economies don't add up"
• "Basic LLM models are adding features so fast that wrapper companies can'd differentiate"
• "Most companies are running out of runway in 6 months given VC money is getting in spotting poor-quality AI"

This depends on taking the approach I covered in Step 4 of conducting effective business interviews - meaning that you recognize during the discussion when you've hit gold, slow down and really start to dig deeper to get the best quotes. The most valuable insights are usually obvious during the interview itself. Someone says something that changes your entire thesis. That's gold. Extract it immediately before it gets buried in 50 pages of transcript.

Time investment: 10 minutes per interview, right after it ends. Send these notes around the team as well to keep everyone on the same page, for example to help others on what to probe in their discussions (to prove or disprove the gold nuggets found this far).

Step 2: Sort the tailings - identify your themes

Now comes the more rigorous part. You need to develop your thematic structure - the buckets that will organize everything you heard.

Read through all your materials (or at minimum, skim each transcript and your notes). Create a working document where you list themes as they emerge. By the end, you should have 20-30 distinct themes grouped into 3-6 major categories. To accomplish this, you need to add, merge and delete categories as you proceed through your reading list. If you were not the one conducting the interviews or it's been a while since you did them, it might take two rounds to do this step: first time just skim them through to remind yourself of the overall context, and only second time start penning down the categories.

Example for market sizing project:

Category: Market Structure

• Market size estimates and definitions of market
• Growth trajectory disagreements and rationale
• Geographic differences (US vs EU vs APAC)
• Segment definitions

Category: Competitive Dynamics

• Comments on individual companies
• Startup opportunities vs. incumbent advantages
• Pricing pressure points
• Partnership vs build decisions

Category: Customer Buying Process

• Decision-maker identification
• Budget cycle timing
• Evaluation criteria

Discontinuities

• Next generation technology
• End user evolving needs
• Geopolitical changes

Category: Regulatory & Risk

• Compliance requirements
• Data privacy concerns
• Upcoming regulation impact
• Certification needs

Don't try to make themes perfect at this stage. You're creating a sorting system, not the final deliverable structure. Themes can overlap. You'll refine them later. A common mistake is trying to force everything into 5 perfect categories too early. Better to have 25 messy themes you'll consolidate than miss important topics because they didn't fit your predetermined structure.

Time investment: 2-3 hours for 30 interviews.

Step 3: Search for "gold dust" per theme

This is where you systematically go through your materials again, but now with your theme structure in hand. For each theme, copy the relevant insights, quotes, and supporting evidence.

I use a simple document structure:

Theme: Market size estimates

"We model TAM at $15B based on Gartner projections, but honestly those numbers are fantasy. Our actual serviceable market is probably $2-3B when you exclude enterprise customers who'll never switch from SAP." Interview 12 - CFO, Series C SaaS company:

"Official market size is $12B, growing 23% CAGR. But that includes a lot of shelfware. Real spending on actively-used solutions is half that." Interview 7 - Industry analyst:

"We do $400M annually. Our three main competitors are roughly the same size. That's $1.6B right there, and we're probably 60% of the real market." Interview 23 - VP Sales, incumbent vendor:

Unlike academic research where you need comprehensive coding of every passage, business analysis follows the 80/20 rule. Capture the representative quotes and insights for each theme. You're building evidence, not conducting grounded theory research. Don't worry about insights or synthesis yet, just make sure you have the stuff sorted. Pro tip: Use different colors or tags or excel columns to mark quotes by stakeholder type (customer, competitor, analyst, practitioner). This makes cross-case analysis easier later.

Time investment: 2-3 hours for 30 interviews with good transcripts.

Step 4: Melt the gold to trophies - translate directly to deliverable format

Now, look at what categories are relevant for your overall storyline and focus. Often there are areas that are interesting (e.g., next generation technology) and that experts keep bringing up, but if you're trying to understand cash burn for next year, they are irrelevant.

Don't create an intermediate summary document that you'll later need to translate again.

If you're creating a due diligence deck, write the actual slide titles and bullet points. If you're writing a market research report, draft the actual section content. If you're preparing an executive brief, write the brief itself.

Once clear on the storyline and structures, ttranslate your theme notes directly into your end deliverable format and think of the so-what / insights. In the market size estimate above, you could conclude that Official market sizing ($12-15B) is 3-5x higher than reality. True market is $2-4B based on actual vendor revenues and customer spending.

Example: Due Diligence Deck Slide

Slide Title: Market Size Reality Check - TAM Claims Don't Match Evidence

Bullets:

• Management claims $15B TAM based on Gartner projections
• Reality: True addressable market is $2-4B based on vendor revenues and expert input
  • Top 4 vendors = $1.6B combined revenue (interview with incumbent VP Sales)
  • Customers report "official" estimates include 50%+ shelfware (industry analyst interview)
  • Enterprise segment (40% of Gartner TAM) unwilling to switch from SAP/Oracle (CFO feedback)
• Implication: Revenue projections assuming 5% market share ($750M) are unrealistic. The sell side case overstates market opportunity by 3-5x

Notice how the theme notes from Step 3 translate directly into slide content, with specific evidence in backup or footnotes.

Don't try to use every theme. Focus on the storyline-relevant themes first. If a theme or quote doesn't support a key decision or insight, save it for the appendix or skip it entirely.

In a typical project, 60% of the themes I identify in Step 2 make it into the main deliverable. The other 40% are either background context or interesting-but-not-critical findings. Don't waste time polishing fool's gold!

Time investment: 3-5 hours to draft main deliverable from organized themes.

Step 5: Add the base - integrate numbers, graphs and facts to make it solid

The final step is integrating your qualitative insights with quantitative backings. This is where interview synthesis becomes truly powerful.

You can get quantitative data from a range of sources:

Frequency counts from interviews:

• "75% of customers mentioned pricing as a top-3 concern" (15 of 20 customer interviews)
• "Only 2 of 12 enterprise buyers had heard of [Startup X]"
• "All 8 technical evaluators mentioned integration complexity"

Note: keep in mind that these numbers should be taken with a grain of salt. Not mentioning something doesn't mean it's not relevant for the specific person - maybe you just omitted the question or they didn't find the right time to answer. That is why in academic research this practice is not really used.

Numbers from the experts:

• "Experts estimated 3 to 5x size in three years"
• "When asked to allocate 100 points to these priorities, on averager 40 points went to priority 1"

Corroboration with secondary data:

• Expert says "market is consolidating" → Add chart showing M&A activity doubled from 2022-2024
• Customer complains about price increases → Pull competitor pricing from public filings
• Analyst predicts regulatory changes → Link to actual pending legislation

Reconciliation of conflicting views: If 10 experts say the market is growing and 5 say it's declining, don't just average them. Look at the pattern:

• All 5 pessimists are incumbents facing disruption
• 8 of 10 optimists are in high-growth segments
• Geographic split: US optimistic, EU cautious
• Synthesis: Market is bifurcating - legacy segments declining, new segments growing rapidly

Create simple visuals: You don't need fancy graphics. A simple table showing stakeholder perspectives is powerful:

Time investment: 2-3 hours to add quantitative backup and create simple visuals.

Love the output... but isn't there a faster way?

The approach above works. I used it successfully many times at McKinsey. But it's still fundamentally manual and time-intensive.

The alternative - just slapping a few quotes to the slides - feels like a waste of all the interviews done and also seriously undermines the quality of the answer. But I have seen in many business settings that quality analysis is the first victim of time pressure.

AI becomes a gamechanger here. Done smartly, you can get extremely high quality analysis with record speed. But this requires a thoughtful approach, not just trying to drop all 30 transcripts to ChatGPT with a plea to "analyse them"... If you try to do that, you will get mediocre answers, no ability to verify for accuracy, and huge issues with client confidentiality.

We built Skimle as a tool to help consultants analyse and structure large sets of data.

Skimle automates Steps 2-3 (the mechanical work) while keeping you in control:

Automated transcription: Upload audio/video from Zoom, Teams, or phone calls. Skimle transcribes automatically in 100+ languages.

Automatic theme identification: Instead of manually reading 30 transcripts to develop themes, Skimle processes everything and suggests 20-30 themes grouped into categories. Every theme links to specific quotes.

Systematic organization: See all 30 interviews in a spreadsheet-like view. Each theme shows which interviews mentioned it and the relevant quotes. No more "I remember someone said something about pricing but which interview was it?"

Editable structure: Don't like the AI's suggested themes? Merge them, split them, rename them, or add your own. You control the final structure.

Direct export: Export organized themes with quotes directly to your deliverable format.

What used to take days hours takes 2-3 hours with Skimle. You still do Steps 1, 4, and 5 where human judgment is essential, Skimle helps set you up for success and do some of the grunt work.

Try Skimle for free - upload your interview transcripts and see how AI-powered systematic analysis compares to manual note-taking.

Contact us if you're a consulting firm, research team, or organization that regularly conducts expert interviews and wants to discuss team licensing.

Olli from the Skimle team