To synthesise expert network calls: structure each call note with consistent metadata (expert background, topics covered, direct quotes), then code across the full set for themes rather than reading sequentially. Prioritise divergence over consensus, because the sharpest signals are often where experts disagree. Translate findings into a deliverable that maps claims directly back to source calls. Tools like Skimle can process 20–30 call transcripts in minutes and surface cross-call patterns that sequential reading misses.
Expert network calls are expensive. A single call through GLG, AlphaSights, or Guidepoint typically costs $500–$2,500 (€460–€2,300), and a commercial due diligence engagement or strategy project might run 15 to 30 of them. The total spend on primary research for a single CDD assignment can easily reach $30,000–$60,000 (€27,000–€55,000).
Most of that investment is wasted at the synthesis stage.
Not because the calls were poor. Experienced analysts know how to run a good expert call. The problem is what happens after: notes sitting in individual documents, a rushed readout meeting, and a senior person writing a synthesis paragraph from memory two days before the deck is due. What ends up in the deliverable is not a faithful representation of what 20 experts collectively said. It is a reconstruction shaped by recency bias, by which calls the synthesis author happened to read most recently, and by which findings aligned most comfortably with the existing hypothesis.
This guide covers how to do expert call synthesis properly: from structuring your notes before the first call begins, through coding and divergence analysis, to writing a deliverable that can survive an investment committee challenge.
Why sequential note-reading fails
The most common synthesis approach is also the worst one: read all the call notes in sequence, take notes on interesting observations, and write a summary. It feels systematic. It is not.
Sequential reading produces three well-documented cognitive biases that corrupt expert call synthesis.
Recency bias. The experts you read last have disproportionate influence on your summary. If the last three calls were bearish on the market thesis, your synthesis will read bearish, even if twelve of the fifteen calls were constructive.
Confirmation bias. When you already have a hypothesis about the investment thesis, you will unconsciously weight evidence that confirms it. This is not dishonesty. It is a feature of how human memory works under time pressure. You read 300 pages of call notes and your brain retains the passages that fit the pattern you expect.
The loudest voice problem. One particularly articulate expert, or one memorable phrase that appeared in the third call you read, ends up anchoring the narrative. This happens because language sticks. "The moat is an inch wide and a mile deep" is a quotation from one call. It does not represent the collective view of your expert set. But it will end up in the deck.
The alternative is to treat expert calls as structured data and apply the same analytical discipline to synthesis that you would apply to any other part of the diligence process. That means coding, not reading. And it means systematic divergence analysis, not consensus-hunting.
A well-run expert call programme is a form of qualitative primary research, and the synthesis methodology should reflect that. The thematic analysis guide covers the analytical foundations; what follows is the application to an expert call programme specifically.
How to structure call notes before analysis
The single most valuable thing you can do to improve synthesis quality costs nothing and takes five minutes per call. Before your first call, agree a standard note structure across the team.
A good expert call note has four components:
Expert profile. Name (or anonymised reference), background summary (former role, years in the industry, specific area of expertise), and any relevant conflicts or limitations to their perspective. A former CMO at a competitor will see pricing dynamics differently from a distribution channel partner. Both views are valuable; both need to be labelled.
Topics covered. A list of the main subjects discussed, with a note of which were covered in depth versus touched briefly. This is essential for synthesis: if competitive dynamics were discussed in 25 of 30 calls but management quality only came up in 8, that distribution matters.
Direct quotes. At least three to five verbatim quotations per call. Not paraphrases. Not "the expert felt that..." but actual words. Quotes are the raw material of any rigorous synthesis. They are what you cite in the deliverable when an investment committee asks "what is the evidence for this?"
Call synthesis statement. A two-to-three sentence summary written immediately after the call, when the analyst's memory is fresh. This is a working note, not the final synthesis. Its value is that it captures the analyst's immediate reaction before the next call has already begun to overwrite it.
If your team is running calls in parallel across multiple analysts, standardising this structure before fieldwork begins is essential. When documents are inconsistent in format and depth, cross-call analysis becomes much harder. Skimle accepts PDF, DOCX, and TXT formats, so the notes can be written in whichever tool the team uses. The important thing is that the structure is consistent across documents.
What metadata to capture per call
Metadata is the multiplier for expert call synthesis. It turns a flat pile of call notes into a dataset that can be sliced and interrogated.
At minimum, capture these variables per call:
| Metadata variable | Why it matters for synthesis |
|---|---|
| Expert type (former executive / analyst / channel partner / customer) | Different expert types have systematically different perspectives. Analysing by type separates signal from vantage-point effects. |
| Industry tenure (years in sector) | Long-tenured experts often see structural dynamics; shorter-tenured see current conditions. |
| Geography (if relevant) | Market dynamics differ by region. A North American view on competitive pressure may not translate to Europe. |
| Recency of operational role | An executive who left the sector three years ago has different visibility than one who left six months ago. |
| Holding or relationship | Does the expert have any current financial interest in the target or competitors? |
When you run synthesis in Skimle, these metadata variables become filters. You can ask: "What did former operators say about management quality, compared to independent analysts?" or "How did views on pricing pressure differ between US and European experts?" That level of segmentation is impossible with manual synthesis. It is the difference between "experts were generally constructive on pricing" and "former operators, particularly those from the past two years, consistently flagged pricing pressure in the SMB segment, while independent analysts focused on enterprise strength."
For more on metadata-driven analysis, see discovering themes in the data using metadata variables and the metadata analysis docs.
How to code across a call programme: open coding vs predefined frameworks
When you have structured your notes and uploaded them, the analytical question is how to code across the full set. There are two approaches, and the right choice depends on where you are in the project.
Predefined (deductive) coding means starting with a framework you already have. In a CDD context, this typically maps to the thesis you are testing: market growth drivers, competitive positioning, customer retention, management quality, execution risk. You define these categories in advance and assign call content to them. This is fast and produces findings that map directly onto the deliverable structure. The risk is that you miss what is in the data that does not fit your framework.
Open (inductive) coding means letting the themes emerge from the data rather than imposing them. You read across the calls without predetermined categories and identify what topics actually appear, with what frequency and in what relationship to each other. This is slower but often surfaces the unexpected: a risk that was not on your hypothesis list, a competitive development that experts were mentioning that the desk research had not flagged.
In practice, the best approach for a CDD expert call programme is a hybrid. Start with open coding on a subset (five to eight calls) to check whether your predefined framework matches what the experts are actually talking about. Then switch to your predefined categories for the full set, while leaving a residual category for content that does not fit. The residual category often contains the most interesting material.
Skimle supports both approaches. You can run inductive analysis and let the AI surface themes from the data, or set up predefined categories that map to your existing framework. For expert call programmes, running inductive first and then mapping to predefined is the approach most CDD teams find most useful.
The complete guide to qualitative coding covers the methodological options in depth if you want to understand the theory behind the choice.
Why divergence matters more than consensus
This is the most underappreciated aspect of expert call synthesis: the most valuable signal in a call programme is often where experts disagree.
When all 20 experts agree that the market is growing, you have confirmed something you probably already knew. Consensus on well-known facts is not particularly valuable. What is valuable is the outlier: the three experts who are concerned about something everyone else is discounting, or the two who have seen something in the data that contradicts the conventional view.
Divergence has several possible interpretations:
- Information asymmetry. The dissenting experts have seen or know something that the majority have not. This is the most important case: it means the consensus view may be incomplete.
- Vantage point. Experts from different positions in the value chain see different things. Channel partners see pricing dynamics that manufacturers do not. This is a difference in visibility, not a factual disagreement.
- Time horizon. Short-term and long-term views on the same market can look like divergence but are actually different questions. Understanding which time horizon each expert is answering is essential for correct interpretation.
- Real uncertainty. Sometimes experts disagree because the market itself is uncertain and no one has better visibility. That uncertainty is important to represent faithfully in the deliverable rather than paper over with a consensus narrative.
The operational implication for synthesis is that you should not treat divergence as an inconvenience to be resolved. You should surface it explicitly, investigate it, and represent it in the deliverable. "Twelve of 20 experts described pricing as stable or improving; six flagged material pressure in the lower-tier segment. The divergence maps to expert type: former operators are the most concerned, independent analysts the least" is a better synthesis finding than "experts were broadly constructive on pricing."
For more on finding and interpreting divergence, see how to find themes across interviews and the consultant's guide to interview themes.
How to handle conflicting expert views without papering over them
Conflicting views create a temptation to average. "Some experts were bullish, some bearish, overall the picture is mixed" is technically accurate but analytically worthless. An investment committee cannot make a decision from that sentence.
The right approach is to characterise the conflict structurally. Who holds which view? What are their credentials and information basis? What would have to be true for each view to be correct? This last question is particularly useful: if the bear case depends on a regulatory development that three experts consider unlikely, that information is relevant to the probability-weighted view.
A conflict map across the expert set looks something like this:
| View | Held by | Basis | What would have to be true |
|---|---|---|---|
| Competitive position is durable | 14/20 experts (mix of operators and analysts) | Brand strength, switching costs in enterprise | Challenger has not closed the product gap in 18 months |
| Competitive position is eroding | 6/20 experts (predominantly former employees of challenger) | Product improvements, pricing aggression | Challenger's recent hire of VP Product is a leading indicator |
This kind of structured representation is much more useful than a narrative summary. It also makes explicit that six of the sceptics are former employees of the competitor, which is a potential conflict of interest the committee should weight accordingly.
When you are working in Skimle, the categories view shows you which documents contributed to each theme, making it straightforward to build this kind of conflict map from the structured data rather than from memory.
5 steps to translate synthesis into a deliverable
Most expert call programmes feed into one of three deliverables: an investment committee memo, a section of a commercial due diligence report, or a client slide in a strategy presentation. The translation from synthesis to deliverable follows the same logic regardless of format.
Step 1: Map findings to the thesis questions. Your predefined framework should have been built around the questions that matter for the investment or strategy decision. Map each major theme from the synthesis back to those questions. Which themes provide evidence for or against each thesis element?
Step 2: Assign confidence levels. Not all findings are equally well-supported. A theme that appeared across 22 of 30 calls in consistent terms deserves higher confidence than one that appeared in 4 calls and was mentioned only briefly. Make confidence levels explicit. "Strong evidence from 22 of 30 expert calls" reads very differently from "emerging signal from 4 calls."
Step 3: Select quotes strategically. The deliverable should include direct quotes, not just paraphrased findings. Choose quotes that are specific, credible (clearly from someone with direct experience), and representative of the broader finding rather than outliers. Two or three quotes per major finding is typically right for an IC memo. More than five per finding and you are providing raw data rather than synthesis.
Step 4: Represent divergence explicitly. As discussed above, do not suppress conflicting views. A CDD section or IC memo that acknowledges divergent expert views and explains it is more credible than one that presents a falsely unified picture. Committees that are good at investment decisions have learned to distrust synthesis documents that are too clean.
Step 5: Provide the evidence appendix. Include a summary of the full expert call programme as an appendix: how many calls, what types of experts, what geographies, what time period. This gives reviewers the context to assess the weight of evidence and demonstrates that the synthesis is grounded in a programme, not cherry-picked calls.
For more on structuring a primary research section of a client deliverable, see from expert calls to client deliverable and presenting qualitative research findings to executives.
If you are working in a private equity or VC context, qualitative primary research in private equity due diligence covers the specific requirements of deal-speed synthesis in more depth.
How Skimle fits into an expert call synthesis workflow
Skimle is built for exactly this kind of analysis: a bounded set of documents (call notes or transcripts), a need for fast but rigorous cross-document synthesis, and a deliverable requirement that traces every claim back to source material.
The typical workflow is: import your structured call notes or transcripts (Skimle supports all common formats), run the inductive analysis to see what themes emerge across the full set, then map those themes to your predefined framework using Skimle's category management. The automatic thematic analysis processes a typical call programme in minutes, not hours.
The output is a structured category hierarchy with every insight traceable to the source document and quote. When an investment committee member asks "where does this claim come from?", you can show them. That traceability is what separates AI-assisted synthesis from AI-generated summaries that cannot be interrogated.
You can see more about how Skimle fits consulting and due diligence workflows at the consultants and investors use-case page.
Frequently asked questions
How do you synthesise expert network calls without losing nuance?
Nuance comes from treating calls as structured data rather than reading material. Capture direct quotes per call (not just paraphrases), code thematically across the full set rather than summarising sequentially, and represent divergence explicitly rather than averaging it away. The nuance is in the divergence: who said what, from what vantage point, on what basis. Flatten that and you flatten the most valuable signal in the dataset.
How many expert calls do you need before synthesis is meaningful?
There is no universal answer, but most practitioners find that patterns begin to stabilise around 10–15 calls for a focused question, and that diminishing returns set in after 25–30 calls unless the expert universe is highly heterogeneous. For commercial due diligence, the practical constraint is usually the timeline rather than the sample size. What matters more than call volume is expert type coverage: make sure your call set includes multiple vantage points (former operators, independent analysts, channel partners, customers) rather than 25 calls from one type.
What is the difference between expert call synthesis and just reading all the notes?
Reading notes sequentially produces a reconstruction shaped by recency, confirmation, and memory constraints. Synthesis is a systematic process: structured notes with consistent metadata, thematic coding across the full set, explicit divergence analysis, and claims tied to specific quotes from specific calls. The output of synthesis is defensible and auditable. The output of sequential reading is a narrative that reflects what was most memorable, not what was most representative.
How do you present conflicting expert views in a CDD deliverable?
Present them structurally, not as a narrative muddle. Quantify the split (e.g., 14 of 20 experts held view A; 6 held view B). Characterise who holds each view and their information basis. Identify the structural conditions under which each view would be correct. This gives the committee what they need to form a probability-weighted view rather than resolving the conflict by averaging.
Ready to turn your expert calls into structured, defensible findings? Try Skimle for free and see how fast systematic analysis of a call programme can be, without losing the rigour that an investment committee expects.
Related reading:
- How to get more from expert network calls
- Commercial due diligence: qualitative primary research guide
- How to analyse interview transcripts
About the authors
Henri Schildt is a Professor of Strategy at Aalto University School of Business and co-founder of Skimle. He has published over a dozen peer-reviewed articles using qualitative methods, including work in Academy of Management Journal, Organisation Science, and Strategic Management Journal. His research focuses on organisational strategy, innovation, and qualitative methodology. Google Scholar profile
Olli Salo is a former Partner at McKinsey & Company where he spent 18 years helping clients understand the markets and themselves, develop winning strategies and improve their operating models. He has done over 1000 client interviews and published over 10 articles on McKinsey.com and beyond. LinkedIn profile
Sources
- GLG Expert Network — How It Works (GLG)
- AlphaSights Expert Research — Pricing and Process (AlphaSights)
- Doing Qualitative Research: Braun & Clarke — Thematic Analysis (Sage)
- The Impact of Cognitive Bias on Decision Making in Private Equity — Journal of Private Equity
- Expert networks: how institutions use primary intelligence in investment research — Greenwich Associates / Coalition Greenwich
