A consulting deliverable is defensible when every material claim can be traced to a specific, attributed source. Without that traceability, even the most rigorous analysis collapses under client scrutiny — not because the thinking is wrong, but because the evidence cannot be produced on demand.
Expert calls are the primary source of proprietary insight in most strategy engagements. They offer access to knowledge that cannot be extracted from public filings, industry reports, or secondary research. But the way most expert interviews are run produces insights that are untraceable, unattributed, and therefore indefensible under client scrutiny. The notes are vague, the claims are composite, and the confidence level is implicit rather than stated.
This guide covers how to structure the expert interview from prep through synthesis so that every claim that makes it into a deliverable is sourced, confidence-weighted, and audit-ready. The structure is not bureaucratic overhead — it is the mechanism that transforms a good conversation into a defensible research asset.
“The hardest client question to answer is: 'How do you know that?' If your answer is 'our experts told us,' you've lost the room. If your answer is 'three of our four experts with direct operational experience in this market confirmed X, with the fourth providing partial corroboration,' you've won it.”
Interview Prep: The Hypothesis Brief
Every expert interview should begin with a hypothesis brief: a one-page internal document that states what the analyst is trying to learn, why this particular expert is the right person to ask, and what the analyst currently believes before the conversation starts.
The brief covers three components:
1. The thesis component being tested in this call — not the overall engagement thesis, but the specific sub-claim that this expert's background equips them to evaluate.
2. The specific questions designed to test it — written out in advance, not improvised in the moment. Questions should be sequenced to move from open exploration toward targeted confirmation or disconfirmation.
3. The expected expert view, based on their background, and what would constitute a surprise. This is the calibration layer: it forces the analyst to have a prior before the call, which is the only way to recognise when something genuinely unexpected is being said.
The hypothesis brief forces the analyst to articulate what they are trying to learn before the call, not during it. Analysts who skip this step often spend the first fifteen minutes of a fifty-minute call figuring out what they want to ask — a waste of the expert's time and a structural failure that produces unfocused notes.
Before writing the brief, the expert's background should be reviewed with three questions in mind: their role relative to the phenomenon being studied (were they a direct participant, an adjacent observer, or an informed commentator?), their last date of direct involvement (recent direct experience carries more evidentiary weight than older adjacent experience), and any potential bias — including former employee loyalty, competitive animus toward a sector rival, or a financial interest in a particular narrative.
None of these factors disqualify an expert. They calibrate how their testimony should be weighted in synthesis. An expert with a known bias can still provide useful signal — but only if the analyst has recorded the bias and applied the appropriate discount in the deliverable.
“The calls where you learn the most are the ones where the expert says something you didn't expect. But you can only recognize the unexpected if you knew what you expected going in.”
— Senior associate, strategy consulting firmIn-Call Structure: The Four-Phase Expert Interview
A fifty-minute expert interview has four distinct phases. Each phase serves a different research function, and the sequencing is not arbitrary — it mirrors the cognitive structure of credibility assessment. Moving straight to thesis testing without context setting produces answers that cannot be interpreted. Ending without pressure questions leaves the strongest counter-evidence untested.
Phase 1: Context Setting (5 minutes)
Ask the expert to describe their role, tenure, and proximity to the subject matter in their own words. This is not small talk — it is the calibration layer that determines how everything that follows should be weighted. An expert who describes a role that is more adjacent than their profile suggested should be noted; their subsequent claims may need to be reclassified from 'direct' to 'adjacent' source proximity in the post-call protocol.
The context-setting phase also establishes the expert's most recent direct involvement. Experience that is more than five years old should be treated as historical context rather than current market intelligence, unless the expert can demonstrate ongoing engagement with the sector.
Phase 2: Open Exploration (15 minutes)
Open-ended questions about the market, sector, or phenomenon — the goal is to let the expert's frame emerge before imposing yours. This phase surfaces the expert's mental model of the problem, which may differ significantly from the hypothesis brief. That divergence is data.
The most productive question in this phase: "What are the two or three most important things happening in this market right now that most outsiders miss?" This question inverts the usual information asymmetry — instead of asking the expert to confirm your priors, it asks them to surface theirs. The answers are often the most valuable material in the call.
Do not rush this phase to get to the hypothesis brief questions. The open exploration phase is where unexpected signals emerge — signals that the analyst could not have anticipated in prep and that often become the most defensible insights in the final deliverable, precisely because they were unprompted.
Phase 3: Thesis Testing (20 minutes)
This is the core of the hypothesis brief: specific questions designed to test each thesis component. Work through the brief systematically, but do not treat it as a rigid script — if the expert's open exploration phase surfaced material that is more relevant than a planned question, pursue it.
Pay attention to hedging, qualification, and declination. An expert who hedges on a claim they were expected to confirm strongly is providing signal. An expert who declines to answer a question — citing legal constraints, confidentiality obligations, or discomfort — is also providing signal. Both should be recorded as findings, not omissions.
The analyst should note, for each claim, whether the expert is speaking from direct operational experience, secondhand knowledge, or inference. These distinctions matter in synthesis and should be captured in real time, not reconstructed afterward.
Phase 4: Pressure Questions (10 minutes)
Present the strongest version of the counter-thesis and ask the expert to engage with it directly. The framing: "Some analysts argue that X is actually driven by Y — how does that square with what you've seen?" The goal is not to win an argument with the expert. It is to test the robustness of the thesis under the best available challenge.
Experts who push back strongly on the counter-thesis and provide specific, operational reasons for doing so are producing high-confidence confirmatory evidence. Experts who immediately concede the counter-thesis — or who say something like "that's a valid way to look at it" — are often signalling that their prior views were based on reputation or inference rather than direct knowledge.
“Phase 4 is where the most valuable signals come from. The experts who push back strongly on the counter-thesis almost always have specific evidence. The ones who immediately concede are often operating on reputation, not knowledge.”
Post-Call Protocol: The 24-Hour Rule
Claims degrade within 72 hours. The texture of a claim — whether the expert was certain or hedging, direct or inferred, corroborating or contradicting prior calls — is precisely the kind of detail that evaporates from memory before the end of the week. The 24-hour rule exists to capture it while it is still recoverable.
Within 24 hours of any expert call, the analyst must complete four steps:
1. Extract all discrete claims from the transcript or notes. A claim is any assertion the expert made about a market condition, causal mechanism, competitive dynamic, or structural factor. Each claim should be extracted as a standalone statement, not embedded in a summary.
2. Tag each claim with a confidence level — High, Medium, or Low — based on the expert's stated certainty, their source proximity, and whether the claim was volunteered or prompted. Also tag source proximity: direct (the expert experienced this firsthand), adjacent (the expert observed it from a related role), or removed (the expert is reporting secondhand).
3. Map each claim to the relevant hypothesis brief component or thesis section. Claims that do not map to any component should be flagged as potential new signals — they may represent thesis extensions or counter-evidence that requires further investigation.
4. Flag any contradictions with prior calls. A contradiction is not a problem — it is a research finding. It tells the team that the evidence on a particular question is mixed, which should be reflected in the confidence level assigned to the relevant deliverable claim.
This is not an optional best practice. It is the protocol that makes the deliverable defensible. Analysts who defer post-call processing to the synthesis phase — when all six calls are done and the deck is due — consistently produce deliverables where the underlying evidence cannot be reconstructed at the claim level.
From Interview to Defensible Deliverable
Every insight that appears in the final deliverable should be traceable back to three things: a specific claim, from a specific expert (identified by EXP-XXX and descriptor), with a specific confidence level. If any of those three elements is missing, the insight is not ready for the deliverable.
In practice, this means the deliverable should not say: "Industry experts believe that margins are under pressure." It should say: "Four of six experts with direct category experience described margin compression as structural, driven by input cost inflation that distributors have been unable to pass through [EXP-003, EXP-019, EXP-027, EXP-044; confidence high]."
The difference between those two formulations is the difference between a claim a client will accept and a claim a client will push back on. The first formulation invites the question "which experts?" The second formulation answers it before it can be asked.
The client pushback test is a useful quality gate at the synthesis stage. For every material claim in the deliverable, the analyst should be able to answer the question: "How do we know this?" The answer must include a specific claim count, an expert descriptor, and a confidence assessment. "Our research suggests" or "experts we spoke with indicated" are not answers — they are deferrals that will be exposed in any serious client review.
Teams that apply this standard consistently find that their expert panels become smaller and better-targeted over time. When every claim must be traceable to a specific expert, the incentive to conduct unfocused calls disappears. The question is no longer "how many experts have we spoken with?" but "which specific experts can speak to which specific claims?" That shift in framing produces both better research and more defensible deliverables.
Closing: Structure as the Standard
Defensible deliverables are built interview by interview, from prep brief to post-call protocol. The quality of the final product is determined by the quality of the process — and the process is determined by whether the structural elements are treated as mandatory or optional.
The structure is the point. Every element — the hypothesis brief, the four-phase interview, the 24-hour post-call protocol, the claim-level traceability standard — exists to make the final deliverable something you can defend under the hardest client scrutiny. Teams that build this structure into their standard research workflow find that they spend less time defending their conclusions and more time acting on them.