Primary Research for Hedge Fund Analysts: A Practitioner's Guide

Sell-side research is thinner than it's been in a decade. Transcript libraries are crowded — if 80% of top hedge funds are reading the same AlphaSense transcripts, the insights in them aren't proprietary anymore. And yet hedge fund AUM sits north of $5 trillion, with 30% more allocators planning to increase hedge fund exposure in 2025 than decrease it.

The math is straightforward: more capital is chasing the same set of public companies, while the traditional information sources that used to generate edge have been commoditised. The analysts who consistently generate alpha are the ones who build proprietary insight through original primary research — not the ones reading more broker notes.

This guide is a practitioner's playbook. It covers exactly how to structure primary research around an investment thesis, run expert calls that actually produce signal, use surveys and channel checks to quantify what calls can only suggest, and synthesise everything into actionable conviction. No theory. No platitudes about "the importance of research." Just the process, step by step.

What Primary Research Actually Means (and Doesn't)

Primary research is the process of gathering original, first-hand information — through expert interviews, B2B surveys, channel checks, and site visits — to inform, validate, or challenge an investment thesis on a public company.

It is not reading sell-side reports, parsing SEC filings, or scanning news. That's secondary research. It's essential, but it's table stakes. Every analyst at every fund has access to the same filings and the same Bloomberg terminal.

Primary research is also not synonymous with "expert calls." Expert calls are one tool in the toolkit. Surveys, channel checks, customer reference calls, and site visits are equally important — and in many situations, more rigorous and more useful.

The distinction matters because many analysts default to a narrow workflow: read the filings, schedule some expert calls, form a view. That workflow leaves significant value on the table.

The Primary Research Toolkit

Before diving into process, let's define the tools available to you. Each method has specific strengths, and the best research programmes combine multiple methods to triangulate findings.

Expert Calls

One-on-one consultations, typically 30–60 minutes, with subject-matter experts: former executives of your target company, competitors, customers, channel partners, or technical specialists. This is the workhorse of hedge fund primary research and the primary reason most funds use expert networks.

Best for: Deep qualitative insight, understanding management quality, exploring nuance around business model dynamics, testing specific hypotheses with people who have direct operational experience.

Limitations: Inherently qualitative, small sample size, subject to individual bias. One former VP of Sales telling you "the pipeline was weak when I left" is a data point, not a conclusion.

B2B Surveys

Structured questionnaires distributed to a targeted sample of industry professionals — customers, suppliers, decision-makers, or employees. Surveys can reach dozens or hundreds of respondents, providing quantitative data points that expert calls simply cannot.

Best for: Measuring trends quantitatively (e.g., "what percentage of CIOs plan to increase spend on cybersecurity in the next 12 months?"), validating patterns you've heard in qualitative calls, generating statistically defensible findings.

Limitations: Requires careful design to avoid leading questions. Lower depth per respondent than a 45-minute call. Response quality depends heavily on targeting.

Channel Checks

The practice of examining upstream and downstream trends in a company's supply chain. This involves contacting distributors, retailers, suppliers, or customers to assess real-time demand, pricing, inventory levels, and competitive dynamics.

Best for: Real-time demand signals, pricing trends, inventory build-up or drawdowns, competitive wins and losses. Especially powerful in consumer, industrial, and healthcare sectors.

Limitations: Can be time-intensive to execute at scale. Data points are often anecdotal unless systematised across multiple channel participants.

Site Visits and Store Checks

Physical observation of operations: foot traffic, shelf placement, construction progress, warehouse activity, store condition. Often undervalued, frequently revealing.

Best for: Retail and consumer theses, verifying management claims about store openings or renovations, assessing brand presence at the shelf level.

Customer Reference Calls

Interviews with a target company's actual customers — critically, including off-list references you source independently, not just the happy customers management provides. This is where you learn about churn risk, competitive threats, and product weaknesses that will never show up in a management presentation.

When to Use What

Structuring Research Around an Investment Thesis

This is where most analysts either get it right or waste enormous amounts of time. The difference between productive primary research and a "fishing expedition" is whether you've defined what you're looking for before you start.

The Thesis-First Approach

Before you schedule a single expert call, write your investment thesis as a series of testable hypotheses. Not "I want to learn about the market." Instead:

• Hypothesis 1: "Company X's net revenue retention rate is above 120% and sustainable because customers are expanding usage of Module Y."
• Hypothesis 2: "Competitor Z's new product launch will take less than 5% share in the next 12 months because of switching costs."
• Hypothesis 3: "The company's channel partner programme is a meaningful growth driver, contributing >15% of new bookings."

Each hypothesis can be confirmed, challenged, or refined through primary research. This framing does three things:

1. It forces you to articulate what you need to believe for the investment to work.
2. It tells you exactly who to speak with — the hypothesis dictates the expert profile.
3. It gives you a clear stopping rule — you stop when you've adequately tested each hypothesis, not when you run out of credits.

Designing a Research Programme

Once you have your hypotheses, design the research programme around them:

1. Start with secondary research. Read the filings, the transcripts, the sell-side notes. Understand what the consensus view is on each of your hypotheses. Primary research is most valuable when it either challenges consensus or surfaces something consensus hasn't considered.
2. Map hypotheses to methods. For each hypothesis, decide: Do I need expert calls? A survey? Channel checks? Often you need a combination.
3. Determine sample size. For expert calls, 5-8 calls per thesis is typically sufficient — after that, insights start repeating. For surveys, you generally want 30+ respondents to identify meaningful patterns, and 50-100+ for statistical confidence.
4. Sequence the work. Expert calls first (to build qualitative understanding), then surveys (to quantify what you've heard), then channel checks (to verify in real-time). Each method's findings should inform the next.
5. Set a time budget. Primary research expands to fill available time. Set a deadline — "I need to have a view on this name in two weeks" — and work backward.

Example: Researching a SaaS Company's Net Retention

Suppose you're evaluating a mid-cap SaaS company whose bull case rests on 130%+ net revenue retention. Here's how to structure it:

Hypotheses to test:

• NRR is genuinely above 130% and not inflated by a few large upsells
• Expansion is driven by organic product adoption, not forced bundling
• Churn is concentrated in SMB, not enterprise (which matters for sustainability)

Research design:

• 3 expert calls with former sales leaders at the company — to understand the expansion motion and where churn comes from
• 2 calls with customers who churned — to understand why they left
• 2 calls with competitors — to understand win/loss dynamics
• A survey of 40-50 current customers — asking about expansion plans, satisfaction, competitive alternatives they've evaluated

Total: ~7 calls + 1 survey. Targeted, efficient, and directly tied to the investment thesis.

Running Expert Calls That Actually Produce Signal

The difference between a call that changes your conviction and one that wastes 45 minutes comes down to preparation and technique.

Pre-Call Preparation

Write a discussion guide. Every call should have a written discussion guide with 8-12 questions, organised around your hypotheses. Not a script — a guide. You want to be able to follow the conversation where it leads, but you need a structure to ensure you cover what matters.

A good discussion guide follows this architecture:

• Background questions (2-3): Confirm the expert's relevant experience, role, tenure, and how current their knowledge is. This helps you weight their input appropriately.
• Hypothesis-testing questions (5-7): Your core questions, each linked to a specific hypothesis. Be direct. "In your experience, what was the typical expansion path for enterprise customers?" is better than "Tell me about the company."
• Calibration questions (2-3): Questions that help you assess the expert's accuracy and potential biases. "What's something the market gets wrong about this company?" is a reliable signal generator.

Review the expert's background. Before the call, understand their role, tenure, and what they'd plausibly know. A former regional sales director can speak to pipeline dynamics; they probably can't speak to gross margin trajectory. Match your questions to their expertise.

During the Call

Recommended structure:

• Context-setting (2 minutes): Briefly explain what you're researching and what you're hoping to learn. This helps the expert give you relevant answers.
• Hypothesis-testing (20-25 minutes): Work through your discussion guide. Listen for specifics — numbers, names, timeframes. Push back gently when answers are vague: "Can you help me size that? Are we talking about a 5% share loss or a 20% share loss?"
• Open-ended exploration (5 minutes): "What haven't I asked about that you think is important?" This is where unexpected insights emerge.
• Follow-ups (2 minutes): Clarify any ambiguous points. Ask if they can suggest other people you should speak with.

Common Mistakes That Kill Call Quality

• The fishing expedition: "Tell me about the industry." This wastes the expert's time and yours. You should already know the industry basics from secondary research. Use the expert for what only they can tell you.
• Leading questions: "Don't you think the company is losing share?" invites confirmation. "How has the competitive landscape changed in the last 12 months?" is neutral and more likely to produce honest signal.
• Not pushing for specifics: If an expert says "the product was having issues," you need to ask: "What kind of issues? How widespread? Was this affecting enterprise customers or SMB? When did this start?" Vague inputs produce vague outputs.
• Treating one expert as gospel: Expert insights are directional, not definitive. A single expert's view is shaped by their specific role, geography, and timeframe. Always triangulate across multiple sources.
• Scheduling too many calls without pausing to synthesise: If you schedule 15 calls back-to-back before reviewing your notes, you'll miss the opportunity to refine your questions as your understanding deepens. After every 3-4 calls, stop, review, and adjust your discussion guide.

Surveys and Channel Checks: The Quantitative Layer

Surveys are the single most underutilised tool in the hedge fund analyst's toolkit. Most analysts default to expert calls for everything — but there's an entire category of research questions that calls can't answer well and surveys can.

When Surveys Beat Calls

Use surveys when you need:

• Quantification: "What percentage of IT buyers are considering switching vendors?" You can't answer this with 5 expert calls. You need a sample.
• Breadth over depth: You want to hear from 50 customers, not 5. Surveys let you reach a much larger sample within a shorter timeframe.
• Statistical defensibility: When you're building a case for your PM or IC, "73% of 60 surveyed customers said they plan to renew" is significantly more persuasive than "the three people I spoke with seemed happy."
• Contrarian insights at scale: Targeted surveys have helped leading hedge funds discover contrarian insights that one-on-one consultations couldn't deliver — patterns that only emerge with larger sample sizes.

Designing Surveys for Investment Research

Investment surveys are different from market research surveys. Key principles:

• Keep them short. 10-15 questions maximum. Busy industry professionals won't complete a 40-question survey.
• Mix closed and open-ended questions. Closed-ended for quantification ("On a scale of 1-10, how likely are you to renew?"), open-ended for colour ("What's the primary reason you'd consider switching?").
• Target precisely. A survey of 50 relevant decision-makers beats a survey of 200 loosely-related respondents. The quality of your targeting determines the quality of your data.
• Benchmark when possible. Ask questions you can re-ask in 3 or 6 months to track trends over time. The real power of survey data often comes from longitudinal comparison, not a single snapshot.

Channel Checks: Real-Time Intelligence

Channel checks give you something neither calls nor surveys can: real-time demand signals from the people who see transactions happen. A distributor knows whether orders are accelerating before the company reports it. A retailer knows whether a product is sitting on shelves before inventory data hits the filings.

Running effective channel checks by sector:

• Software/SaaS: Talk to resellers, system integrators, and implementation partners. Ask about deal velocity, competitive displacements, customer objections during the sales cycle.
• Healthcare: Talk to distributors, GPO contacts, hospital procurement, and prescribers. Ask about formulary changes, utilisation trends, competitive product adoption.
• Consumer/Retail: Visit stores. Check shelf placement, inventory levels, promotional activity. Talk to store managers. Supplement with foot traffic and credit card data.
• Industrials: Talk to distributors, contractors, and regional sales reps. Ask about order backlogs, pricing changes, and customer capex plans.

Synthesis: Turning Research Into Conviction

Gathering data is getting easier. The bottleneck has shifted to synthesis — turning 10-20 data points from expert calls, survey results, and channel checks into a clear, actionable investment view.

A Practical Synthesis Framework

1. Map findings back to hypotheses. For each hypothesis you started with, what did the research say? Confirmed? Challenged? Inconclusive? Be honest — inconclusive is a valid finding that tells you where to dig deeper or where the risk lies.
2. Weight the evidence. Not all data points are equal. A survey of 60 customers carries more weight than a single expert call. A current channel partner's perspective is more current than a former executive who left two years ago. Weight accordingly.
3. Identify consensus vs. variant perception. Where does your research align with what the market already believes? That's not alpha. Where does it diverge? That's where the edge is.
4. Document dissenting views. If 7 out of 8 experts were positive but one raised a specific concern, don't dismiss it. Investigate it. The minority view is sometimes the most valuable.
5. Quantify the impact. Translate qualitative findings into model implications. If channel checks suggest pricing is under more pressure than the market expects, what does that mean for gross margins? For revenue? For the stock?

Using AI to Accelerate Synthesis

AI tools — both general-purpose (Claude, ChatGPT) and specialised (AlphaSense) — can meaningfully compress the synthesis step:

• Transcript summarisation: Feed expert call transcripts into an AI tool and ask it to extract key themes, data points, and areas of agreement/disagreement across calls.
• Pattern identification: AI can scan across 15 transcripts and flag recurring themes or contradictions you might miss reading sequentially.
• First-draft synthesis: Use AI to generate a first pass at a research summary, then layer in your own judgment and interpretation.

But be clear-eyed about what AI cannot do: it cannot generate the proprietary data. It cannot tell you whether the expert was credible or hedging. It cannot weigh conflicting evidence the way an analyst with sector expertise can. AI compresses the synthesis time — the human judgment in interpreting what the research means for the stock remains irreplaceable.

Compliance and MNPI: What Every Analyst Must Know

Primary research operates in a compliance-sensitive environment, and the rules are non-negotiable. The expert network industry now operates under a stringent compliance framework shaped by the SEC's insider trading crackdown, which exposed systemic vulnerabilities in how MNPI was handled.

The Core Rules

• Material Non-Public Information (MNPI): Information that is both material to an investment decision and not publicly available. You cannot trade on it. Full stop.
• What you can't ask: Anything that would elicit MNPI — unreleased financial results, pending M&A activity, undisclosed regulatory actions, upcoming product approvals that aren't public.
• What you can ask: An expert's general industry knowledge, their perspective on publicly known trends, their historical experience at a company, their opinion on competitive dynamics based on publicly available information.
• Cooling-off periods: Expert networks implement mandatory cooling-off periods for former company insiders — a waiting period between when someone leaves a company and when they can consult on that company. This prevents fresh, potentially material knowledge from being shared.
• Reg FD (Regulation Fair Disclosure): Public companies cannot selectively disclose material information. This regulation, implemented in 2000, is a key reason expert networks exist — investors needed alternative, compliant ways to gather industry intelligence when companies could no longer share information selectively.

Practical Compliance Best Practices

• Document everything. Keep records of who you spoke with, when, what was discussed, and how it informed your investment decision.
• Use compliance-monitored platforms. Reputable expert networks vet experts and monitor calls. This protects you.
• If an expert starts sharing something that sounds like MNPI, stop the call. Immediately. Report it to your compliance team. This is not optional.
• Post-engagement review. Institutional best practice is to review expert interactions after the fact to ensure nothing crossed the line. Build this into your workflow.

Choosing and Managing Research Providers

The expert network market has grown to roughly $3 billion in 2025, with over 120 providers operating globally. The landscape breaks into distinct categories, and understanding which model fits your needs is critical to getting the most from your research budget.

The Self-Service Expert Network Model

This is the traditional model: you identify a need, request an expert through the platform, schedule the call, conduct the call yourself, and synthesise the notes. The major players include:

• GLG (Gerson Lehrman Group): The pioneer. Access to 900,000+ experts across 150+ countries. Premium pricing (~$1,350/hour). Still the most widely used — roughly 50% of expert network clients use GLG.
• AlphaSense/Tegus: Merged in 2024 to create a $4 billion platform combining AlphaSense's AI-powered search (10,000+ content sources) with Tegus's 100,000+ expert call transcripts. Strong on the transcript-first research approach. Now offering AI-led expert interviews — an early-stage but notable development.
• Guidepoint: 1 million+ experts across 200 industries. Known for flexibility and custom solutions including surveys and long-term engagements.
• AlphaSights: Known for speed — connects clients to experts within hours. Strong service model.
• Third Bridge: Forum product provides pre-existing transcripts. Also offers traditional expert calls.

The trade-off: Self-service networks give you access. You still do all the work — scheduling, question design, conducting calls, synthesising. For a 3-person analyst team covering 30+ names, the time burden is significant. A senior analyst's time is extraordinarily valuable to the fund; every hour spent on call logistics is an hour not spent on analysis.

The Done-for-You Model

A fundamentally different approach: you brief a research provider on your thesis question, and they handle the entire process — designing the methodology, sourcing and vetting experts, writing discussion guides, conducting interviews or surveys, and delivering synthesised findings.

This model is gaining share for several reasons:

• Resource-constrained teams get leverage. Most fundamental hedge fund teams are small — 3 to 10 investment professionals. They don't have dedicated research operations teams. Done-for-you providers effectively become an extension of the team.
• Better research design. Providers who run hundreds of research projects develop pattern recognition on methodology — which questions work, which expert profiles produce the best signal, how to structure surveys that generate actionable data.
• Faster time-to-insight. Instead of spending a week scheduling and conducting 8 calls yourself, you get a synthesised deliverable — often faster because the provider is working on your project in parallel with your other analysis.
• Cost-effectiveness when you account for analyst time. The direct cost may be comparable to self-service, but the total cost — including the opportunity cost of analyst time — is often lower.

This is what we do at Woozle Research. Our clients brief us on what they need to know — whether it's testing a specific thesis, running a customer survey, or executing a set of channel checks — and we deliver the finished research. No scheduling. No discussion guide writing. No transcript synthesis. If that model fits how your team operates, we'd welcome a conversation.

Aggregator Platforms

Platforms like Inex One combine 30+ expert networks into a single dashboard, allowing you to compare expert availability, pricing, and speed across providers. Useful if you work with multiple networks and want to streamline procurement.

How to Evaluate Providers

When selecting or reviewing research providers, optimise for these factors in order of priority:

1. Speed: How quickly can they deliver a relevant expert or a finished research output? In markets that move on earnings and events, 48 hours versus 5 days is the difference between actionable insight and stale information.
2. Relevance: Are the experts actually relevant to your specific question? A generic "industry expert" is far less valuable than someone who was the VP of Product at the exact company you're researching.
3. Compliance: Is the provider's vetting, monitoring, and documentation rigorous? This is table stakes, but not all providers are equally thorough.
4. Pricing model alignment: Does the economic model match how you actually use the service? Price is rarely the deciding factor, but buyers need predictable economics. Hidden fees, rigid terms, or models that penalise volume make adoption harder. Unused credits that expire are a common sore point.

A practical tip: Work with multiple providers. Institutions often use 2-4 providers to diversify access and reduce dependency. Different providers have different expert networks, different strengths by sector or geography, and different service models. There's no reason to be monogamous.

AI and the Future of Primary Research

95% of hedge fund managers now allow access to generative AI tools. 90% are using AI somewhere in their investment management process. The adoption is real. But the application to primary research specifically is more nuanced than the headlines suggest.

What AI Can Do Now

• Compress secondary research: AI-powered platforms like AlphaSense let you search across earnings transcripts, filings, broker research, and news using natural language queries. What used to take a day of reading takes an hour.
• Summarise and analyse transcripts: Feed 15 expert call transcripts into an AI tool and get key themes, contradictions, and data points extracted in minutes.
• Draft discussion guides: Given a thesis and target company, AI can produce a solid first draft of questions to ask an expert.
• Expand coverage: Multi-strategy funds are deploying AI agents to expand stock coverage dramatically — analysts who covered 20 stocks may soon cover 200 with AI support.

What AI Cannot Do (Yet)

• Generate proprietary primary data. AI can process data that already exists. It cannot call a former VP of Sales and ask them why the company's enterprise pipeline softened in Q3. That requires a human conversation or a carefully designed survey.
• Judge expert credibility. Was the expert hedging? Were they bitter about being let go? Do they have an axe to grind with the current management team? Reading these signals requires human judgment.
• Ask the right follow-up question. When an expert says something unexpected, the analyst who catches it and probes deeper is doing something AI can't replicate — connecting a novel data point to their existing mental model of the company.
• Make the investment decision. Synthesis, weighting of evidence, and conviction-building remain fundamentally human processes.

The Emerging Development: AI-Led Expert Interviews

AlphaSense has introduced AI-led expert calls, where an AI interviewer conducts the conversation with a client-chosen expert. The pitch: 70% cost savings compared to traditional expert network calls, and the ability to scale the number of calls dramatically.

This is early-stage. The technology works best for structured, factual questions. It works less well for the nuanced, follow-up-heavy conversations that produce the most valuable insights. Worth watching, but not yet a replacement for a skilled analyst conducting their own calls — or a research provider conducting calls on their behalf.

Putting It All Together: A Practical Workflow

Here's the step-by-step process from new investment idea to research-backed conviction:

Step 1: Secondary Research and Thesis Formation (Day 1-2)

Read the filings, earnings transcripts, sell-side notes, and relevant industry reports. Use AI tools to accelerate this. Form your initial investment thesis and articulate it as 3-5 testable hypotheses.

Step 2: Research Design (Day 2)

For each hypothesis, determine: What primary research method will test this? Who do I need to speak with? How many data points do I need? Map out your expert profiles, survey design, and channel check targets.

Step 3: Expert Calls — First Wave (Day 3-7)

Conduct 3-5 initial expert calls to build qualitative understanding. Write discussion guides tied to your hypotheses. After each call, update your notes and refine your questions for the next call.

Step 4: Survey and Channel Checks (Day 5-10)

Based on what you've heard in calls, design and deploy a survey to quantify key findings. Simultaneously, run channel checks with distributors, partners, or customers for real-time data points. These can run in parallel with later expert calls.

Step 5: Expert Calls — Second Wave (Day 8-12)

Conduct 2-4 additional calls, now informed by survey data and channel check findings. These calls should be highly targeted — you're filling specific gaps, resolving contradictions, and pressure-testing your emerging view.

Step 6: Synthesis and Decision (Day 12-14)

Map all findings back to your original hypotheses. Weight the evidence. Identify where your view differs from consensus. Quantify the model implications. Write your research summary and make your recommendation.

Step 7: Ongoing Monitoring (Post-Investment)

Primary research doesn't end when you put on the position. Set up quarterly channel checks or periodic expert calls to monitor the thesis. A fund invested in a retail stock might survey consumers each quarter to track brand perception, or check in with distributors monthly to monitor order trends. Early detection of thesis deterioration is how you protect capital.

Key Takeaways

• Start with hypotheses, not calls. Define what you need to believe for the investment to work, then design research to test those beliefs.
• Use the full toolkit. Expert calls are one method. Surveys, channel checks, customer references, and site visits each provide different — and often more rigorous — forms of evidence.
• Invest in synthesis. The bottleneck isn't data gathering; it's turning data into a clear investment view. Use AI to compress synthesis time, but don't outsource judgment.
• Know when to stop. After 5-8 expert calls on a topic, insights start to repeat. Diminishing returns are real. Move on to a different method or make a decision.
• Triangulate everything. No single data source is definitive. Cross-reference expert calls with surveys, channel checks with financial data, qualitative impressions with quantitative evidence.
• Consider the total cost of research. The direct cost of an expert call is easy to measure. The opportunity cost of the analyst's time spent scheduling, conducting, and synthesising is not — but it's often the larger number. Whether you use a done-for-you provider or build an efficient internal process, optimise for total cost, not unit cost.
• Compliance is not optional. Document everything. Use vetted platforms. Stop any conversation that ventures into MNPI territory. This protects you and your firm.

Primary research remains the single most reliable way for fundamental hedge fund analysts to generate differentiated insight. The tools are evolving — AI is compressing timelines, transcript libraries are growing, survey technology is improving — but the core skill hasn't changed: asking the right questions of the right people and synthesising what you learn into a view the market hasn't priced.

The analysts who do this well, systematically and repeatedly, are the ones who generate alpha. Everything else is noise.

Woozle Research is a done-for-you primary research provider for investment professionals. We handle the entire research process — expert interviews, surveys, channel checks — so your team can focus on analysis and decision-making, not logistics. If you'd like to see how we can support your next research project, get in touch.