Data Collection and Research Techniques

4.2 Primary Research Design – Step-by-Step Guide

Primary research is how you move beyond “what’s on the internet” to what real buyers, users, partners, and competitors will actually do next. It should be fast, narrow, and hypothesis‑driven—not a fishing expedition. The aim is to generate proprietary, decision‑grade evidence that either proves the thesis or breaks it early enough to change price, structure, or go/no‑go. What follows is a step‑by‑step guide you can lift and run in any diligence sprint.

Start with the investment decision and work backward. Each interview, survey, or channel check must tie to a specific “would‑have‑to‑be‑true” statement and a dollar value at stake. You are designing a test, not a conversation. Keep the footprint tight, bias toward revealed behavior (contracts, invoices, win/loss outcomes) over opinions, and use a second method to cross‑check any killer variable.

Step 1: Translate hypotheses into research objectives

 Restate each top hypothesis as a research objective with a measurable endpoint. If your hypothesis is “realized price can rise 2–4% annually without elevating churn,” your objective might be “estimate willingness‑to‑pay corridors and procurement pushback thresholds in segments A/B/C; identify the three most credible value messages that justify increases.” Decide up front how this evidence would change the model or the term sheet. If it wouldn’t, drop it.

Step 2: Choose the right method for each objective

 Match the question to the instrument. In diligence, you rarely need every method; pick the shortest path to a credible answer.

• Customer depth interviews: Decision process, switching costs, perceived differentiation, drivers of churn/expansion, and how procurement actually negotiates.
• Win/loss calls: Why you won or lost in the last 6–12 months; competitor claims that landed; price vs. non‑price drivers.
• Procurement interviews: Guardrails, MFNs, discount ladders, how increases flow through, and who signs off.
• Channel/distributor checks: Shelf, territory coverage, sell‑in vs. sell‑through, rebate pressure, and private‑label risk.
• Expert calls: Market structure, competitor capabilities, and likely responses.
• Quantitative surveys: Directional sizing of behaviors, pricing experiments, feature trade‑offs, and segmentation hypotheses.
• Mystery shopping/secret buyer: Price quotes, discounting rules‑of‑thumb, lead handling, and sales process friction (only where ethical and legal).
• Observational/usage shadowing: In field‑service, industrial, or healthcare settings where workflow and ergonomics drive stickiness.

Step 3: Define cohorts and a sampling frame

 Design your sample to reflect where the revenue and risk truly sit—not a generic census. Typical diligence cohorts include: top accounts by revenue; churned accounts from the last 12–18 months; active prospects and recent losses; channel partners or distributors by tier; procurement leaders; and, where relevant, end users vs. economic buyers vs. technical approvers. Set quotas by segment, region, ACV band, tenure, and product bundle so you can cut results along the same axes you’ll model.

Rules of thumb for sample sizes (directional, not academic power studies):

• Qualitative depth: 12–20 interviews per key cohort will surface 80% of recurring themes.
• Directional quant survey (single market): ~100–150 completes yields a 95% margin of error around ±8–10 percentage points; ~300–400 narrows to ~±5–6 points.
• Pricing tasks (Gabor‑Granger/Van Westendorp): 150–300 completes per major segment for stable ranges.
• Discrete choice/conjoint: 200–400 completes per segment with 8–12 choice tasks can estimate main effects reliably; keep designs simple in a diligence sprint.

Step 4: Build a tight screener that protects quality

 Your screener is a filter, not a courtesy form. Confirm role (economic buyer, technical approver, user), decision authority, recent purchase/renewal event, and category familiarity. Exclude agency employees and competitors. Add quality controls: attention checks, time‑in‑survey minimums, red‑herring questions, IP/device deduping, and open‑end validation. For B2B, include revenue band and industry codes to match your bottom‑up sizing cuts. For churned or lost‑deal cohorts, require recency and proof (e.g., PO or vendor name remembered).

Step 5: Draft instruments that test, not lead

 Interview guides should be semi‑structured and written to disconfirm.

• Opening context: role, responsibilities, and what “good” looks like in their job.
• Decision journey: trigger, shortlist, evaluation criteria, proof points, buyer committee.
• Alternatives and trade‑offs: who else they considered, perceived pros/cons, and where the target underperformed.
• Price and terms: how budgets are set, discount norms, rebates, price‑increase history, triggers for re‑bid.
• Stickiness: switching costs (data, integrations, retraining), lock‑in clauses, and what would make them leave or expand.
• Future state: what would have to be true for them to pay more, buy more, or switch.
• Disconfirming prompts: “What are we missing?” “Under what conditions would the opposite of your last answer be true?”

Survey design should emphasize forced trade‑offs and revealed preferences over generic satisfaction bars: short batteries (≤ 15 minutes total), randomized item order, balanced scales with labeled anchors, and minimal open ends. Pick pricing modules with intent: Van Westendorp for quick corridor bracketing, Gabor‑Granger for acceptance at discrete prices, discrete choice for attribute‑price trade‑offs when time allows. Use MaxDiff for feature importance; translate results into clear “keep/kill” product messages for the deck.

Step 6: Secure compliance and clean‑team approvals

 Before fieldwork, lock a one‑page compliance memo: who you will contact, what you will say, and what you will not ask. Confirm consent language, sponsor disclosure (blind or disclosed), compensation ranges, and data handling. If operating with a clean team, list who can see raw transcripts or PII, how notes will be redacted, and what aggregation rules apply before sharing with non‑clean members. Reiterate antitrust guardrails (no price coordination, no customer allocation) to anyone making outreach.

Step 7: Select recruiting channels and vendors

 Choose the shortest credible path to your quotas: existing customer lists via counsel‑approved outreach; panel providers for broad B2B/B2C samples; expert networks for niche roles; targeted LinkedIn outreach for senior titles; distributor networks for channel checks. Pre‑approve POs and incentives to avoid delays. For sensitive cohorts (C‑suite procurement, clinicians), expect smaller yield and higher incentives—budget and timelines accordingly.

Step 8: Launch fieldwork and monitor in real time

 Stand up a daily dashboard with four numbers: invites sent, completed by cohort vs. quota, book rate, and drop‑offs at key questions. Watch mix balance: if Segment B is under‑represented by Day 2, shift effort and incentives. Tag each interview with segment metadata at booking so analysis can move as calls land. Record interviews (with consent) and transcribe within 24 hours. Keep a live log of notable quotes and contradictions linked to hypotheses.

Step 9: Enforce data‑quality gates

 Reject suspicious completion immediately. Signals include implausible speed, straight‑lining, inconsistent answers, nonsensical open‑ends, duplicate IPs/devices, or geographic mismatch. For interviews, watch for vendor coaching, unusually polished talking points, or reluctance to discuss alternatives. Keep a “do not use” folder—never pad counts with weak data. Document acceptance/rejection rules in your data book.

Step 10: Code, quantify, and triangulate

 Convert qualitative notes into a small set of coded themes tied to your hypotheses: price drivers, defectors’ reasons, perceived differentiation, procurement veto points. Quantify frequency and strength by segment. For surveys, run simple but decision‑relevant cuts: top‑two box shares, corridor estimates, and differences between loyalists, switchers, and churned. Weight only when you have a defensible population frame (e.g., revenue by segment); otherwise report unweighted and state limits. Triangulate: reconcile stated willingness to pay with realized invoices or competitor quotes; reconcile churn stories with cohort math; reconcile distributor assertions with sell‑through where available.

Step 11: Translate findings into model‑ready inputs

 Every research output should land in the model or the risk ledger within 24 hours. Examples: update realized price assumptions by segment and range; adjust NRR/GRR by cohort and confidence; change conversion rates or ramp times in the funnel; add downside triggers where procurement pushback or competitor aggression is likely. If evidence is thin, widen sensitivity ranges and lower confidence; do not round directional answers into false precision.

Step 12: Close the loop with synthesis and auditability

 Select three to five quotes (anonymized) that illustrate the core finding, footnote the cohort, and show the number of interviews/surveys that support it. Maintain a source register with date stamps, licenses, and storage locations for audio, transcripts, and raw survey data. Before any exhibit leaves the workbench, a verifier should replicate the numbers from raw files and sign off.

Practical blueprints you can copy

• 72‑hour launch plan
  Day 0: finalize hypotheses, cohorts, quotas, and compliance memo; approve vendors and incentives; publish interview guide and survey draft.
  Day 1: program survey and schedule first 10–15 interviews; launch expert calls; begin outreach to churned/lost cohorts; counsel signs off on scripts.
  Day 2: first interviews complete; survey soft‑launch and quality check; adjust quotas, incentives, and screening logic; start coding early themes tied to hypotheses.
• Interview guide skeleton (15–30 minutes)
  Context → Decision journey → Alternatives → Price/terms → Stickiness/switching → Future state → Disconfirming prompts → Close and permission to follow up.
• Survey module menu (pick 3–5, keep ≤ 15 minutes)
  Awareness/consideration; short value messaging test; purchase criteria stack‑ranking; pricing module (Van Westendorp or Gabor‑Granger; DCE if time allows); retention and expansion intent; competitor shortlist; open‑end “what would make you switch/expand.”

B2B vs. B2C nuances that matter

• B2B: The Decision‑Making Unit (DMU) is real—separate the voice of the user, the technical approver, and the economic buyer. Procurement’s memory of discounts and clauses is usually better than end users’. Expect longer scheduling cycles and lower survey response rates; over‑recruit and pre‑book backups.
• B2C: Panels and digital intercepts can scale quickly; watch channel bias (marketplaces vs. DTC vs. retail) and over‑representation of deal‑seekers. Pair stated intent with digital‑shelf and scanner data from secondary sources to keep stories honest.

International and regulated contexts

Localize instruments; avoid idioms; test translations. Confirm data‑privacy rules (e.g., cross‑border transfer, consent), and be explicit about compensation norms by market. In healthcare, financial services, and public sector, clear scripts with counsel and avoid any questions that could elicit confidential or regulated information.

Compliance and ethics you cannot skip

Always disclose who you are and the purpose of the research unless counsel approves a blinded approach. Obtain informed consent before recording. Do not solicit or store material nonpublic information. Never ask about competitor pricing intentions or future coordination. Store PII in encrypted locations with limited access and a defined destruction date.

Common pitfalls—and how to avoid them

• Over‑scoping: Too many questions with no path to a decision. Fix by aligning each question to a hypothesis and model input.
• Leading questions: “How valuable is our feature?” becomes “Which alternatives solved this better?”
• Survivorship bias: Talking only to happy customers. Force churned and lost‑deal cohorts into the mix.
• Thin evidence: Relying on 12 interviews to make a market‑wide claim. Use surveys or a second method to quantify.
• Slow fieldwork: Waiting for perfect scripts. Ship v1, monitor, and iterate by Day 2.
• Legal drag: Launching before clean‑team and counsel approvals. Treat compliance as a gating dependency, not a parallel task.

Primary research checklist (paste into your workspace)

• Hypotheses mapped to specific objectives and dollar value at stake.
• Cohorts and quotas defined; sampling frame matched to revenue mix.
• Screener with role filters and quality checks approved.
• Interview guide and survey instrument written to disconfirm.
• Compliance memo signed; clean‑team rules and consent language locked.
• Vendors, incentives, and POs approved; booking links live.
• Daily dashboard tracking invites, completes, and cohort balance.
• QA gates enforced; rejects logged; transcripts stored securely.
• Triangulation complete for killer variables (second method identified or executed).
• Findings translated to model inputs, ranges, and confidence ratings; residual risk documented.

Run this play with discipline and your primary research will do what diligence needs most: deliver fast, defensible proof on the few questions that move value—without drowning the team in noise or creating compliance risk.