Clarify ambiguous requirements under pressure

## Part 1 — Example STAR Response (Misunderstood Problem → Recovery) - Situation: Our sales VP asked for “a list of accounts similar to our top customers to accelerate Q3 pipeline.” As a data scientist, I assumed “similar” meant product-usage similarity and built a model using app telemetry (features used, seat counts, growth rate) to generate lookalikes. - Task: Deliver a ranked list of 500 lookalike accounts for outbound in two weeks. - Action: 1) I engineered a cosine-similarity model on normalized telemetry features, applied territory deduping, and pushed the top 500 into the CRM. 2) Within days, AEs flagged that many accounts were small SaaS firms outside target industries. In a review, I realized the VP meant “enterprise CPG accounts with similar buying centers and budgets,” not “similar product behavior.” 3) I paused rollout, owned the mismatch, and ran a 45‑minute alignment session with Sales Ops and two senior AEs to define: - Objective: Generate net‑new enterprise prospects. - Must‑haves: Industry = CPG/Beverage, Revenue > $5B, North America, multi‑brand portfolio, large distributor network. - Success: AE acceptance rate ≥ 60% on top‑100, meetings booked in 30 days +30% vs. baseline. 4) I rebuilt the approach: rule‑based hard filters (industry/SIC/NAICS, revenue, geography), then a logistic regression ranking on firmographics, web signals (job postings for data/analytics), and tech‑stack tags. I created a one‑page spec and sample output for sign‑off before full run. - Result: The revised top‑100 list had 65% AE acceptance (vs. 32% initially) and produced 18 first meetings in 30 days (vs. 10 baseline, +80%). I published a two‑page post‑mortem and introduced an intake checklist and written confirmation step for all future requests. Trust with Sales improved; the checklist became a team standard. - What prevented fallout: - Immediate acknowledgement and a clear correction plan with dates. - A small pilot (top‑50) before global CRM write‑back. - Written definitions of “objective,” “must‑haves,” and “success metric,” approved by stakeholders. ## Part 2 — Role‑Play: Five Clarifying Questions for “Find clients similar to Coca‑Cola” 1) Objective and action: What will you do with the list (prospecting net‑new accounts, upsell/cross‑sell to existing clients, benchmarking, or partnership outreach)? 2) Entity definition: When you say “Coca‑Cola,” do you mean the parent company (The Coca‑Cola Company), specific brands, or regional bottlers—and should “clients” be current customers or net‑new prospects? 3) Similarity criteria and hard filters: Which attributes define “similar” for you—industry (CPG/beverage), revenue/employee size, geography, distribution model, ad spend, product portfolio, tech stack—and are there any must‑include/exclude filters? 4) Scope and output: How many matches do you need, for which markets/regions and time horizon, and in what format should I deliver them (CSV, CRM upload, dashboard)? Any refresh cadence? 5) Success and constraints: How will we measure success (e.g., AE acceptance rate, meetings booked in 30 days, precision@K), by when do you need this, and are there data, budget, or compliance constraints on sources we can use? ## Part 3 — Techniques to Prevent Misunderstandings Under Pressure - Reflective listening and paraphrasing: - Repeat back the request in plain language: “You want net‑new enterprise CPG prospects in North America similar to the parent company, evaluated by AE acceptance within 2 weeks. Did I get that right?” - Ask for a 10‑second example and a non‑example to expose hidden criteria. - Assumption list and anti‑goals: - Write explicit assumptions (e.g., “Coca‑Cola = parent company; exclude bottlers; revenue > $5B”). - Capture anti‑goals (e.g., “Not a brand‑affinity audience model; not consumer lookalikes”). - Written confirmation (single‑page brief): - Problem statement, objective, success metric, scope, constraints, sample output, timeline, owners. Share within 24 hours; proceed only after sign‑off. - Early sample and metric check: - Deliver a small sample (top‑10) with rationales and a proposed metric (e.g., precision@10). Quick feedback beats perfect silence. - Define “done” with measurable criteria: - Example: “Deliver top‑100 accounts in NA CPG with revenue > $5B; AE acceptance ≥ 60%; 2‑week turnaround.” - Boundary cases and glossary: - Clarify tricky terms (client, account, parent vs. subsidiary, region). Use a shared glossary to avoid repeat confusion. - Cadence and decision log: - Schedule a 15‑minute midpoint check; maintain a brief decision log (date, decision, rationale) to ensure alignment and continuity. - Communication tactics for accents and time pressure: - Slow pace, paraphrase key terms, confirm spellings and entities (e.g., parent vs. brand), use chat to capture names, and follow up with a written summary. When appropriate, share visuals (sample list, filters) to reduce ambiguity. - Guardrails and pilots: - Start with rule‑based must‑haves, then layer modeling. Pilot with a small cohort and pre‑defined acceptance criteria before full deployment. These habits convert ambiguous asks into well‑scoped, measurable projects, reduce rework, and build stakeholder trust under time pressure.