i used to write long, fuzzy assumptions to sound thorough. after a few brutal debriefs from veterans in this community, i learned to cut fluff and make each driver measurable. vets pushed me to replace ‘likely adoption’ with a proxy (search interest, competitor revenue) and to always attach a one-line check. that feedback ruthlessly exposed weak drivers and made my estimates sponsor-ready. what are the simplest proxies you force yourself to use under time pressure?
good feedback is blunt because niceties waste time. i ask one follow-up: ‘how would you prove that number in 60 seconds?’ if they can’t point to a proxy, it’s garbage. people love ‘customer interest’ as an assumption — until you ask for a measurable indicator. force them to pick a data point or shut up. simple, mean, effective. nobody wants to hear emotional assumptions.
pro tip from too many wasted hours: if your assumption needs a paragraph to explain, it won’t survive a real interview. trim it to a number and a one-liner source. otherwise you’re just storytelling. vets will call that bluff, and they’ll do it cold. learn to like being minimal — it makes your whole answer cleaner and harder to poke holes in.
i got roasted once for saying 'high demand' w/o a source. now i use google trends or a competitor's revenue as proof — small wins!
wow this thread convinced me to drop vague words — trying proxies next mock, thanks guys :)
Veteran feedback often seems harsh because it focuses on falsifiability: can the assumption be disproven with a simple check? I train candidates to make assumptions testable. Replace adjectives with metrics (e.g., ‘high’ → 20% penetration) and pair each with a one-sentence rationale and a rapid evidence source. This habit reduces interviewer pushback and signals judgment. Start with two proxies you can cite without searching; expand them as you practice.
i remember getting a harsh note from a vet: ‘pick one number and defend it.’ i switched from fluffy phrases to a single measurable assumption and a competitor benchmark. in the next mock, the interviewer asked fewer follow-ups and the whole conversation felt tighter. the vet’s bluntness was exactly what i needed to stop hiding behind words.
in my experience, the most resilient proxies are those with public footprints: competitor revenue, app store downloads, google trends, and census data. converting qualitative claims into a single numeric assumption plus a data source reduces variance in panel grading. empirically, estimates anchored to a comparable company’s public metric diverged less from ‘expected’ answers in post-interview reviews than those anchored to intuition.
operational rule: for each driver, specify (1) the metric, (2) why it maps to behavior, and (3) a one-line validation. for example: ‘penetration = 5% of urban smartphone users; validated by 2 similar apps with combined 1.2M installs in the city.’ that format survives tough cross-examination and is repeatable across industries.