What the work actually involves

You are producing and judging business reasoning at expert level. A typical assignment looks like one of these: write a prompt asking for a bottom-up TAM estimate in a niche vertical, then supply the reference answer with your assumptions and sources laid out; take three model attempts at a working-capital model and rank them, writing out exactly where each one broke; audit a model's process-improvement recommendation against what an operations lead would actually be able to implement given headcount and system constraints. Volumes come in batches, and the same rubric usually governs dozens of items so your early calibration decisions compound.

The hard part is rarely knowing the answer. It is writing down why an answer is wrong in language another reviewer can apply consistently — distinguishing a wrong number from a wrong method, a defensible assumption from a fabricated one, and a genuinely useful business recommendation from one that merely reads like a consulting deck. Models are fluent enough now that surface plausibility tells you nothing; the failures hide in double-counted revenue, invented benchmark figures, and frameworks applied without checking whether the premises hold.

What the screen looks for

  • Verifiable operating history: named employers, the scope you actually owned, the decisions your analysis fed into
  • Real fluency with financial and operational modelling — you can be asked to reconstruct a build-up or spot a broken assumption chain on the spot
  • Judgment under a rubric: can you hold a scoring standard across items instead of drifting toward whichever answer sounds smartest
  • Written clarity, since your rationale is the deliverable a downstream reviewer inherits

Logistics

Fully remote and asynchronous, with work claimed from a queue rather than scheduled. Most contributors commit 10–20 hours a week; some tracks require a short calibration set and a review period before full rates apply. Assignments are commonly English-language, and rates are agreed per-project at onboarding rather than fixed across the platform.