What the work involves

You'll spend most of your time on two loops. In the first, you write the gold-standard artifact: given a brief and a dataset, you build the readout deck a real client would receive — the storyline, the exhibit choices, the chart types, the headline callouts that actually say something rather than restating the axis label. In the second, you grade what the model produced against that standard, and your written rationale matters as much as the score. Task families rotate across research readouts, survey results exhibits, segmentation presentations, tracker reports, qualitative findings storytelling, and insight-to-action recommendations.

The judgment being harvested is the part of insights work that never gets written down: why a stacked bar beats a pie for brand funnel data, when a significance marker belongs on the slide versus the appendix, how a tracker deck should lead with movement rather than levels, and what separates a finding from a recommendation. Expect to justify calls like base-size disclosure, weighting footnotes, and whether a segmentation portrait is telling a story or laundering a cluster label.

What the platform screens for

Ethos runs an AI voice screen. It probes for concrete tenure — where you did research, what categories, whether decks went to a client or an internal exec committee — and then pushes on method depth: significance testing, weighting, sample composition, tracker wave comparability, qual coding. It also tests evaluation temperament: whether you can articulate why a slide fails in language another reviewer could act on, rather than saying it feels off. Vague agency-speak and unverifiable claims about "C-suite exposure" get followed up hard.

Logistics

  • Fully remote, fully async — no standing meetings, no client calls
  • 5–20 hours per week, self-scheduled; higher volume available on some task queues
  • $70/hour observed on this listing; rates vary by queue and are not guaranteed
  • Work is delivered through the platform's task interface; PowerPoint fluency is assumed, not taught