Placement sourcing

Overall median lag 39 days. Of traced placements: 44% placed within 30 days of first touch, 23% in 31–90 days, 14% in 91–180 days, and 19% took more than 180 days. So most placements in any given week were sourced months earlier — the marketing that drove them is not this week's spend. ZipRecruiter and Meta convert fast (median <8 weeks); Indeed and LinkedIn are slow (median 4–10 months), so a same-week efficiency read understates the slow-maturing channels. ~4% of traced placements show a placement date before the earliest traced application (data quirk) and are excluded from lag. Lag = placement week − first-application date.

Blended-mix warning: these board rates blend ALL disciplines, and board job mixes differ sharply — SLP-family traced placements in the matched windows: Indeed 38 vs ZipRecruiter 3 — so do not move discipline-specific budget (e.g. SLP) on these blended numbers; use /pivot's role rows. Touch-priced: each touch priced at its board's cost-per-application over the window (Indeed $22, ZipRecruiter $21/apply; Google Ads at cost-per-click $2.60 as an apply proxy; organic $0). It deliberately excludes spend on applicants who never placed, which is most of the spend — use it as a floor, never as total cost. Matched $ / placement divides a board's full paid spend by its traced placements over the SAME weeks (numerator and denominator never cover different week sets). Feed coverage differs by board: the Indeed spend feed starts 2026-06-08, so its matched window is 7 weeks vs ZipRecruiter's 7 — boards are each self-consistent but their windows differ. Google Ads spend is shown but excluded from every blend: its placements are structurally untraceable (first-touch strings put them under Google organic/website), so charging its spend with no creditable placements would distort the blend. Timing caveat: placements landing in a week were mostly sourced months earlier (median 39d), so these are run-rate ratios (spend rate vs placement rate), not cohort costs. Meta & LinkedIn spend is not yet in either view. Cost covers traced placements only (~42%).

Priced on actual billing, at discipline × assignment type. Rates come from `clockb_wow.disc_rev_by_type` — mean gross billing over a full calendar year of settled placement start months (2025-01–2025-12), keyed on the placement's own discipline and assignment type, falling back to the discipline rate then the overall blend ($38,217 across this page's traced mix). This replaces the flat $41,721 an earlier version used, which is retracted. Two reasons the flat value was wrong here: revenue per placement is not constant (the planning model alone holds 12 brand-level values, $30,032–$49,366, and realized billing varies 4.6–11x by discipline), and this page is anchored on the placement week, where the assignment-type mix swings hard by season — in the week of 2026-06-01, 82% of placements were Summer/ESY, and ESY bills roughly 8x less than Traditional inside the same discipline (Paraprofessional $33,031 vs $4,331; SLP $71,748 vs $7,038). A flat rate overstated the summer peak several-fold. Two caveats remain: a settled per-discipline rate is an annual average that cannot express the ~2.6x swing by school-year start month (a July start bills a full year, a March start four months), and the numerator's placements were mostly sourced months before this window's spend (median 39d lag), so window ROAS is a run-rate ratio, not a cohort return.