Cost per candidate and spin, by campaign. Based on our own hiring system (ATS), not Google's conversion count. The gold columns are real placements credited to the campaign.
from to
How to read the placement columns
Sourced plc counts REAL placements (billable pla_status) in the week the person’s Google lead arrived, so a window’s spend and its placements are the same people. It takes about 6 months from apply to placement, so recent windows read low or zero and grow at each refresh. A low recent number means too early, not bad. Hover a cell for that campaign’s total so far this year.
Placed counts real placements that closed in the window, no matter when the person first applied. Use it for what closed. It is never divided into spend, so Cost / placement and ROAS pair with Sourced only.
Real, not made. A placement counts once its pla_status is billable. Made (all statuses) is shown beside it as a secondary column: on older groups about 45% of made placements never start, and recent ones are still resolving, so made runs ahead of real.
Billed is billed gross to date from , the same file /compare reads. Billing runs about 4 months behind, so a recent placement can read $0 billed while still counting. Read it as a floor. Billed ROAS = billed / spend. Modeled ROAS is a model (discipline revenue rate x real placements / spend), not billing.
Candidates are distinct people per week with an apply to a real job or an intake touch (a lead form, resume pull or site resume form with no job) on the campaign, summed over the weeks in the window, the same basis as /compare. Someone active in two weeks counts in both. Intake counts those no-job touches; they are candidates, never applies.
Spins on two bases, never blended.Spins is today’s basis for this page: the spun lead’s own campaign, dated to the week of the candidate’s most recent apply (else intake touch) on or before the spin, the rule /compare uses, one per candidate per week. A spin with no such touch in this year’s weeks stays on its lead week. Spins · last paid touch credits each spin to the candidate’s most recent paid lead on or before the spin, which can be another campaign here or another platform (then it leaves this page). Same people, same NMQ rule.
Spins = a recruiter actively worked the candidate (our “qualified”). The newest week undercounts a little, which is what the “exclude latest week” toggle is for.
Weekly spend and spins per brand on the same weeks. Each line has its own scale. The dashed tail is the newest weeks, which are still filling in because it takes about 6 months from apply to placement, so recent placed points are incomplete.
Weekly trend by channel
Full year, real spend per channel. Dashed tails are still settling: the newest spend and lead weeks lag a little, and recent placements undercount for a month or two while Pending and Credentialing statuses resolve.
How to read this view
Spins = recruiter spin events (our “qualified”), matched to a discipline by the candidate’s own skill.
Placements here credit each person once, to the channel of their earliest application anywhere, by the week they were placed. That compares channels; it is not this window’s spend efficiency, since these placements came from people sourced months earlier.
Spend by discipline is exact where the platform reports it (Indeed jobs, Zip categories, named campaigns). The rest is split by spin mix, so grey values are mostly estimated: read them as the channel’s blend, not a per-discipline price. Channel totals are real.
Google paid vs organic is split by matching applies to known paid campaign tags, so it is a floor: some “Google organic” placements are really paid.
The multi-brand Shared account account is its own row, “Shared account (multi-brand)” (about $700 a week), the same row /compare carries: it can’t be credited to any single brand, so its spend, spins and placements sit there and in no brand.
Applies = every application; candidates = distinct people (about 93% of applies). Both come from the same weekly data /review uses, so the pages agree.
On a discipline row, applies and qualified are counted differently: applies by the job the person applied to, qualified by the candidate’s own skill. So a discipline can show more qualified than applies (a recruiter worked them as an SLP even though they applied to a Para job). Read them as two views of the row, not a ratio. Brand and platform rows are not affected.
Two cells show spend and qualified but no applies on purpose: Google paid applies can’t be separated from organic in our system, and Brand D barely appears as a department there even though the warehouse shows it spinning leads.
The multi-brand Shared account account is its own row, “Shared account (multi-brand)” (about $700 a week), the same row /compare carries: it can’t be credited to any single brand, so its spend, spins and placements sit there and in no brand.
Leads = worked leads (undercounts raw applies). Qualified = spins. Placed = first-touch placements by placement week; recent weeks undercount until billing posts, so read Placed from the sparkline shape, not the last cell.
What changed or broke, and when. Entries are found in the data automatically, not from memory: a detector watches job counts, knockout rates, apply volume, spend, and Google conversion tracking, and logs each shift with the before and after and the source. Each one shows as a dashed marker on the trend charts. You can also add your own note below, kept separate and never counted as a detection.
Add an annotation
For context the data can't show. Stored as an annotation,
labelled unverified, and kept out of the detected list.