Stop buying the same candidates twice. Fill from the database you already paid for.
Most staffing databases hold tens of thousands of workers who were screened, worked a few shifts and went quiet. Every new job-board lead you buy is often one of them, at full price. The Sourcing Engine fills from what you own first, and spends only on the gap.
What this work looks like today, and what changes.
Today, per open order
- Order comes in. Coordinator posts the job on two or three boards and waits.
- Leads arrive over several days. Many are already in your ATS under a slightly different phone number.
- Recruiter screens them again from scratch.
- Workers who did well last quarter are never contacted because nobody has time to search the database.
- Referrals happen only when a worker thinks of it.
- Cost per hire is tracked as a total, never as "how much of this did we already own?".
With the Sourcing Engine
- Open order lands. The engine matches it against your whole database: skills, distance, past sites, last worked, availability, attendance history.
- Matched workers get a personal message on the channel they answer, at the hour they read it, with the actual shift and pay.
- Interested replies are pre-screened by the AI Recruiter and appear as a shortlist.
- Only the remaining gap goes to job boards, with the posting written for that gap.
- After a good shift, the worker is asked for a referral, with a link that credits them.
- Duplicate leads are merged on arrival, so you never pay twice for one person.
Step by step: what goes in, what the automation does, where a person decides.
Green steps run on their own. Amber steps are where a human approves or decides, by design.
From the Client Desk, your ATS or a coordinator. Role, count, site, dates, rate.
Score every worker on fit: certifications, distance, past sites, attendance, availability, last contact. Exclude anyone who opted out.
AutomatedWhatsApp, SMS or a short voice call in the worker's language, timed to when they historically read messages. Real shift, real pay, one-tap interest.
AutomatedInterested workers flow into the AI Recruiter to confirm availability and documents. Still-valid records skip the questions already answered.
AutomatedIf the shortlist is short, the engine posts to boards with a description written for the missing profile, and dedupes incoming leads against the database.
AutomatedWorkers who know your sites, at a fraction of acquisition cost, plus a referral loop that compounds every month.
One screen, three panes: the input, the reasoning, the result.
| Worker | Fit | Last shift |
|---|---|---|
| Kwame A. | 14 d | |
| Fatima S. | 31 d | |
| Joseph M. | 62 d | |
| Grace O. | 9 d | |
| +57 more | worked this site or within 15 km | |
Fit uses skills, distance, past sites and attendance. Never age, gender or nationality.
- 38 messaged, 14 interested, 9 confirmedown
- Gap: 3 · posting to 1 board, forklift-certified onlygap
- 2 duplicates from board merged into existing recordsdeduped
Board spend covers three seats, not twelve.
A composite fill. The point is the order of operations: own database first, boards for the gap, and every lead deduplicated against what you already have.
Where the advantage comes from, step by step.
| Step | Today, by hand | With ShiftMargin | Why it matters |
|---|---|---|---|
| Finding candidates | Post to boards. Wait. Screen strangers. | Search your own database in seconds, scored on fit and recency. Boards only for the gap. | You already paid to acquire and screen these people. |
| Reaching out | A mass SMS blast, or nothing. | Personal messages with the real shift and pay, in the worker's language, timed to when they read. One-tap reply. | Relevance and timing are what get replies. Blasts train workers to ignore you. |
| Re-screening | Full screen again, as if new. | Only what changed. Valid documents and known availability are skipped; expired ones are asked for. | Faster for the worker, cheaper for you, and it respects their time. |
| Duplicates | Same person, three records, three fees. | Merged on arrival by phone, ID and name similarity, with a human check on ambiguous merges. | Paying twice for one worker is the most common invisible leak in acquisition. |
| Referrals | Ad hoc, when someone remembers. | Asked after a good shift, with a tracked link and an automatic credit. | Referred workers show up and stay. The ask has to be timely to work. |
| Reporting | Cost per hire as one number. | Own-database fill rate, cost per lead by source, board spend avoided. Weekly. | You can only cut acquisition cost if you can see where each hire came from. |
The components, in plain language, and why each one matters to you.
Everything here is chosen so your ops team can maintain it after we leave. The only pieces we keep maintaining are the ones that move money.
Skills, distance and travel time, past sites, attendance record, last contact and opt-outs. Explainable weights you can see and change. No demographic inputs.
Through a regional provider for the best rates, with per-country templates. Service replies inside the 24-hour window become billable from October 2026; the cost model already assumes it.
Personalised from the order and the worker record, translated, then checked against your tone rules. Costs a tenth of a cent per message.
Learned from when each worker historically replies. Quiet hours by country are enforced.
Phone, national ID and fuzzy name match. Confident merges are automatic; ambiguous ones queue for a person.
Bullhorn, zvoove or your system via API, or scheduled import. n8n handles the glue so your ops team can see and edit the flow.
What stays human, and what happens when something goes wrong.
A worker who says stop is never contacted again by any channel. That flag lives in your ATS, not in our system.
Fit is built from work history, skills and distance. Age, gender, nationality and similar fields are excluded from the model and from the message.
Messages respect local quiet hours and channel consent per country. Voice only where the worker has agreed to calls.
We never buy leads on your behalf. The engine recommends what to post and where; your team approves spend.
What it costs by hand, what it costs to run, and where the margin goes.
typical cost per lead in high-volume staffing markets, before the recruiter's time to screen it. Often for someone already in your ATS.
per WhatsApp message by country, plus a fraction of a cent for the model. A 40-message campaign that fills nine seats costs about the price of one coffee.
Job-board prices vary widely by market and role; we measure your actual cost per lead by source in week one and report against it weekly.
Six weeks to production, behind a manual fallback.
Cost per lead by source, share of hires already in the database, duplicate rate, referral rate. Usually the most surprising two weeks of the engagement.
Matching model over your ATS, messaging templates per country, dedupe rules, referral flow, board-posting recommendations.
Engine fills alongside your usual sourcing. We compare fill speed and cost per seat order by order.
Own-database fill rate, board spend avoided, cost per lead. The referral loop keeps compounding.
Before they let this near the business.
Our database is a mess. Old numbers, duplicates, no tags.
That is the normal starting point. Dedupe and enrichment are the first two weeks of the build. Even a messy database usually holds more fillable workers than a month of job-board spend.
Will workers find re-engagement spammy?
Not if it is relevant and timely. One personal message with a real shift and real pay, sent when they usually read messages, with a one-tap reply and an absolute opt-out. Blasts are what workers hate, and we do not send them.
Does this replace job boards entirely?
No. It reorders them. Boards are for the gap your database cannot fill, and the posting is written for exactly that gap. Most operators see board spend fall by half without losing fill rate.
How is a referral tracked and paid?
Each worker gets a personal link after a good shift. Referred candidates are tagged on arrival and the referrer is credited automatically once the referred worker completes an agreed number of shifts. The credit rules are yours.
Start with your number, not our demo.
Bring the volumes and the time your team spends on this workflow today. In thirty minutes we will tell you whether it is your most expensive line, and if it is not, which one is.
Book a free 30-minute teardown →