Recruiting Source Effectiveness
Which recruiting sources send you people who stay, rather than which ones send you the most people.
Reads .xlsx and .csv inside your browser. Nothing is uploaded.
The hires
Drop a spreadsheet here
One row per hire, with where they came from, when they started, and when they left if they have. Leavers are the whole point, so leave them in.
Counting hires rewards volume
The usual source report counts hires, and the source at the top is the one everybody names as the best. But a source that fills a third of your roles with people who are gone by the spring is filling some of those roles twice. What a source is worth is the people it sent who stayed.
So every source here is reported twice: its share of your hires, and its share of the hires who were still there a year later. A source whose second number is smaller than its first is supplying more of your turnover than of your workforce.
“Still employed today” is the wrong retention number
The obvious fix is to report the share of each source’s hires who are still employed. That flatters whichever source you have used most recently, because its hires have not had time to leave. A job board you started using in March looks remarkably loyal in June.
This tool judges each source only on hires from more than a year ago, every one of whom has had the chance to reach a first anniversary, and shows the flattering figure beside the honest one so you can see how far apart they are.
Small sources lie in both directions
Two referrals who both stayed is a hundred percent. Four agency hires who all left is zero. Neither means anything. So a source is only compared with the rest once it has enough hires to compare, and the comparison is an exact statistical test rather than a glance at two percentages.
Because every source is tested at once, a correction is applied: look at enough sources and one will seem unusual by chance. Findings come in two tiers. A source that stands out survives the correction. One that is worth a look would be unusual on its own but not once everything is tested together.
One source, written four ways
A source column is usually typed by hand, so “LinkedIn”, “linkedin” and “Linked In” turn up as three separate small sources. Names that differ only in case, spacing or punctuation are joined automatically and the joining is reported. Names that merely look related, like Referral and Employee Referral, are offered as a suggestion and only joined if you say so.
Cost per hire, and cost per hire who stayed
Type in what each source cost you and the table works out both. The cheapest source per hire is often not the cheapest per hire who stayed, because a cheap hire who leaves in the spring gets paid for twice. For the fuller cost picture, the Cost per Hire Calculator separates recruiting costs from all-in costs, and the Recruiting Funnel Analyzer covers what happens before the hire.
What this does not do
Staying is not the same as being good at the job. A roster records who stayed and who left, so that is what this measures, and it says nothing about the people any source sent. The source is also whatever your system recorded, which is usually where the candidate applied rather than where they first heard of you.