People & Exec SupportFree
Recruiting Funnel Analysis From Your ATS
Whether the role is stuck on sourcing or on closing, worked out from your ATS export rather than argued in a staffing meeting.
River's recruiting funnel analysis reads the export out of Greenhouse, Lever or Workable and rebuilds the pipeline from the rows themselves. Back comes a sheet carrying conversion, median time in stage and source for every role in the file, a funnel chart per role, and a short written diagnosis naming the one stage worth working on. The first thing it repairs is the denominator, because the number of rows in an export and the number of applicants for a role are different numbers.
Unlike the funnel spreadsheets that rank for this query, it will not compute a rate on a stage that still holds live candidates. A req open five months has people sitting mid-process, and counting them in a denominator with no outcome attached makes every late stage look worse than it is. Rates here come from candidates whose outcome is settled, and the unresolved ones are reported separately with their age in days. Pinning what a rate even counts is the same discipline the metric definitions pack applies to reporting.
Written for in-house recruiters running several reqs at once, talent leads asked to explain a slipping plan, and the hiring manager who has been told for five months that the market is tight. Reach for it when a role has been open long enough that people have stopped asking why. When the diagnosis points at screening, resume screening on evidence is the next run; when it points at the requirements, the job description pack is where they get rebuilt.
Most recruiting funnels are read from the wrong end
Rank the stages of any hiring funnel by conversion rate and the application-to-screen step comes first every time, because applying is free and screening is not. It is the largest loss in almost every pipeline and almost never the thing to fix. Product analytics learned the same lesson and now ranks by people lost rather than by rate, which is the argument the product funnel analysis makes. In hiring the useful version is narrower still: where does a candidate who met the advertised bar get lost.
That question needs a denominator, and the raw export does not supply one. The Internet Applicant rule sets four conditions. The person expressed interest. The employer considered them for that position. They met the basic qualifications advertised in advance. They never withdrew. Run those four over a real export and the row count falls. In one worked file, 412 rows on a req became 322 applicants once 24 sourced profiles, 47 people below a posted requirement and 19 withdrawals came out.
The diagnosis turns on one distinction. A sourcing problem is a volume problem: fewer qualified people arrive and every downstream rate holds. A closing problem is a rate problem at the end: volume holds and offers get declined. In the same file, 322 applicants arrived over 21 weeks while two other reqs filled from 210 and 96. Median days from hiring manager interview to work sample were 19 against 4 elsewhere, and two of three declined offers had waited 61 and 74 days. That is a calendar problem wearing a sourcing problem's clothes.
How it works
Send the export
Greenhouse, Lever or Workable, with stage dates and dispositions. Several reqs in one file is better than one.
The denominator gets fixed
Sourced profiles, applications below the posted bar and withdrawals come out, and every exclusion is counted.
Read the diagnosis
One stage named, whether its volume or its rate moved, and the days a scheduling change would return.
Argue with it
Move the qualification bar and see who returns, or run the same cut against a req that filled.
What you get
- The applicant denominator rebuilt against four conditions, with every excluded row counted and explained
- Conversion by stage on candidates whose outcome is settled, with the live ones listed separately
- Median and 90th percentile days in each stage, so one stalled candidate cannot set the number
- A funnel chart per role, plus the same chart for roles that filled, as the comparison
- Source performance measured at the offer accepted step rather than at application volume
- A written diagnosis naming one stage, whether volume or rate moved, and what changes
Common questions
My ATS already ships a funnel report. What does this add?
A denominator you can defend and a diagnosis. Built-in reports count rows and divide, which mixes sourced profiles, unqualified applications and withdrawals into the same number, then compute late-stage rates over candidates who have not finished. This rebuilds both, then says which single stage to work on and whether its volume or its rate is what moved.
How does it tell a sourcing problem from a closing problem?
By checking whether volume or rate moved, separately. Qualified arrivals per week against the same company's other reqs answers the first. Offer acceptance, days in process for the candidates who declined, and reject reasons at the final stages answer the second. When arrivals hold and acceptance falls, more sourcing makes the wait longer. Where the delay is candidates awaiting a reply, the candidate communication pack counts who is overdue.
Half my candidates are still in process. Does that break it?
No, they get their own section. Every rate is computed on candidates whose outcome is settled, and the live ones are listed with their current stage and their age in days. That list is often the more useful output, because a candidate sitting at one stage for six weeks is a decision nobody has made rather than a number.
What exactly do I need in the export?
One row per candidate with the requisition, the source, the current or final stage, the dates each stage was entered, and the disposition. Rejection reasons help and blanks are fine. Six months across several reqs beats one req in isolation, because a single funnel has nothing to be compared against.
Can I use this to compare recruiters or hiring managers?
It will show throughput and days in stage by owner, and it will say plainly when the counts are too small to mean anything. A recruiter with nine candidates through a stage has not been measured. Where the numbers are real, they get reported as a question to ask rather than as a verdict about a person.
Do I need to keep any of this after the role closes?
Keep the underlying records regardless. Under EEOC recordkeeping rules, application forms and other hiring records are preserved for a year from the record or the personnel action, whichever is later, and longer once a charge is filed. The analysis sits on top of data you already have to hold on to.
Recruiting Funnel Analysis From Your ATS
Fill in the form and your workspace opens with the work already underway.