The recruiting funnel: stages, and where each one breaks
Six stages, and each one fails in a specific way. How to find the one leaking on your own numbers, and which two a sourcing tool can actually touch.
A funnel is only useful if you can name the stage that is leaking.
Most teams have a number for the top and a number for the bottom, and nothing in between. That tells you something is wrong and gives you no idea what.
Six stages. Each fails in its own way, and each has its own fix.
We are not going to hand you benchmark conversion rates.
We have no source we could stand behind, and a made-up benchmark makes people optimise toward a number that means nothing.
What works better anyway: your own funnel, measured twice. The first measurement is the baseline, the second tells you whether the change worked.
This page covers the six stages, the failure signature of each, and which two a sourcing tool actually touches.
TL;DR:
- Six stages: sourced, contacted, replied, screened, interviewed, offered.
- Each has one characteristic failure. Name it before you fix anything.
- Only two are sourcing. The other four are process and offer.
- The biggest drop is your problem. Everything else is noise.
- A benchmark you cannot source is worse than no benchmark.
Key facts:
- Kalent database: 850 million profiles, 250 million enriched with verified contact details
- Share of returned profiles directly reachable: around 80%
- Automated channels: email, LinkedIn, SMS, WhatsApp, on Copilot
- Kalent average placement: 10 days
- Price: $149 and $179 a month, per licence
The six stages
| Stage | The count | What it fails on |
|---|---|---|
| Sourced | profiles you would genuinely send | a brief that is too narrow |
| Contacted | of those, how many you could reach | missing contact details |
| Replied | of those, how many answered | one channel, or a generic first line |
| Screened | of those, how many were actually a fit | a ranking with no stated reason |
| Interviewed | of those, how many reached a manager | scheduling, and the silent gap |
| Offered | of those, how many said yes | the range, the level, a faster competitor |
Fill this in for the last three roles you closed. One role is an anecdote.
Stage 1 and 2: sourced and contacted, the ones you control
These are the two a sourcing tool actually touches, and they are usually the two nobody measures.
Sourced fails when the brief is narrower than the market. Nine requirements, of which three are real. Or when the brief uses the company's vocabulary and candidates do not.
The Kalent agent searches wider than the brief, applies your qualification questions rather than filters, ranks, and writes next to each profile why it is there.
Contacted fails on contact details. A profile with no verified number is a research result, not a candidate. Kalent publishes 250 million enriched profiles of 850 million, with around 80% of returned profiles directly reachable.
Stage 3: replied, and the two reasons it does not
One channel, or a first line that could go to anyone.
At FAB Group, 50 consultants, 98% of approaches left as InMails and the reply rate settled at 15%. Four automated channels took it to 30%, across more than 18,000 qualified profiles in 12 months.
The first line is the second half. The reason for the match, written when the profile was ranked, goes into the opening sentence.
"I came across your profile" is the most sent and least personal sentence in recruiting.
Stage 4: screened, where a bad shortlist shows up
If most people you screen turn out not to fit, the shortlist was never a shortlist.
This is where an unexplained ranking costs you. A tool that returns five hundred results in an order with no reason has moved the sorting into your screening calls.
A stated reason per profile lets you drop the wrong ones before the call, and lets your hiring manager agree or disagree in writing.
Stage 5: interviewed, the calendar stage
Three things account for most of the loss here, and none is a sourcing problem.
Scheduling. It belongs to nobody, so it takes a week. Give it one owner with the authority to book.
Round count. Five rounds over six weeks loses people to whoever ran three.
The silent gap. A candidate who hears nothing for eight days is being interviewed by someone else.
Stage 6: offered, and what a no actually tells you
Write down the stated reason every time someone declines.
The range, the level, the manager, the commute, a competing offer that moved first. Five roles is enough to see the pattern.
And the rule that saves the most money on this page: if candidates reach the offer and say no, stop working on sourcing. More people into the same offer produces more of the same answer.
How to read your own funnel
- Fill the six counts for three closed roles.
- Compute the drop between each pair of stages. The biggest one is your problem. Ignore the rest for now.
- Apply that stage's fix, and only that one.
- Re-measure on the next three roles.
- Repeat. Two measurements beat any benchmark, because they are about your market and your offer.
Resist fixing two stages at once. You will not know which one worked, and next quarter you will be guessing again.
Related reading
The AI sourcing comparison cluster, all read on the same date against the same criteria:
On the same subject: how to reduce time to hire, and recruiting for hard-to-fill roles.
On the definitions: 10 recruiting metrics worth tracking.
And the product pages: pricing, ATS integrations, trust center.
Frequently asked questions
What are the stages of a recruiting funnel?
Sourced, contacted, replied, screened, interviewed, offered. Each fails in its own way, so naming the leaking stage matters more than the totals at either end.
What is a good conversion rate at each stage?
We do not publish benchmarks, because we have no source we could stand behind. Measure your own funnel twice instead: the first measurement is the baseline, the second tells you whether the change worked.
Which stages can a sourcing tool improve?
Sourced and contacted, plus the reply rate through channels and a better first line. Screening, interviewing and the offer are process and compensation.
What if candidates drop out at the offer?
Stop sourcing. More people into the same offer produces more of the same answer. Write down the stated reason every time someone declines.
Should I fix more than one stage at a time?
No. You will not know which change worked, and next quarter you will be guessing again.
