Marketers have understood something for decades that job seekers are rarely told: when a process has stages, the only useful question is which stage is leaking. Pouring more traffic into a funnel with a broken middle is how you burn a quarter and hit no targets.
Your job search is a funnel. Let us put actual numbers on it.
The five stages, with real conversion rates
Data from Ashby, an applicant tracking system that has analysed 100M+ applications across 200K+ jobs, plus widely-reported market benchmarks:
- Stage 1 — Applications sent. Roles now attract 300+ applications per hire, roughly triple 2021 levels. In Ashby's data, 93.8% of applications are inbound, with only about 1% arriving via referral and about 4.8% sourced.
- Stage 2 — Application to interview. For an inbound applicant, this is the bloodbath: roughly 3-5% convert. Ashby found candidates today are about 50% less likely to reach an interview than five years ago. Meanwhile 40% of referred candidates and 42% of internal candidates get interviews.
- Stage 3 — Interview to offer. Once a human is talking to you, the odds transform. Overall the rule of thumb is about 1 in 10; Ashby measured 16% for interviewed referrals, 32% for internal candidates, and 8% for agency candidates.
- Stage 4 — Offer to acceptance. Usually 80-90%. Most people who get an offer take one.
- Stage 5 — Started and stayed. Not your problem today, but it is why interviewers are so cautious.
Run it end to end for a typical inbound search: 100 applications → about 4 interview processes → about 0.4 offers. Which is a statistically elegant way of saying 100 applications is not a lot of applications, and also the problem is almost certainly stage 2.
The two numbers that should change your behaviour
3-5% at the top. 1 in 10 in the middle.
Notice how differently those two numbers behave. The top of the funnel is a volume game with terrible odds and almost no feedback. The middle is a small-numbers game with respectable odds and rich feedback — you find out what they are worried about, you can answer it, you can adjust.
The consequence: effort spent converting stage 2 is worth vastly more than effort spent expanding stage 1. Doubling your applications from 40 to 80 is a week of soul-flattening copy-paste. Doubling your application-to-interview rate from 3% to 6% has the same arithmetic effect and is achievable through tailoring, which is boring but not soul-flattening.
Also notice the referral number, because it is the loudest signal in the dataset. Referrals are 1% of applications but convert to interviews at 40% — roughly a tenfold advantage over inbound. One conversation with someone inside the company is worth, in expected value, something like thirty cold applications. It is not fair. It is simply true, and it is free to act on.
Where funnels actually leak
Leak 1: applying to the wrong stage of role
If you are getting almost no interviews but you keep reaching final rounds when you do, your targeting is fine and your volume is low. If you get interviews and always lose the final round, your positioning got you into rooms you cannot yet win. Different diagnosis, opposite fix.
Leak 2: an untailored document
This is the big one. Generic resumes fail not because a machine rejects them, but because a human cannot tell in four seconds what you are. Recruiters and filters both match literal language, so the fastest available improvement is rewriting your title, summary and top three bullets in the vocabulary of the specific posting.
Leak 3: a formatting or parsing failure
Two-column layouts, text embedded in graphics, exotic fonts with broken character maps. Nobody rejects you here; you simply arrive as gibberish. Send yourself a plain-text extraction of your own file once and read it. It takes two minutes and occasionally saves a search.
Leak 4: mistaking "no direct experience" for "no relevant experience"
Most people fail a requirement they actually partially meet, because they only counted the literal version. Never worked in banking, but ran a product sold into three retail banks? That is adjacent, real, and — stated honestly and specifically — often sufficient. Lying about the requirement is disqualifying. Reframing true experience so its relevance is visible is just competent writing.
Leak 5: treating volume as effort
Twenty applications with a tailored first page will beat a hundred identical PDFs, and take less of your life. The hundred-PDF strategy mostly generates evidence that the market is cruel, which is emotionally satisfying and strategically useless.
Running your search like an operator
- Track four numbers weekly: applications sent, interviews booked, processes reached final stage, offers. Nothing else.
- Set a floor, not a ceiling, on quality. Ten tailored applications a week is a real pipeline. Fifty untailored ones is a hobby.
- Spend 20% of your time on referrals. Given a tenfold conversion advantage, an hour of outreach outperforms an evening of applying.
- Diagnose from the ratios, not from your mood. Job searching feels like continuous rejection even when the funnel is working normally. At 4% application-to-interview, twenty silent rejections is not a signal. It is Tuesday.
- Expect the timeline. With these ratios, a decent search runs 2-4 months. Planning for three weeks is how people conclude they are unemployable in week four.
The honest closing note
The funnel is genuinely harsher than it was five years ago — Ashby's own data says so. But the shape of the harshness is oddly encouraging: the market has become more selective about who it interviews and more decisive about hiring the people it does interview. Offer conversion has risen above 2021 levels even as interview access halved.
Which means the game is almost entirely about getting into the room. And getting into the room is a writing and targeting problem — the one part of this process that responds directly to effort.
Fix stage 2. The rest of the funnel is friendlier than its reputation.