Why volume stopped working, on both sides

An arms race with a predictable end

Before 2023, a tailored application cost an hour, so people sent few and employers received few. When drafting one dropped to about thirty seconds, applications per posting rose sharply: LinkedIn and several large job platforms have reported application volume growing far faster than the number of jobs, and single postings for ordinary remote roles now routinely draw several hundred to several thousand candidates.

Employers responded exactly as you would expect if you had thought about it for a minute:

• More knockout questions on the form, which reject automatically.
• More take-home exercises, earlier, to impose a cost that a generated application cannot pay.
• Heavier weighting on referrals, because a vouched candidate skips the pile.
• Postings closed early, sometimes within 48 hours, once enough applications arrive.
• Fewer replies, because the arithmetic of replying to two thousand people does not work.

The net effect is that the cold conversion rate in the funnel lesson fell, for everybody, at once. Nothing about you changed.

The maths, since it is the whole argument

If cold conversion was 4% and is now 1.5%, and warm conversion held at 30%, then one warm approach is now worth twenty cold applications rather than seven. Every hour you spend automating volume is spent on the side of the ledger that got worse, while the other side got relatively better.

This is not a moral point about effort. It is division, and it is the reason this module sits after the one on getting seen rather than before it.

Auto-apply services, specifically

Products that fire hundreds of applications on your behalf are sold on exactly the wrong model of the problem. Three concrete costs:

• They apply to roles you do not fit, because they match on keywords. Your application history is stored in that company’s tracking system, and a recruiter who sees six unsuitable applications from you in a year reads it as noise rather than enthusiasm.
• Some systems detect and block them, through submission-rate patterns or form telemetry.
• They spend the hours you had. The two hundred applications feel like work; the five people you did not contact are the search you did not run.

What replaces it

A weekly quota that a working person can actually hold:

• Five real applications, tailored in the twenty-minute way, to roles you can defend wanting.
• Five direct approaches: a named person, a specific reason, one piece of evidence attached.
• One piece of public work: a write-up, a merged pull request, an answer to somebody’s question.
• Thirty minutes updating the tracker and reading what it says.

Eleven items a week. Roughly six hours. Sustained for three months, that is 60 applications, 60 approaches and 12 published artefacts, and on the conversion rates in the funnel lesson that produces enough conversations to reach offers. Two hundred auto-applications a week for the same three months produces a full inbox and no interviews.

Where volume does still work

Being honest requires naming the exceptions, because they are real and they matter for exactly the readers this course is written for.

• Large services and outsourcing firms hiring in batches. These processes are built for volume, run standardised aptitude tests, and applying widely is the correct strategy.
• Graduate schemes and campus intakes, which open in fixed windows with fixed processes.
• Government and public-sector recruitment, where the process is formal, scored against published criteria, and personal routes do not help.
• Agency-driven contract markets, where recruiters place people at speed and being in many agencies’ databases is straightforwardly useful.

In those channels, apply to everything that matches. Outside them, volume is a way of feeling busy.

The limitation

Nobody knows where this settles. Employers are experimenting with in-person stages, live coding, work trials and stricter referral policies, and some of those are worse for candidates without networks, which is most people reading this. The reasonable response is to build the one thing that survives every version of the process: a small number of people who have seen your work and will say so.