The funnel you can actually read
There is a published number that describes this programme and, by accident, describes the search every applicant is running: 25,000 applications for 25 places per cohort. That is one place per thousand applications. Around 0.1 per cent.
Most people read that as a reason to feel discouraged. It is more useful read as a description of a machine. A programme that receives 25,000 applications and admits 25 is running a funnel with a very narrow throat, and it processes applicants on a schedule rather than on how much any individual wants the place.
Your own search is the same shape. What is different is that most applicants never build the funnel at all. They send ten applications to ten companies, wait, and then describe the result as luck.
Ten is not a search, it is a hope
Run the arithmetic. If one application in ten turns into a conversation, ten applications produce one conversation. If that conversation has to become an interview, an offer and a start date, you have bought yourself one shot at a chain of three events.
Now put a bad week inside it. One late application, one company that freezes hiring, one role that closes before your note lands. A ten-application search has no room for a single interruption, because there is nothing behind it.
Thirty applications produce three conversations, not because the ratio improved but because the base got bigger. That is the whole mechanism. Volume is not a substitute for quality. It is the only thing that makes quality survivable.
What converts at the top of the funnel
The top of the funnel is not your CV. When a reader opens an application the first thing they look for is evidence that you have already done something on the list, and the second is whether the thing you did can be checked.
That is a different instruction than "write well." It means every application should carry one thing that exists outside the document: a repository, a shipped page, a data set, a short write-up with a date on it. The proof does not have to be impressive. It has to be inspectable, and it has to be yours.
The four-week version
A workable search has a cadence rather than a burst. Fifteen applications a week for two weeks beats forty in one Sunday, because a replacement-rate search keeps the funnel full while the early ones are still being read.
Which week matters too. A cohort that runs for six months and takes 25 people has one entry point per cycle, and summer intakes are being finalised while students are writing exams. The applications that land in the quiet weeks are read by a person with time.
Why a referral behaves differently
A referral does not raise your score. It removes the first filter, which is whether a stranger opens the file at all.
This is where the network number matters more than the acceptance rate. Around 210 people have been through the programme across 19 countries. Spread that thin and you get something no job board can offer: in most major markets, somebody who has already done the thing you are trying to do, in a system with different rules, and who answers in under an hour.
That is not an introduction service. It is a group of 210 people who can each tell you what the regulator actually said.
The number to track
Track conversations started, not applications sent. Applications sent is an activity measure; it feels like work and it rewards you for volume that may never be read.
Conversations started is the only number that predicts anything, and it is the one you can still improve in week four when the first rejections arrive and the temptation is to rewrite the CV for the fourth time.
The probability of a search ending badly is a function of how thin it was. Widen it early. Then keep widening it.