Stagger your applications

Every piece of job-search advice on the internet is about volume. Apply more. Apply faster. Apply everywhere. The single most useful piece of advice in the research behind this product is the opposite, and it comes from one person describing what actually worked.

What the account says

The thread is a write-up from a senior full-stack engineer about applying to tier-1 tech companies exclusively through referrals, with a Sankey diagram of where the applications went. It has 463 upvotes and 68 replies. Most of it is about referrals, and the numbers are the kind people write up when they want to be believed: a thousand applications beaten by twenty-eight referrals, four processes reaching onsite, one offer.

That is not the part worth copying, and this post is not about referrals.

It is this, in the interview-prep section:

I staggered a bunch of lower-priority applications first to practice, and by the time my last process rolled around (which was for the role I eventually took) I felt a LOT more capable and my pass rate increased (my last 4 applications I went all the way to onsite).

Read that again, because the important part is the order. The applications were not the practice. The practice was the applications. He was not looking for a job and treating every application as a serious attempt. He was spending the ones he could afford to lose on being visibly, honestly bad at something, so that the one he cared about would arrive last.

That is the inverse of everything the volume advice says to do, and it is the reason this product has a practice flag on a posting, a marker that means this is one I do not mind losing, and I am rehearsing on it.

Why the order is the mechanism

Volume advice treats an application as a ticket with one outcome: offer, or no offer. Under that model, an application to a job you do not care about is pure waste, and the correct number of them is zero.

That model is wrong about what an interview is. An interview is the only place most candidates ever speak at length about their own work, under time pressure, to a stranger who will compare the answer to a scale. Every other rehearsal most people do, whether that is reading a prep guide or rehearsing a story alone in a mirror, does not reproduce the thing that actually fails. The voice, the pace, the moment you forget the number you memorised and have to approximate it, the follow-up you did not see coming.

Practice on a job you want and you have spent a real chance at a rehearsal. Practice on a job you can afford to lose and the same hour is spent, and the downside is bounded by definition.

The people the product is for are the ones who cannot afford to lose anything. That is not an edge case; it is most of them, which is why the flag is a boolean the candidate sets themselves rather than something the product can work out. Nothing in the data can tell you which opportunity you can lose. You know which one it is.

What this is, and what it is not

This is one person's account of their own search. It is n = 1, from a forum post, and the run that found it recorded it as the strongest thread in the research precisely because the mechanism it describes is visible in the outcome, which is the last four applications all the way to onsite after a stagger, and not because a thousand other people have said the same thing. They have not. Treat it as a hypothesis with a person attached, not as a finding.

What makes it worth acting on anyway is that it is cheap and bounded. The downside of practising on a posting you have already decided not to care about is that you did not get it, and nothing else. There is no version of this that costs you the thing you wanted.

Two things this product deliberately will not do with it:

• It will not tell you that practising worked. Once you have marked a few applications as practice and a few as ones you cared about, the product counts how many of each were answered and shows you the two counts side by side. It does not compare them into a verdict. Three applications on each side is arithmetic, not evidence, so below that floor the comparison is absent rather than shown.
• It will not tell you how many practice applications to send. That number is not knowable, and a product that produced one would be inventing the thing this whole exercise is about.

The part that is hard

The advice is easy to agree with and awkward to execute, because the applications you do not care about are the ones you have already mentally filed. Sending one is a small act of deliberate waste, and it has to be a real application, with a real CV, a real posting and a real interview, because the failure mode you are rehearsing for does not occur in a practice round that is only a practice round.

The account also does the other half, which is worth more than the staggering: DIY mock system design interviews with an LLM, and recording yourself answering a behavioural question and reading the transcript back. Neither of those needs an application at all.

If you are mid-search and the job you want is next week, staggering will not help you. If the job you want is two months out, the applications you are going to send in the meantime are the practice, and the only question is whether you are going to treat them as practice or as attempts.

Related

• Read the interview questions off the job posting, most of what can be prepared is already in the posting
• STAR is the wrong half of STAR, on the one rehearsal technique here that needs no application at all