Rui sent 31 applications from a hospital procurement office in Braga and got 2 replies, one of which became an interview. Over the same three months he rewrote the CV four times and never once looked at the list of jobs he was sending it to. Those two numbers say which of the two was the mistake, and they say it before anybody reads a word he wrote.
Read the count from the top rather than the bottom. 31 applications producing 2 replies is a collapse at the first step, and the first step is targeting and wording: either the jobs are wrong for the evidence he has, or the evidence is there and the page does not say it in the words the advert used. A collapse further down would mean something else, and one interview is not a sample of anything. Rui had concluded he interviews badly.
The two bullets that get read, and what has to be in them
The scan is about six seconds and it lands in a fixed order: job title, employer, dates, then the top of the most recent role. The first two bullets under the most recent job are the ones read, and everything below them is insurance. Rui's first two were Responsible for supplier management and purchasing across the department and Liaised with clinical staff and external vendors. Both are duties, and a duty is a list of what the job contained rather than what Rui did inside it. Two candidates with the same duties are indistinguishable on paper, which is the real failure: three people in that department did supplier management, and their first two bullets would have read the same.
A bullet has a verb, a number and an outcome, in that order. Consolidated the ward's consumables from eleven suppliers to three and took order to delivery from 9 days to 4, across 2025. The verb says he did it rather than was present while it happened. The numbers make it checkable, which is the only reason anybody believes a CV. The outcome is the thing a procurement manager is measured on, so it is legible to the person deciding. Rui had done that work. It was on page two, seventh bullet, described as involvement in a supplier rationalisation project.
The other half of the fix is not on the CV at all. The words in the advert are the words to use back: if the advert says stock control, the bullet says stock control rather than inventory management, because the first reader is a parser matching strings and the second is a person who chose those words on purpose. That is not stuffing a page with keywords, it is answering the question that was actually asked.
Evidence beats the paragraph you wrote about yourself
Deciding what to aim at is the one part of this topic with no form to fill in, and it is the part that decides whether the other four are wasted. The method is not introspection, which returns whatever mood you are in. It is a search of the record: list the last ten things people came to you for by name. Not what you believe you are good at, but what other people already believed, at the cost of walking over and asking.
Rui's ten came out as six about suppliers and contract terms, one about a spreadsheet nobody else could fix, and three about rotas. None of them were about the clinical side he had been describing as his strength, because that was the part he found interesting rather than the part he was reliably better at than the person asking. Those are different lists and only one of them is evidence.
The stall is the fifth rewrite of the only thing you control
Here is where people stop, and it looks like diligence. The CV can be improved alone, on a Sunday, and every pass makes it a little better, so the hours go there. The list of jobs cannot be improved alone: it needs somebody to say what a role is really like, which means asking, and asking is the half where a person can decline. So the artefact gets its fifth version and the list stays whatever the search results returned in week one.
The check is arithmetic rather than mood. Count the applications since the last change to the CV and the replies to them. If the reply rate did not move, the fifth version has told you the CV was not the constraint, and a sixth will tell you the same thing more slowly.
The number, and the sentence that carries it
The offer conversation is the one people prepare for least and it has one move in it: give a range with a reason attached, not a single number. A range with nothing behind it is a hedge and gets heard as the lower end. A range whose ends are attached to something, the framework renewals sitting inside the job or outside it, is a question about scope, and scope is the ground where a candidate has something to say that the other person does not already know.
What the range should contain is not on this page and will not be anywhere on this site. That is a number about one market in one month, and it comes from people doing the job now, from the advert when it publishes one, and from a recruiter who has no reason to mislead a candidate they might place next year.
The climb
Five camps, and the last stretch is the long one
Every camp before the last is reached by answering something somebody advertised. The last one is not advertised, and nothing about the first four prepares anybody for that.
- 01Knowing what you are aiming atNo form to fill in, which is why it is skipped
- 02Materials that survive six secondsA parser first, a person second
- 03The conversationThe only stage where both sides are being assessed
- 04The numberOne conversation, prepared for less than any of the others
- 05Being handed the next job before the titleNobody announces that this climb has started, or that it has finished
The spacing is not time. The camps are placed by how much has to change in the person, which is why the ground steepens at the end and why people arrive at the last stretch expecting the gradient of the first.
The question that tells you most in an interview
Ask what happened to the last person in the role. The answers divide cleanly and none of them is bad news by itself: promoted internally, left after eight months, still there and the team grew around them, or the post is new. What the question really buys is the pause before the answer, and how much of it is about the work rather than about the person. Rui asked it once, in his one interview, and got a fluent paragraph about restructuring that answered a different question, which is itself an answer.
What to open in 2026
LinkedIn is where the searching actually happens, and it replaced the recruiter's card index and the speculative letter to a company that was not hiring. The free account is enough to be found and to apply; the paid tier mostly buys messages to people who did not ask for them. The part worth the hour is the headline and the top of the profile, because those are what a recruiter's search returns and reads, in the same fixed order as the CV scan.
Export the CV as a plain PDF out of Word or Google Docs, both free in a browser, and it replaced the designed template with two columns and a sidebar. A parser reads a two column layout in the order the boxes appear in the file rather than the order the eye sees them, which turns a good CV into interleaved nonsense before a person ever opens it. The visual cost of a single column is small and the failure it avoids is total and silent.
A model drafts the covering letter now, which is the change since 2023, and it replaced the blank page and the hour of dread attached to it. Every major assistant covers it on a free tier. What it cannot supply is the evidence, and a letter with no specific evidence in it reads exactly as what it is, which is worse than three honest sentences. Give it the ten things people came to you for and the advert, and it produces something worth editing; give it nothing and it produces something that has been received four hundred times that week.
Five clusters, and the last one has no advert to answer
Deciding what to aim at is first and gets skipped first, for the reason above. The materials, and getting them read is where the six seconds and the parser live. The conversation that decides it is the only stage where the other side is also being assessed, which is what the last-person question is for. The part where money is decided is one conversation long and it is worth preparing more than the four weeks that led to it.
Moving up, or sideways, on purpose is the cluster with the longest gap between doing the work and being paid for it, which is why it is read after somebody has been passed over rather than before. Somebody selling the same skill without an employer in the middle is reading selling your own skill, directly instead, where the pricing question replaces the offer conversation entirely. Once the next job is other people's output, it is getting work done through other people, and for anyone doing all of this in a second language the English that decides whether you get the role teaches the interview vocabulary rather than assuming it. All of it sits under running work, getting paid, and the software it happens in.
What none of the five clusters settles is whether Rui should be aiming at procurement at all. Every page below assumes the target is chosen and improves the aim, and the evidence method above tells him what he is reliably good at without telling him whether he wants ten more years of it. Those are different questions, the second one has no procedure, and a page claiming otherwise would be the one dishonest page here.