Hiring

A ten-person company runs fine with no hiring process at all. The founder messages a former colleague, they get coffee once, and the person starts the following week. That works because there is exactly one person judging and that person already knows the candidate.

Past forty people the same approach breaks. The pool of people you already know runs out, the number of people doing the judging grows, and three or four roles are open at the same time. Pagedeck went through this. On one backend engineering role Theo said yes and Grace said no, and the two of them had looked at completely different things — Theo spent thirty minutes talking about code, Grace explained where the company was heading and watched the reaction. They met the same candidate and had nothing in common to compare, so the decision went to whoever argued harder. The question that came out of that meeting was the starting point: had anyone actually written down what kind of person this role needed?

Leave it unwritten and three things collapse together. Every interviewer asks about something different, nobody can explain why a candidate was rejected, and the next time the same role opens there is no record of the bar that was used. It is no better from the candidate's side. Two weeks pass with no word, they take another offer, and the company never finds out why they disappeared.

Hiring got bundled into one named practice in order to turn those loose pieces into a flow with an order and an owner. Write down who is needed, look where that person is likely to be, filter against the criteria set in advance, put terms on the table, and get to an acceptance. That whole span is the unit. It is also why this branch sits at the front of the wider people function: onboarding, performance and pay all deal with someone who is already inside, so they cannot begin until it is settled who came in.

The terms under hiring fall into six groups: what has to exist before you start (preparation), who owns the work (role), the routes by which you meet candidates (sourcing paths), the ways you judge them (stages), the step where terms get agreed (closing), and the numbers that tell you whether the flow is working (metrics). The first five follow the order of time; the last one looks back across all of them.

Preparation is what gets done before the door opens. It splits in two, along the line of whether it serves one role or the whole company.

1. Job description — the document stating what the person will own and what they need to be able to do. It reads like an advertisement for attracting candidates, but it works more like an agreement that forces interviewers onto the same criteria. The three lines under required qualifications become the three things checked in the interview, so a vague document makes every later step vague with it.
2. Employer branding — the work of making the company look like somewhere people want to work. Writing up how the team works on a public blog, or publishing an engineering talk, belongs here. It pays off slowly, so it slides down the list easily, and without any of it built up, every sourcing message has to explain the company from scratch. A team with two years of posts behind it can open with a link; a team with none spends its first three paragraphs establishing that the company exists at all.

Role settles who owns the work, and it has one slot. The hiring manager is the manager who actually needs the seat filled and makes the final call. A recruiter keeps the process moving, but deciding what kind of person is needed and whether this one passes belongs to the manager. Where that line is blurry, applications sit untouched for three days and decisions slip a week past the last interview.

Sourcing paths cover the routes by which candidates arrive, and there are four. Whether you wait or go looking separates the first three.

1. Inbound application — people who see the posting and apply on their own. It takes the least effort and brings in exactly as much as the company is already known for.
2. Employee referral — someone on the team introducing a person they know. The candidate has already been judged by somebody who worked with them, so the pass rate tends to be high. The catch is that people with similar backgrounds cluster, and leaning only on this route tilts the shape of the team.
3. Sourcing — going out and contacting people rather than waiting for applications. It reaches people who are not currently looking, which is the closest route to strong candidates, but twenty messages producing two or three replies is normal, so it costs a lot of hours.
4. Talent pool — a kept list of candidates who are not right now but may be later. People who narrowly missed at the final stage, or who declined because of timing, stay here. When the same role reopens six months on, it saves starting the search from nothing.

Stages cover the ways candidates get judged. There are five, and they usually sit in this order.

1. Screening — deciding from the application alone whether to move someone forward. What is being read is not how impressive the person is but whether anything in the application overlaps the requirements written in the posting.
2. Technical interview — checking the skills of the job directly. Code for an engineer; for a designer, the reasoning behind decisions in past work.
3. Take-home assignment — a practical task done at home and submitted. It removes the pressure of the room and shows real output, at the cost of several days of the candidate's time. Without a stated limit on scope, the number of people who abandon the application goes up. A task described as four hours of work routinely eats a candidate's whole weekend, and the ones already employed drop out first.
4. Culture interview — checking whether the way someone works and decides fits the company. It gets misread as testing whether personalities match, when what it looks for is behavior: what the person does when opinions split, for instance. Asking how someone handled a review comment they disagreed with gets further than asking what kind of team they enjoy.
5. Reference check — asking people who worked with the candidate directly. It exists to narrow in on what the interviews could not settle, and it runs with the candidate's consent.

Closing is where terms get agreed, and there are two slots. An offer carries salary, level and start date, and it is the first point where the company becomes the one being judged. A counter offer runs the other direction: it is the raise the current employer puts up to keep someone who is leaving, which makes it the last wall the offering side hits. This is where a candidate everyone considered confirmed changes their mind three days later.

Metrics are the two numbers that show whether the flow is working. Time to hire counts the days from the posting going up to the hire being confirmed, and offer acceptance rate is the share of offers that get accepted. The two get read together because they light up different stretches. A long time to hire on its own usually points at the front — candidates are not arriving, or interviews are not getting scheduled. When the acceptance rate falls at the same time, the cause is not at the back end but at the very beginning, meaning the work described in the posting and the salary band attached to it are out of step with the market. If a team that used to fill six roles in 60 days suddenly needs 90, with acceptance dropping from 80 percent to 50, reopening the posting gets there faster than adding interviewers.

"This posting has been up six weeks and we have twelve applications. Let's look at the posting again before we book more interviews."

A proposal to look for the cause of thin applications in the document rather than in the interview stage.

"Two offers declined back to back. We need to check whether our salary bands still match the market."

Reading a falling acceptance rate as a problem with the terms rather than a change of heart, and asking to reset the pay ranges per level.

"All five interviews this week are blocked on the hiring manager's calendar."

Pointing out that the process is stuck on the decision-maker's time rather than on candidates, and warning that time to hire grows if it stays that way.

Pagedeck opened two engineering roles and one design role in the same quarter. Thirty minutes every Tuesday morning is the hiring meeting, and on screen each role gets a line showing how many people sit at which stage. Theo is the hiring manager for the two engineering roles and Noa took the design one.

In the first two weeks the engineering roles drew forty applications and the design role drew six. Noa reread the posting and cut the required qualifications from five lines to two. Out went five or more years as a product designer, expertise in one specific tool, startup experience, familiarity with collaboration tools, and a portfolio; week three brought nineteen applications. Grace pointed at a different line. The engineering roles had plenty of applications but were averaging nine days to a first interview, and two candidates had gone elsewhere in the meantime. Theo's calendar turned out to be the bottleneck, so first interviews moved from Theo alone to two engineers splitting them. One meeting produced two decisions, each touching a different stretch of the flow.

The most common mistake is treating hiring as the HR team's job. Postings, scheduling and application tracking all come out of that team, so it looks that way. But deciding what kind of person is needed, and making the final call, belongs to the manager who will work alongside them. Hand the whole thing over and the process runs while no conclusion arrives: decisions sit after the last interview, candidates get tired of waiting and go elsewhere, and six months later all the company has is the line that there was nobody good out there. Naming one owner per role is the cheapest fix for this.

The belief that more stages produce better hires also tends not to hold. Adding a round feels like it will catch someone the earlier rounds would have missed, but what an extra round catches is mostly what the previous one already saw. Ask a candidate similar questions five times and the information stops growing after the third. What does grow, reliably, is elapsed time. Five stages add about three weeks of scheduling alone, and inside those three weeks the strong candidates get answers somewhere else. People talking to several companies at once leave first, so adding rounds tends to leave behind the candidates with fewer options. Deciding what each stage is there to check, one thing per stage, and removing the ones that overlap makes the judgment sharper than stacking more of them.