Deciding what to do first

There is never a shortage of postings. There is always a shortage of the hours someone has to write good applications, especially while working. Ranking by match quality alone ignores the constraint that actually binds.

The question a prioritisation system answers is: given four hours this week, what should be done with them. That is a scheduling problem, not a scoring one.

The budget is also smaller than any tool assumes. Someone job hunting alongside a full-time role has perhaps three or four usable hours a week, and a plan that implicitly requires ten is a plan that gets abandoned in week two.

A strong match requiring a portfolio submission and a written exercise may be a worse use of two hours than three good matches with short forms. Value per hour is the ranking that matters.

Effort is estimable from the application route: number of steps, custom questions, whether tailored documents are needed, whether an account must be created. None of that requires guessing.

Improve the estimate from the candidate’s own history. How long their last four applications through this platform actually took is a better predictor than any general model, and recording it costs a timestamp at each end.

• Expected value — match quality, employer responsiveness, how much they want it
• Estimated effort — steps, custom questions, documents required
• Urgency — stated closing dates and how long the posting has been live
• Rank by value per hour, not by match score

A stated closing date is a hard constraint and should dominate ranking as it approaches. A posting that has simply been open for three weeks is a softer signal — the best applications arrive early, and the odds decline.

Model both. Without them a system happily recommends a role that closed yesterday over one closing tonight, and the user loses an opportunity to an ordering mistake.

Let a deadline pre-empt the weekly plan rather than waiting for the next cycle. A role closing tomorrow that clears the quality floor should surface today, and a system too tidy to interrupt itself is useless at the moment it matters most.

Ranking purely by immediate expected value concentrates everything on the safest, most similar roles. That maximises the chance of some offer and minimises the chance of a better one.

Reserve part of the plan for higher-variance applications — a stretch role, a different direction. It is the same exploration argument as in recommendation, and it matters more here because the stakes for the individual are higher.

Include what is not an application at all. An hour on a conversation, a referral request or a piece of public work has a different and often better return than the marginal fourth application, and no system optimising a queue of postings will ever suggest it.

Order matters beyond which roles make the list. The application someone cares most about should not be written last, when they are tired and the quality floor has quietly dropped — that is how the best opportunity gets the worst effort.

Put the highest-stakes application first in the week and batch the mechanical ones together. Switching between a considered cover letter and four portal forms costs more attention than either task does alone.

Leave slack deliberately. A plan that consumes every available hour has no room for the interview that appears on Wednesday, and a plan that cannot absorb good news is a plan that gets abandoned the first time something goes right.

A ranked list of forty is still forty decisions. What changes behaviour is a small, concrete plan: these three this week, in this order, with roughly this much time each.

Explain each choice in one line and let the plan be reordered. A prioritisation people can adjust is one they follow; an opaque ranking is one they override and then ignore.

Treat a reordering as information rather than as a correction to absorb silently. A candidate consistently promoting roles the system ranked low is telling you something about their preferences that no dismissal signal captures.