Screening candidates without wasting everyone's week

Hiring is expensive in the currency small agencies have least of: senior attention. The people best qualified to judge candidates are the same people billing at $150 an hour, and every hour they spend on a shortlist is an hour that isn't on a client project.

So the goal isn't a thorough process. It's a process where the expensive hours land on the decisions that are actually close.

What the process costs

A hypothetical funnel from 120 applications to one hire, with all review time done by someone senior at a $95 loaded hourly cost:

Stage	Candidates	Time each	Total
Scan applications	120	2 min	4.0 hrs
Review written answers	30	10 min	5.0 hrs
Screening call + notes	12	30 min	6.0 hrs
Review paid trial task	4	90 min	6.0 hrs
Final conversation + debrief	2	90 min	3.0 hrs
Total			24.0 hrs

At $95 an hour that's $2,280 of internal cost, plus whatever you pay for the trial tasks — four candidates at four hours each at $80 an hour is another $1,280. Call it $3,560 before anyone signs anything, and that's a lean version.

The shape matters more than the total. Notice that the 120-candidate stage costs four hours and the four-candidate stage costs six. That's correct. Cheap decisions at the top, expensive attention at the bottom. Most agencies get this backwards — they agonise over the longlist and then make the final call on a gut feeling after one conversation.

Stage 1: the two-minute scan

You are not evaluating anyone here. You're sorting into "no" and "maybe" as fast as honestly possible.

Decide your two or three hard criteria before you open the first application, write them down, and apply only those. If the criterion is "has run projects directly with clients", then a beautiful portfolio from someone who has only ever executed someone else's brief is a no, and you move on in forty seconds.

Two minutes each is enough for one portfolio piece or code sample and a glance at the trajectory. It is not enough to be fair to a borderline application, which is why borderline goes in "maybe" and gets its two minutes back later.

Do this in one sitting. Standards drift badly across days.

Stage 2: the written answers do the heavy lifting

If your job post asked two judgement questions — the change-of-brief question, the work-a-client-didn't-love question — this is where they earn their keep.

Read for three things:

• Specificity. Real projects have awkward details. Generic answers describe a process rather than an incident.
• Ownership. Does the difficulty in the story belong to them, or is everything the client's fault or the previous agency's fault?
• What they optimised for. Someone who solved a scope change by working the weekend and someone who solved it by renegotiating the deadline are different hires. Neither is wrong; you should know which you're getting.

Ten minutes each, notes in one shared doc, one line of reasoning per candidate. The line of reasoning is the point — it's what stops the shortlist from becoming a popularity contest three weeks later when you've forgotten why someone was on it.

Stage 3: the 25-minute call

Twenty-five minutes, on the calendar as twenty-five minutes. Not an interview — a mutual filter.

Three questions, same for everyone:

1. "Walk me through what your last two weeks of work actually looked like." Hour-level texture. This is very hard to fake and tells you more about seniority than any credential.
2. "What kind of work do you want more of, and less of?" If more-of is not what this job is, stop here — kindly, and immediately. You've both saved a fortnight.
3. "What questions do you have about how we work?" The quality of the questions is the signal. Someone asking how review cycles work and who signs off has done this job. Someone asking nothing has not thought about it.

Then five minutes of you being honest about the hard parts. Not a pitch. The candidates you lose to honesty were going to leave anyway.

Take notes during, not after. The five minutes you spend writing up straight afterwards is the difference between a decision and a vibe.

Stage 4: the paid trial task

One task, four hours maximum, paid at a fair rate, with a real deadline and a real brief — ideally an anonymised version of something you've actually done.

Rules that keep this useful:

• Pay for it. Unpaid multi-hour tasks filter for people who can afford to work for free, which is not the filter you want.
• Cap it explicitly and mean it. "Four hours. If you're at four hours and it's not finished, stop and tell me what you'd do next." The people who ignore the cap and deliver twenty hours of polish are telling you something about how they'll handle project budgets.
• Ask for the reasoning, not just the artefact. Half a page on what they decided and what they'd do differently with more time. The reasoning is more informative than the output and much harder to outsource.
• Have two people review it independently before discussing. Otherwise the first opinion voiced becomes the shared opinion.

Give the same brief to every finalist so you're comparing like with like.

Stage 5: the conversation that decides it

Ninety minutes, and spend most of it on the trial task. Have them walk you through what they did, why, and what they'd change. Push on one decision and see whether they defend it, revise it, or fold. All three are informative; folding instantly under mild pressure from a future boss is the one to worry about.

Then the practical part: notice period, working pattern, salary, anything that could derail an offer. Find out now.

Debrief within the hour, in writing, before anyone talks themselves into consensus.

Where to cut if you're short on time

Cut stage count, not depth. A four-stage process where the trial task is real is better than a six-stage process of conversations. In particular:

• Don't add a "culture chat" with someone who won't work with the person.
• Don't run panel interviews at agency scale. Four people asking overlapping questions costs four times the money and produces one opinion.
• Don't do a second trial task. If you can't decide after one, you're deciding something the task can't answer.

This week

Write down your two or three hard criteria for the role you're hiring now, before you look at another application. Then time yourself on ten applications at two minutes each. If you can't decide in two minutes, the criteria aren't concrete enough yet — and that's the cheapest problem to fix, because it's costing you at the widest part of the funnel.