What screening applicants really costs

Screening is the one part of hiring that never arrives as an invoice. Job boards bill you. Agencies bill you. Software bills you. The forty applications read on Sunday evening, the six calls booked and the two not attended: none of it appears on a budget line, so it is rarely counted and never compared.

That is why teams tolerate it far longer than a tool costing the same. A founder who would refuse a few hundred dollars a month will spend two evenings a week reading applications without flinching, and the second one is more expensive.

For most small teams, screening one role breaks into the same stages.

• Reading applications. Deciding in a minute or two whether each is worth a call. The part people picture, and on a busy role the smallest share of it.
• Sifting, then sifting again. Resumes are not comparable to one another, so the pile gets sorted more than once: a pass for obvious mismatches, one to rank what is left, another when the shortlist comes out too long or short.
• Scheduling. Proposing times, waiting for a reply, working around the candidate's current job, sending the link, sending the reminder. Dead time for your team and elapsed days for the role, per candidate, per round.
• Calls that do not happen. Some candidates do not attend, some cancel the night before, some accepted elsewhere last week. The slot has gone anyway.
• The same screen, twenty times. The identical opening questions, asked by somebody whose attention at call nineteen is not what it was at two.
• Writing it up and comparing. Notes in different shapes from different interviewers, then a meeting to reconcile them. Where nobody wrote the answers down, it happens from memory.

Two of those stages, scheduling and the calls that never happen, tell you nothing about any candidate. They are pure overhead that scales with the number of people you speak to, which is the number you want to increase.

You do not need an industry benchmark to price your own screening. You need one role you filled recently and four numbers you already have.

1. How many applications the role received.
2. How many you actually read. Most teams stop well before the bottom of the pile, and that gap is worth noticing on its own.
3. How many first-round calls you held, and how long each took including the scheduling messages and the write-up.
4. The loaded hourly cost of whoever did it: salary plus employer costs over working hours. If that is you, use what an hour is worth against the next best use of it.

Then account for it stage by stage rather than as one lump.

Swipe the table sideways to see every column.

Multiply hours by the loaded rate and you have a figure you can hold next to the price of any tool, this one included. Do it per role, not per hire: a role that ran three rounds and ended in a declined offer still cost every hour.

Time is the cost you can compute. These arrive later.

The bottom of the pile is never read

Attention runs out before applications do. Whoever reaches application 140 at eleven at night is not applying the standard they applied to application four. Every unread application is a candidate you paid a job board to reach and then did not evaluate.

Inconsistency, and decisions made on thin evidence

When the first round has no structure, every candidate gets a slightly different interview and the comparison at the end is between notes that do not line up. If it produced an impression rather than a record, the second round starts from nothing and the final call goes to whoever was most confident in the room. Both are quality problems right up until somebody asks why the rejected candidates look different from the hired ones.

Search for the cost of hiring and you will find averages: a cost per hire in dollars, recruiter hours per role, a share of resumes supposedly rejected before any person sees them. Following the trail usually ends at another article quoting the same figure. The most persistent of them is the claim that roughly three quarters of resumes are rejected by applicant tracking software before a human reads them. It circulates without a study behind it, and is worth naming as folklore rather than repeating. Filters do reject applications, because employers configure them to, but that famous fraction is not a finding.

Some things about screening are established well enough to build a process on, and they concern how you interview, not how fast you read.

Structure predicts better than conversation. A re-analysis of decades of personnel selection research by Sackett, Zhang, Berry and Lievens in the Journal of Applied Psychology revised long-standing validity estimates downwards across the board, by their account between .10 and .20, and still found that structured interviews came out top-ranked among the procedures it re-examined.

Reading resumes is not a neutral act. In the field experiment behind Bertrand and Mullainathan's callback study, otherwise identical resumes sent to real job adverts in Boston and Chicago drew 50 percent more callbacks when the name at the top sounded white, and a stronger resume raised callbacks by about 30 percent for those names while doing considerably less for the others.

Overriding a structured signal on instinct tends to cost you. In Hoffman, Kahn and Li's study of job testing across fifteen firms, managers who hired against the test recommendation ended up with worse hires on average, which points at bias or error rather than superior private knowledge.

None of that says software should decide. It says the first round is worth making structured, recorded and reviewable by a person who can see what it rested on.

The alternative to screening faster is letting every applicant have the first conversation, then reading the results.

On Career1, applying to a job starts the interview. It is spoken, runs in the browser on a laptop or a phone with nothing to install, takes about eight minutes and is capped at ten. Nothing gets booked, so it happens at whatever hour suits the candidate. Questions come from the role and the applicant's own resume, and the interviewer follows up when an answer is thin, which is what separates work somebody did from work that happened near them. Because the ranking comes from the answers rather than the document, it is resume screening that verifies the resume rather than parsing it.

Each applicant's report carries a match score against the role, an overall interview score, a recommendation from strong hire to no hire, scored sections with written reasoning for technical depth, communication, problem solving, experience fit and culture fit, a ledger of which claimed skills were actually demonstrated and the quote that shows it, strengths, risks, and questions for the human round. The transcript and recording sit beside it, and there is a complete example to look at.

That removes the scheduling, the slots lost to no-shows and the twentieth repetition of the same question. What it does not remove is the reading: somebody still reviews the shortlist, deliberately. Career1 ranks and recommends, and rejects nobody.

Pricing is the easy half of the comparison, because it is published. Career1 charges by the number of candidates vetted each month: the free plan covers two and needs no card, and paid tiers run from $49 to $199 a month for 25, 60 and 150 interviews. Reports, transcripts and recordings are included at every level, and when an allowance runs out live jobs pause until it renews, so nothing arrives as an overage. Current plans are on the pricing page.

Put that next to your own figure, per role. The question is not which number is smaller but which hours come back, and whose. Candidates never pay.

• Your bottleneck is finding people, not screening them. Career1 does not source or find candidates. If three people applied, screening is not the cost.
• You need an applicant tracking system. Career1 interviews and ranks applicants and keeps simple stages for them; it is not a record of every role and interviewer across a company. If coordination is what is breaking, read ATS or AI interviewer.
• Your first screen is a work sample. A coding exercise or a portfolio review measures something a conversation does not.
• Screening is pure eligibility. If the only real questions are a licence, a shift pattern or the right to work, ask them on the application form.
• You interview in a language other than English. The interview is English only at the time of writing.
• You hire twice a year from a handful of applicants. Call them. The wider list of what to check before buying any of this is in how to evaluate an AI interview platform.

1. Create a workspace. An email address and a password, no card.
2. Post one role you are genuinely trying to fill, or paste the description you already wrote.
3. Copy the job link into wherever you advertise, or send it to the applicants already sitting in your inbox.
4. Applicants interview themselves, whenever they get to it. Nobody books or chases anything.
5. Open the shortlist and read the reports at the top, transcript beside each score.

The free plan vets two candidates a month with no card, enough to see a real report on a real applicant before committing to anything.

Each source below was opened and checked on 26 September 2026.

• Sackett, Zhang, Berry and Lievens, "Revisiting meta-analytic estimates of validity in personnel selection", Journal of Applied Psychology: doi.org/10.1037/apl0000994.
• Bertrand and Mullainathan, "Are Emily and Greg More Employable than Lakisha and Jamal?", NBER working paper 9873: nber.org/papers/w9873.
• Hoffman, Kahn and Li, "Discretion in Hiring", NBER working paper 21709: nber.org/papers/w21709.

Deliberately absent, for want of a source we could open: average cost per hire, recruiter hours per role, and no-show rates.