Automated Shortlisting ROI: Results from a 3-Role UK Pilot

Hiring teams do not need another dashboard. They need a faster path to a clean shortlist, evidence that interview quality is rising, and a process that stands up with legal. We ran a three-role pilot to measure the return on automated candidate shortlisting with numbers a finance lead can accept and a hiring manager can trust.

Pilot baseline and goals

Company: a UK-based B2B software firm with 220 employees. Roles: Customer Support Associate, Sales Development Representative (SDR), and Full-Stack Developer. Each opening drew between 120 and 210 CVs from job boards, referrals, and the career page.

Baseline, recruiters worked from a shared inbox and an ATS queue. First-pass screening took 3 to 4 minutes per CV for a quick read and tagging. Hiring managers then skimmed the top 30 to 40 profiles and moved 12 to 18 forward. Across the three searches we logged 34.5 hours of human time to produce shortlists and five business days from job post to final slate. Interviewer feedback showed a 48% pass rate from first interview to next stage, with frequent notes about unclear fit.

Goals were specific. Cut first-pass time by half. Improve interview pass-through by at least 15% by sending better aligned candidates. Keep a record that supports GDPR-aware screening with traceable decisions, because this is often the first question for cv screening software uk buyers.

Setup: criteria and a reusable rubric

We used Marxel for ai cv screening and resume screening software tasks. Setup followed four steps per role with recruiters and hiring managers working in the same session.

1. Criteria from the brief. We pasted each job description and briefing notes into the tool. It produced draft criteria with weights. Examples:
   • Support: customer contact volume, ticketing systems, SLA handling, shift readiness.
   • Developer: languages (TypeScript, Python), shipped features, production ownership, version control, on-call.
   • SDR: outbound volume, quota consistency, CRM hygiene, objections handled.
2. Editable weighted criteria. We tuned weights to focus on outcomes over labels. Illustrative adjustments:
   • Developer: reduced a specific framework from 20% to 5%; increased shipped features from 15% to 25%; on-call readiness from 5% to 10%.
   • SDR: added coachability at 10% with evidence from reviews when present; raised quota consistency from 15% to 25%.
   • Support: ticket complexity at 20% outranked years in role at 5%.
3. Bias-aware checks. The tool flagged risky or vague phrases in the briefs. We removed “native speaker” and rewrote “rockstar” to “meets quota for 3 consecutive quarters” before any CVs were scored. This made “good” concrete.
4. Reusable scoring rubric. Each approved rubric was saved and applied to all candidates for that role. Two weeks later, when we reopened Support, the same yardstick preserved consistency.

We then ran bulk CV screening. For each role we uploaded up to 200 CVs in one run. Processing progress tracking showed runtime and percent complete so recruiters could plan their day instead of hovering. When runs finished, candidates landed in four buckets: Aligned, Potential, Hold, or Unclear.

Explainable reasoning showed which criteria were met, where evidence was thin, and a confidence indicator. A typical SDR note read: “Outbound volume 80–100 calls/day (met), quota attainment 3/4 quarters (partial), CRM notes detailed with next steps (met), coaching feedback present (met). Concern: enterprise prospecting experience missing.” Where the tool missed context, manual rebucketing allowed a move with a note. Candidate-pool queries helped compare borderline profiles, for example “show all Support candidates with high contact volume but limited system coverage.”

We exported shortlists to CSV for reporting and manager handoff. The file included bucket, score, top evidence snippets per criterion, and reviewer notes. Managers could scan columnar evidence instead of opening every CV.

Data handling. We worked under gdpr compliant cv screening expectations. Marxel uses encryption and access controls, supports GDPR-conscious workflows, and does not use uploaded CVs to train Marxel-owned models. Our legal team reviewed the pilot terms and the audit record before go-live with hiring managers.

Input quality also mattered. On the candidate side, tools like ApplyTOP help applicants align language to the job, which increases relevant signals. Clear, weighted criteria kept our screening anchored to evidence rather than only to phrasing.

Results: time, quality, ROI, and audit trail

Throughput and time saved

• CV volume. Support: 186 CVs. SDR: 209 CVs. Developer: 124 CVs.
• Processing time. Average runtime per batch was 7 to 10 minutes. Review and calibration added 90 to 120 minutes per role, including one manager sync.
• Shortlist speed. Time to a manager-ready shortlist dropped from five business days to two. Human hours fell from 34.5 to 11.0 across the three roles, a 68% reduction in screening effort.

Shortlist quality

• Interview pass-through. Pass rate from first interview to next stage rose from 48% baseline to 59% with the AI shortlist, a 22% relative improvement. This metric was tracked per role and rolled up weekly.
• Bucket fidelity. 81% of interviews came from the Aligned bucket. Only 9% of candidates were manually rebucketed after review, mostly moving Potential to Aligned when human reviewers validated context the model had marked as partial.
• Manager confidence. “Why are we seeing this person?” declined from a frequent comment to a rarity. Time spent debating the slate in calibration meetings dropped from 45 minutes to 18 minutes per role because evidence sat next to each score.

Cost and ROI snapshot

• Time cost reduction. Using a blended loaded cost of £45 per recruiter hour, cutting 23.5 hours equates to £1,058 saved in the first month across three roles.
• Opportunity cost. Faster shortlists pulled interviews forward. Two roles closed a week sooner than the prior quarter’s average, protecting Support coverage and outbound prospecting pipeline. We did not assign a pound value to this in the pilot, but hiring managers flagged it as material.
• Reporting. The CSV export fed a simple sheet that tracked per-role time, pass-through, and notes. Finance requested this view for monthly recruiting ops review.

Audit trail and explainability

• Governance. Compliance reviewed criteria and saw risky language removed before screening began. Criteria approvals and later edits were timestamped with editor names.
• Consistency checks. When two SDRs with similar tenure landed in different buckets, the reasoning showed one had documented quota attainment and detailed CRM notes that met the weighted threshold. The other had activity volume but missing evidence. The call was visible and defensible.
• Calibration insight. In the Developer search, three Aligned candidates shared a noted concern around production on-call. We updated the brief and increased the on-call weight for the next intake. Pass-through for that criterion improved the following week.
• Planning. Processing progress tracking let recruiters kick off morning batches, check runtime, and schedule reviews for lunch, instead of refreshing tabs.

Lessons and how to scale next

• Spend the time on weights. The biggest quality gains came from shifting weight to outcomes. For SDR, quota consistency outranked brand names. For Support, ticket complexity outranked years in role.
• Use bias-aware checks early. Removing vague and risky phrases at setup kept the rubric clean and shortened approval cycles. It also supported our GDPR-conscious approach.
• Read the reasoning. Scores are pointers, not verdicts. Reviewers made the best calls by scanning the evidence excerpts next to each criterion, then using manual rebucketing with a note when needed.
• Close the loop with interviews. We tracked interviewer pass-through by bucket and fed that back into weights. That is how the Developer role surfaced the on-call signal we missed in the first pass.
• Standardise reporting. The CSV shortlist created a shared view of candidates, reasons, and next steps. We now attach it to every kickoff and wrap-up note.

Next, we will extend to five roles and reuse saved rubrics where they fit. We will keep the pilot discipline for every new search: criteria from the brief, edits and approvals, one bulk run, a quick calibration loop, then export and report. If you are comparing resume screening software for a UK team, or searching for cv screening software uk with GDPR in mind, start with one role and measure the same three numbers we did: time to shortlist, interview pass-through, and the quality of your paper trail.

Key takeaways

• Automated shortlisting cut first-pass screening hours by 68% across three roles.
• Interview pass-through improved by 22% with a weighted, approved rubric.
• The audit trail, explainable reasoning, and CSV export made decisions clear and reportable.
• GDPR-conscious handling supported a defensible process without slowing hiring.