Home of science. Built for hiring.

Alva's hiring system is grounded in decades of psychometric research. Our contribution: adaptive testing powered by IRT and Bayesian inference, calibrated on 1.5 million candidates, and independently certified by DNV.

Where the science comes from


The research that Alva's system builds on comes from decades of peer-reviewed work in personnel selection. Alva's own contribution: adaptive assessments powered by Bayesian IRT, a methodology that allows continuous recalibration without destabilizing the model.

 

Built and maintained by a full-time science team of psychometricians, ML engineers, and people scientists.

Dr-Paul-Sackett

Dr. Paul Sackett: long-standing collaborator

One of the world's most cited researchers on hiring and selection. His 2022 meta-analysis with colleagues aggregated thousands of validity studies to answer a deceptively simple question: which hiring methods actually predict job performance? The findings changed the consensus, and directly shaped how Alva's system is designed.

 

Dig deeply into a job, figure out what's important, put together questions linked to the crucial attributes of a job and treat every candidate consistently. And man, you've got a tool that's really quite effective.

 

Dr. Paul Sackett, Professor of Psychology at the University of Minnesota and past president of SIOP.

The methods that predict performance, ranked by evidence

Sackett's meta-analysis ranked hiring methods by how well they predict job performance, based on thousands of studies. The results are clear: structured methods outperform unstructured ones by a wide margin. And combining methods produces the strongest predictions of all.

  • Structured interviews (.42) rose to the top of the list, outperforming every other single method

  • GMA/logic tests (.31) and Big Five personality (.25) remain solid standalone predictors

  • Unstructured interviews (.19) deliver less than half the predictive power of structured ones

  • Years of experience (.07) and level of education (.10) are near the bottom

  • Combining methods (structured interviews + GMA + work sample) reaches up to .60, the strongest result in the study
Structured-interviews

“Bayesian IRT gives us something no one else in this market has: control. We can add items, remove bad ones, recalibrate the model, and push updates to every candidate and recruiter at once. Most competitors train their model once and leave it untouched for five to ten years. We retrain continuously.”

Morgan Pihl, former Head of R&D, now scientific advisor to Alva Labs

Morgan-Phil

The People Behind The Science

I came to Alva to do psychology. I ended up building an entire department. When a company trusts you with that kind of ownership, you don't just do your job. You shape it.

Anna Brodin

Lic. Psychologist & Expert Consultant

Alva Labs

Learn more

The best frontend work happens when you're close to the product decisions, not just executing specs. At Alva, I shape how people experience AI-driven hiring, and the stack lets me move fast without cutting corners.

Sumit Bajracharya

Sumit Bajracharya

Alva Labs

Learn more

I spent my career so far in TA at Spotify, Amazon and King. I understand what hiring teams need because I've led several and supported many more. That's why I joined Alva – the product solves problems I lived with for a decade.

Frida Ihreblom

Senior Customer Success Manager

Alva Labs

Full story

“A valid test score means nothing if it gets overruled by a gut feeling in the final meeting. The system exists to protect the evidence all the way to the decision.”

Kajsa Asplund , Psychologist, PhD (SSE), Lead People Science at Alva Labs since 2018

Kajsa-Asplund

How we predict job performance

Every assessment in our hiring system combines psychometric theory with machine learning. IRT models calibrate question difficulty and trait measurement. ML monitors for drift, bias, and degradation over time. The science doesn't just build the tests. It keeps them honest.

Logic Test

Measures general mental ability. Adaptive: tailors questions to each candidate's level. Takes about 15 minutes.

DNV-certified globally (.83 validity, EFPA: Excellent), the first logic test in the world to achieve that. Works across all languages without translation.

Personality Test

Measures the Big Five personality traits, the most reliable framework in personality psychologyCc. Same adaptive precision. Takes about 15 minutes.

DNV-certified globally (.88–.92 validity, EFPA: Excellent). Validated in 21 languages, each through a dedicated translation and recalibration pipeline.

Coding Test

Practical challenges that mirror real day-to-day engineering work. Candidates work in a GitHub-based environment with clear instructions.

Two layers of evaluation: automated functional testing and AI-driven code review for quality, structure, and best practices.

Structured Interviews

Role-specific templates with rating anchors that define what strong, sufficient, and weak answers look like. Consistent scoring regardless of who's in the room.

All scores on a 1-5 scale.

Built once. Recalibrated continuously

Every item in our assessments follows the same rigorous development process. Written by psychometricians, piloted with real candidates, and stress-tested before going live. Items that underperform get pulled.

Construct definition

Grounded in independent published
research.

Pilot testing

New items appear in real assessments
but don't affect candidate scores.

Fairness analysis.

Every item checked for group
differences (DIF analysis). Biased items
get flagged, revised, or removed.

Item development.

Written by psychometricians.
Reviewed for clarity and fairness.

Calibration

Items placed on the IRT scale using
real candidate data.

Monitoring and refresh

Parameters re-examined continuously.
Full item bank recalibrated every 18 months.

Fairness at every stage. Verified continuously.

Validity

Research foundation

Assessments built on GMA and the Five Factor Model, the two most validated constructs in personnel selection.

Calibrated using a working population sample

Benchmarked against 1M+ candidates and growing. No arbitrary norm groups. One calibrated scale.

Continuous monitoring

Every item re-examined against predefined thresholds. Underperforming or exposed items get archived.

Fairness

Differential item functioning

Every item checked for group differences before going live. Biased items get flagged, revised, or removed.

Transparent results

Candidates receive the same full report as recruiters. They own their data and get clear feedback.

Human oversight

Recruiters decide how to weight test results. Scores inform decisions, they don't make them.

Certified. Auditable. Compliant.

Every assessment is independently certified by DNV. Every AI decision is auditable. Every data point is protected by ISO 27001 and stored within the EU. We didn't wait for regulators to set the bar. We set it ourselves.

certificates_1

Certified by DNV

Independent third-party validation of scientific rigour and practical utility. Two reviewers, documented non-conformities, 3-year recertification cycle. Next audit: March 2028.

Only vendor with ISO 10667 globally

Full assessment methodology (Define, Assess, Decide) independently certified under the global standard for assessment service delivery and quality.

EU AI Act ready

ISO 42001 certified for AI management systems, one of the first companies globally. Built for compliance before the regulation requires it.

AI built and run completely in-house

All psychometric tests developed using IRT and Bayesian inference. No candidate data processed by external LLMs. No black-box scoring. Full control over every model update.

Every decision is auditable

We can show what data the model used, how it was weighted, and why it produced a particular score. Full traceability from input to output.

GDPR compliant by architecture

ISO 27001 certified. Candidates own their data. 24-month default retention with automated deletion. All data stored on Google Cloud within the EU.

“For us, it has been important to find a way of working that makes recruitment more fair, more accurate, and less dependent solely on CVs, without compromising on our requirements for security and compliance. Alva has become an important part of that work.”

Sofia Linden Falkenlöv, HR-Specialist & Team Lead TA, Post- och telestyrelsen (PTS)

Sofia-Linden-Falkenlow-Post-och-telestyrelsen

650+ companies already trust the science. Here's what they're seeing.

candidate assessments

.5M+

candidates assessed using our certified instruments

Candidate experience

%

of candidates complete their assessment

Candidate experience

%

candidate satisfaction across all assessments

Efficiency

%

reduction in time to hire

See the science on your roles

Book a walkthrough with our team. We'll show you how this translates into better hiring decisions for your organisation.