Assuring good job performance for Hemfrid using Hubert’s structured AI interviews
August 9, 2022
Viktor Nordmark
Hemfrid: Outcome and results from 15.000 completed interviews and 700 hires.
After almost a year of collaboration and more than 15,000 conducted interviews which led to approximately 700 hires, we evaluated Hubert’s accuracy as a first screening step. The results show that about 94% of all hires were considered successful hires.
“Since initiating our collaboration with Hubert, we’ve monitored the outcome meticulously when it comes to time saved, rating quality, and candidate experience. Our analysis of Hubert as a performance predictor further confirms our belief that it’s a great screening tool for high-volume recruitment.”
Jessica Ivarsson

To meet the large flow of incoming candidates, Hemfrid is using Hubert as a first screening step. All applicants are invited to a short introductory chat-interview where they can express themselves freely and explain in their own words why they are right for the job. The interview can be taken at any time and place through any smartphone, desktop or tablet and measures for example:

-      Hard/soft skills

-      Motivation

-      Previous experience

-      Relevant competencies

The outcome of the interview is a score between 0-100 which indicates how well the candidate meets the requirements. A threshold was set at 50%.

Score distribution

This is what the distribution of scores looks for the ~ 15.000 conducted interviews.

Roughly one-third of the scores were above the threshold (indicated in green). Candidates receiving a Hubert score above the threshold are considered a good fit for the role. These candidates are then manually reviewed by the recruiters at Hemfrid and many proceed to the next stages in the process. The middle group (blue) received a score just below the threshold. During times of exceptionally high demand, candidates that scored just below the threshold are being considered and evaluated manually by the Hemfrid recruiters.

The lower group (orange) consists of candidates with a lower-than-average score. These candidates do not meet the qualifications for the position and are likely to be rejected by the recruiters.

Interview score vs. Job performance

From a total of ~700 hires over the past year, all employees were evaluated by their team leaders and sorted into one of three groups. Either exceptional performance, average performance, or below-expectations performance.

Out of the 700 hired candidates, about 6% were placed in the below-expectations performance-group, 2% resigned or fell into long-term sickness (early dropouts), 12% in the exceptional performance-group and ~80% placed in the average performance-group.

Below-expectations performance among the total amount of applicants is very rare (0.3%)

From a total of 15.000 interviews, about 5% of the candidates were ultimately hired. Looking at the number of in the below-expectations performers from the perspective of the total amount of applications, the share is disappearingly low.


We had a hunch that Hubert did his job well, but the outcome of this study truly exceeded all our expectations. Being able to predict successful hires with over 90% accuracy is a staggering result that most screening methods don’t come close to. We are very proud and happy about the result, especially since we believe that there’s even more room for improvement and further optimization to the current method.

This is strong proof that the indications Hubert gives towards future job performance are reliable. Last but not least, the automation Hubert introduced has effectively removed both human bias as well as 75% of the manual work previously involved at the first screening stage at Hemfrid.

August 9, 2022
Viktor Nordmark
Give us a call
General inquiries
Swedish office
Vasagatan 28, 111 20 Stockholm, Sweden
Update cookies preferences