How do RCXUE algorithms compare with other assessment tool?

As a multinational corporate that has operation in China, one of the largest European food-product companies pays a great deal of attention to its talent cultivation. That’s why they have been investing a lot of its resources in campus recruitment.

When business grows, so does its need for a diverse type of talents. Thus it became challenging to identify the right candidate for the company, especially when there are all sorts of assessment tools focusing on different criteria to evaluate a candidate.

What makes the situation even worse is the fact that traditional assessment tools are very easy to be gamed. Largely based on multiple-choice questions, these tools are susceptible to applicant’s cheat. For instance, those who complete the test in the morning can publish the answer right away and it would be read by thousands of candidates by the afternoon.

In order to get reliable results, this company has to utilize several assessment tools to evaluate a candidate, making the work very time-consuming and inefficient.

Which assessment tool gives the best evaluation of a candidate? Is it possible to use only one system to solve the problems at hand? Such questions becomes the core challenges facing this company.

Solution

They decided to use RCXUE system together with 3 other assessment tools to score the 8,400 candidates of the year at each round of the selecting process. During each round of the recruitment, a candidate’s performance will be scored by the four assessment tools, and these system would give their own recommendations for candidates to enter the next round.

In fact, the decision on who to enter the next phase was still made by human interviewers, because later we’d like to compare the systems’ results with human judgements.

Results

RCXUE system demonstrated over 20% predictive power than the other three assessment tools. Among the candidates who passed the final interview, a larger share of them had been recommended by RCXUE. The predictability was even slightly larger than all other three tools combined – over 20% better.

The robustness of RCXUE system was also backed up by data of the candidates rejected during the selecting process. For those who were rejected by human interviewers, over 90% of them occupied the bottom 20% of our system-generated rankings.