The more variation there is in applicant responses, the more accurate the predictive model can be, producing even more accurate predictions and recommendations for you.
Much better variance is detected through text than from the multi-choice questions. You and I might pick the same response for a number of multi-choice questions, but the way we would each respond to a question like ‘what motivates you’ will have much more variation.
More variance = greater ability for the machine to differentiate between applicants traits = higher predictive accuracy.
Additionally, gaming is reduced through free text questions as there is no right answer.
At the same time, we are always reviewing the predictive accuracy of our multi choice questions against performance data we receive from our customers. We remove those questions that show consistently lower predictive accuracy and so over time have reduced the number of multi-choice questions to a core set.
The uniqueness of our product model is in the combination of behavioural science and data science. We define the personality profile of an individual and put it through the lens of Machine Learning algorithms. This takes into account many thousands of combinations and finds correlation between those profiles and performance.
Other elements of uniqueness and IP are in the questions themselves, which are designed to gauge a particular construct without being influenced by unrelated elements (like gender, age, etc.). This allows us to collect objective data on individuals.
Text as data is a highly reliable mirror to personality and beliefs. Text responses serve as a digital truth serum. Our algorithm translates the response into personality and we match that personality text data with the same text data from others performing in the role. This combined with the answers to the multi-choice questions gives you much higher accuracy levels for predictions.