Our machine learning measures each and every one these different factors, or data points, which fall under 5 main areas.

Thematic alignment: this indicates how close a given text response is to themes found in hired and performing hires in the past. A machine learning model using a rapidly growing pool of thousands of candidate responses is used for the identification of the relevant themes. These themes are aggregated into the Traits that are reported on in the TalentInsights report.

Readability: AKA how easy the text is to understand. This includes the complexity of vocabulary and syntax. We calculate a set of scores for standard readability measures such as Flesch Kincaid Grade, Dale Chall Readability Score, and the Coleman Liau Index to measure how readable the given text is. This contributes to the English Proficiency score in the TalentInsights report. 

Text structure and quality: this includes a total sentence count, word count, ratio of verbs/nouns/pronouns etc. We also look at the quality of spelling, grammar, and punctuation. This contributes to the English Fluency score in the TalentInsights report. 

Personality: we map the text answers supplied against the HEXACO framework to analyse the applicant’s personality. This contributes to the Personality Read in the TalentInsights report.  

Sentiment: A measure of how negative or positive (polarity) and how subjective or objective (subjectivity) the given text response is.

Once a candidates' answers are submitted, these different factors are analysed, compared with applicants who have been hired into the same role, and then allocated a recommendation. Generally speaking, in terms of the possibility of being hired,

Like with everything in life, this isn't a guarantee they'll be successful, but it's our prediction based on the factors we've mentioned above. 

We're not here to replace recruiters, but to help you use your time better. Think of us as you would about using Boolean search to locate candidates. It helps you distil a large pool down into a much more manageable one, but you still need to make the decision on who are the right people to move forward with.

It's almost inevitable that you will see someone who doesn't fit or isn't right for your role, particularly in the beginning. As we learn what good looks like for your business you can help us by categorising your applicants regularly so the model can learn from your data over time. 

Did this answer your question?