22 July 2019
Recruiting through Predictive Analytics: A Beginner’s Guide
After reading this blog you will know:
- Predictive Analytics in recruitment: A quick explanation;
- Why data beats instinct;
- The required data set to start implementing Predictive Hiring;
- How to collect this data;
- How to make the first steps towards Predictive Hiring;
- Categorize your data;
- Create a wish list;
- Introduce a feedback loop;
- Equalture’s pre-selection technology for Predictive Hiring.
In other words: You’re ready to rumble!
Predictive Analytics in recruitment: A quick explanation
Using historical data to predict future outcomes. That’s basically how Predictive Analytics work.
90% of all recruitment professionals is used to manually pre-selecting candidates based on a resume. Of course you get better and better, but this shortlist will always be partly based on gut feeling instead of a data analysis when working with a manual resume screening.
Why data beats instinct
HBR conducted a research in 2014 to find out how the use of data in recruitment influences hiring decisions. At that time Artificial Intelligence (AI) wasn’t even a hot topic in recruitment, so this research was just based on ‘simple hiring algorithms’ (imagine how shocking the results must be at this time..).
HBR’s research showed that a simple equation outperforms human decisions by at least 25%. Twenty-five percent! And now we’re just talking about a formula.
These are the three reasons why:
- People aren’t as accurate as formulas. Yes, that’s perfectly normal, but it’s also a recipe for making mistakes. And even the smallest mistakes might lead to bad hiring decisions.
- People are inconsistent. Again, this is not more than normal, but it might lead to the fact that, besides not using all information, we can also interpret the same information differently at different moments. And again, this might lead to the wrong decision.
- People are biased. Based on our preferences, friends, family and habits we might wrongfully reject someone because this person is for instance not interested in the things that you’re interested in. This might lead to the fact that you could have wrongfully filtered out the candidates with the highest potential, therefore increasing the chance of making a bad hire.
So that’s why I believe in using data to make hiring decisions. Not to replace humans in the process, but just to continuously challenge their thoughts.
The required data set to start implementing Predictive Hiring
The infographic above shows the data we believe is required to predict a candidate’s success. These four categories combine track record (i.e. historical data) and potential (i.e. predictive data):
- Education: Shows historical achievements in terms of intelligence.
- Work experience: Shows historical achievements in terms of intelligence and personality.
- Skills and Accomplishments: Show historical achievements in terms of cognitive traits and intelligence.
- Potential: Predicts the future successfulness rate in terms of team fit (personality) and job fit (cognitive traits and intelligence).
How to collect this data set
Now all we need to do is making sure we’re able to gather these different types of data to start building a Predictive Analytics model for recruitment.
(1) LinkedIn (or a resume) for historical achievements
The green ones (education and work experiences) are the easiest ones to collect, because this is the information you will find on someone’s resume or LinkedIn profile. I would recommend you to use LinkedIn as your source for this information since resumes are increasingly being criticized.
Skills and Accomplishments is an interesting one, because some aspects of this category can be gathered by trusting on a candidate’s input as well as conducting a test (for instance a language test). Since this type of data is relatively easy to validate in an interview I would recommend you to also use LinkedIn (or a resume) for this information.
(2) Assessments for potential
Something that’s harder to validate is a candidate’s personality, intelligence and cognitive traits. Therefore I would recommend you to introduce an assessment during the application procedure (preferably in the very first step of the application so that you already have this data when pre-selecting your candidates).
An example could be neuro-assessment games: Gamified assessments that assess candidates on the three aspects mentioned above. You can read here why I would recommend to use games instead of traditional assessments in recruitment.
So now that you know why data beats instinct, which data we need to do so and how to collect this data, let’s go to the final (and coolest) part: The first steps towards Predictive Hiring.
How to make the first steps towards Predictive Hiring
As I mentioned in the first alinea of this blog, Predictive Analytics is all about using historical data to predict future outcomes. This is how to do that for a specific job opening.
Step 1. Categorize your data
This is what we did in the previous chapter of this blog. It’s very important to categorize and structure your data to make it possible to compare different data sets.
When we started creating a recruiting algorithm, the categories shown above were sacred to us, meaning that every single piece of data needed to be categorized. And if something didn’t fit into this infographic, we just left it out of our model. Discipline is everything!
Step 2. Determine the importance per category and create a wish list
Now take one of your job openings (for instance a Development position) and determine (1) your wishes per category and (2) the importance of the different categories.
Here @ Equalture, when we want to hire a Software Engineer we don’t place much value on their education; for us it’s all about the potential and the skills and expertise an applicant has gained over the years. For a position like this, we will use a distribution as displayed underneath:
The next step is to determine what we desire per category. This is an example:
- Work experience
- Min. 2 years of experience
- Preferably in the field of Software Engineering/DevOps Engineering
- Skills and accomplishments
- Experience with the following programming languages:
- Professional level of English
- Experience with the following programming languages:
- Personality/Cognition: Does this candidate fit into our way of working and interaction (for instance agile working, a high level of adaptability and a high level of autonomy);
- Intelligence: High level of creative thinking, analytical thinking and accuracy.
Step 3. Collect this data
Since you’ve finished your wish list, it’s now time to collect this data per candidate and assess how close these candidates match with your wishes. Please keep in mind here that in step 2 you also determined the importance of the three different categories; this should be taken into consideration when evaluating your candidates.
Step 4. Introduce a feedback loop
The last step is the most important one when implementing Predictive Hiring: gathering feedback.
To be able to continuously improve your success predictions it’s necessary to assess how successful your predictions were in the past. In other words: is the candidate with the highest ‘score’ (based on your wish list) as successful as predicted now that (s)he is working at your company? And if the answer is no, which of the three categories is most ‘disappointing’?
After answering these two questions you know where to improve your wish list. Et voilá, you actually started using historical data to make a better prediction of a future outcome.
Equalture’s pre-selection technology for Predictive Hiring.
Although you can always start experimenting with Predictive Analytics manually, there are also a lot of technologies out there to help you collecting, interpreting and improving the data sets you need to successfully predict a candidate’s success. Harver has for instance proven itself as a really valuable pre-selection software for high volume jobs, such as call center- or retail jobs.
Here @ Equalture we are very much focused on jobs on bachelor/master level. Our AI-driven matching algorithm matches all candidates based on the data set you’ve created in our platform. By having all this information you can easily pre-select candidates and create a shortlist for the 1st interview. On the background we continuously track the decisions you make (who did you reject/advance) and we ask you for feedback after hiring someone to be able to continuously let our algorithm improve itself.
Keen to learn how to collect these different types of data during a candidate’s application and how we feed our matching algorithm with Predictive Analytics? Just let us know, we’re more than happy to chat about the magic of Predictive Hiring!