John was carefully screened before the offer. Everyone thought he would be a great performer. He scored highly when compared to previously hired employees who had excelled. He had the right competencies and the right cultural fit. All the analytics agreed that he was the “right” hire. Now it has been 6 months and things are not going well. The work team has not accepted John very well and his boss rates his performance as average.
His boss wonders if the new people analytics team is as good as it is cracked up to be. The leader of the analytics team had explained how the process worked. They gathered data from scores of employees and selected the ones with the highest recommendations and the best performance reviews. The reasoning was simple: if recruiters could find candidates with similar competencies and experience they, too, would be high performers.
If a company is investing thousands of dollars on an analytics team, shouldn’t it expect virtual perfection? Otherwise, recruiters should just stick to the old interview and guess methods of the past.
The reality is that analytics are a tool and need to be used with great understanding and skill. People are complex and we have not yet perfected an algorithm or a method that is foolproof. A simple correlation with previous successful employees may be simplistic. There needs to be a deep understanding of many factors before a reliable prediction.
There are at least 4 things that an analytics team needs to do, as well as make sure the hiring manager has a realistic expectation of the outcomes.
Deep understanding of the positon and desired outcome
Every analytics team needs to spend time learning about the position, the daily routines, the types of people on the team, and the outcomes that are expected. Most work is complex and has many dimensions with aspects that are hard to articulate. There may be subtle interactions between team members that make a huge difference.
There may be the need to communicate outside the work team that requires a different communication style or deeper knowledge. There are many variables and they need to be understood and weighted before a solid prediction model can be developed. No model will ever be perfect and it will need to be continuously tweaked as hires are made and they are successful or not.
Agreement on the candidate’s characteristics
Some jobs require a strong cultural fit, others a foundation of skills and knowledge. Or a job may require that a candidate be highly motivated and passionate while other jobs may simply require that a candidate have done the job before.
Although every position requires all of these, they do not need to be the same proportion. In some jobs it is more important to be motivated (e.g. sales) and in others highly skilled (e.g. Engineering). Each of the four factors needs to be weighted for importance and integrated into the model.
The ability to use our own judgement and integrate both quantitative and qualitative data
Not everything about a candidate can be quantified, no matter how much we wish it could be. The level of modelling we can expect from most analytics functions is not going to be sophisticated enough for us to rely on solely. Humans are still very important in the final decision process. Hiring managers should feel comfortable making a decision that veers off from the recommendation if they feel they have a reasonable objection.
Some researchers believe that we will never be able to rely only on a machine prediction and that intelligent augmentation, or in other words human decision making, will be needed. And there are problems with presenting managers with recommendations. It is human nature to downplay objective facts in favour of our own gut reaction. It is also possible that a hiring manager reacts the opposite way and is so convinced that the recommendation is on target that his expectations for the new hire are unrealistic and he is overly critical.
The methodology to continuously improve and learn from the misses
There needs to be a clear understanding that at the beginning of an analytical process the results may not be correct. The models that an analytics team develops need to be modified on a constant basis incorporating new data and changing how the algorithm weighs various factors. Over time the models will become more useful and more predictive of success. But this takes the understanding of hiring managers that have been prepared for this and work as partners not as adversaries. Good analytics are the result of a journey that takes time to deliver a strong result.
John may not have worked out for many reasons. Maybe the model was too simple or used the wrong variables. Or maybe he was just the exception. What is most important is that no one gives up. Try to analyze why John has not lived up to expectations and whether that was the result of an incorrect prediction, changing job factors, the over expectation of his boss, or something else.
Recruiters have not had a set of tools that could help them make better decisions until recently. It is important to use the tools weighing recommendations with experience and common sense. The danger is that analytics gets a bad reputation before they have had a fair trial.
Two weeks till the People Analytics & Future of Talent conference – book your tickets now!
Leave a Reply