Beware Workforce Vendors Using Predictive Workforce Language But Offering Non-Predictive Solutions

new-blog-pic-from-gretaTalent Analytics, Corp. has a unique approach to workforce predictive analytics.  At our firm, we measure success by how our projects quantifiably benefit the Line of Business.  We watch it, track it, and report success.  Our algorithms get better and smarter using the best Data Science methods available.

I’ve been involved in data science for almost 2 decades, but the workforce arena is newer for me.  I have to admit I’m surprised at how many vendors claiming to reduce employee turnover or increase employee performance do little more than offer a solution that “sounds” effective.  They say the right predictive analytics buzzwords – without showing / proving that their solutions actually work for their customers.

An example is the global multi-billion dollar market of pre-hire talent assessment vendors.

Most talent assessment vendors put their energy into creating “validated questions” that measure interesting human factors.  This is called Content and Construct Validation and just the first step in delivering business usefulness from a pre-hire talent assessment.  This level of validation says nothing about whether or not their surveys are proven to deliver business value.

Few vendors go the extra (and difficult) step of Criterion Validating their talent assessments –

My observation is that in the last 30 years, businesses and their vendors have moved away from using economic measures of business outcomes in solving problems with staffing.

We’ve become more obsessed with employee engagement, job satisfaction, fancy software and recording every single activity our employees do instead of focusing on empirical evidence to prove our initiatives actually deliver value for the business, or not!

Predictive analytics delivers outcomes not possible without algorithms.  Everyone wants their solution to be predictive.  This creates an industry of non-predictive vendors using predictive-sounding phrases like “the highest probability of success” and “pre-hire predictions” and “when the solution was implemented the company stock price increased”.

It’s hard for prospective buyers to be able to tell the difference; but we owe it to our companies to do so.  Ask for proof.  Ask for Case Studies.  Ask for documented results.  Ask for details about their predictive process.  Read about their data scientists on staff.  Push to learn more beyond the words they’re using.  Just because they say “predictive” or have the word predictive in their company name doesn’t mean they really are.

And for sure, if you’re using talent assessments ask if they’ve been able to repeatedly Criterion Validate their talent assessments.   In the end, business results are the only measure that matters!

By Carla Gentry
Data Scientist
Talent Analytics Corp.

Posted in Tags: , , , ,

One Response to “Beware Workforce Vendors Using Predictive Workforce Language But Offering Non-Predictive Solutions”

  1. Sunil Kappal Says:

    December 30th, 2016 at 10:12 am

    I believe you made a great point when you say “Most talent assessment vendors put their energy into creating “validated questions” that measure interesting human factors” and per my understanding this way of looking at the employee turnover reasons might introduce “ecological Fallacy” when the vendors start making conclusion about the employees based on the limited data that they have based on their analysis construct and might not be that robust. For Example; you might run into an employee who belongs to a group of new joinees who left the organization within the first few month due to various behavioral issues and you think to yourself “this employee must be like the other and might leave in few weeks or days to come” Here you go ! Fallacy! just because this employee comes from a group with a specific trait it doesn’t mean that this employee will automatically leave in coming days.

    Therefore, it becomes very important to have an algorithmic way of defining the employee turnover reasons by including various disparate data sources like HR files, probably psychometric tests, previous employment details, social media activities etc.This will not only make the algorithm better to predict outcomes but will also consider a 360 degree view (social, Professional & Personal) of a potential employee.

Leave a Reply