HR Analytics: Interference's in your Inferences...!!!

We often hear about Analytics Being the next big thing whenever we speak about future of HR.  The shift in focus has been from descriptive to insight, from insight (what is happening) to predictive  (what will happen) and from predictive to prescriptive (What should you do?).   There are couple of things to keep in mind before organizations move quickly to jump the bandwagon in their quest for prescriptive outputs.  I am speaking from the HR perspective only.   

Consider this:  We are all data agents in the sense that everything we do in our lives today seems to generate some sort of data on a continuous basis.  The hand phone we use generates data on our movements (tracked by GPS and Maps),  the messages we put up  on Facebook or other social media sites,  the embedded chips in our cars send data back and forth to the servers,  often washing machines are smart and sending data to the transmission and distribution servers.  All of this data says a story about us.   Something that BIG BROTHER (various organizations that aggregate the data and keep analyzing the trends) can watch understand and know more about you.  

Similarly as members of organizations we are generating data all the time.   We have interview records, application formats filled in, employee demographics, performance data, attrition analysis data,  exit interview data,  snap shot surveys etc.

Now You will all agree that everything we do in analytics is based on data. Just like in our personal and social lives data is generated through us, by us in our organizations.   However there is a subtle yet important difference you need to consider.    Unlike data that BIG BROTHER is looking at, the level of consciousness in an individuals behavior is different when it comes to data in organizations.   People don’t know who is looking at the data, who is watching, who is inferring when it comes to BIG BROTHER kind of data.   So as long as the person on the other side is unknown the level of consciousness remains low and therefore the actions we do tend to be like ourselves. 

When it comes to organization data there is a high level of consciousness and therefore the behaviors get altered.   e.g Employees have to really be trusting of the organization if they give right inputs on the reason for their leaving. They have to be themselves.   Else you end up with wrong data.   Experience has shown that  people are very conscious about leaving with a “good feeling” and don’t want to give right inputs on why they leave.   This will surely lead to wrong conclusions if your analysis is on “prediction of which people will leave”.   Your model would be built on data that is not accurate.     

So before we jump to expecting that the predictive models we develop will work couple of important things to be remembered.

a1) Ensure that the process has sufficient amount of design considerations on interfacing parties.  For e.g. if exit interview is being done by manager and person has problems with work environment in the project / department then the responses may be biased and erroneous.  Similarly if  there is dissatisfaction with HR environment it can lead to distorting responses if the interviewer is from HR group to which employee belongs.   Choosing the right actors (Participants)  during the process is very important

b2) Validate the data you are generating in various parts of the organization. E.g.  take data from off line delayed feedback survey with employees who left.    Compare that with what was said at the time of departure. If you find a good correlation (higher the better) you have reliable data to infer. The same can be done for appraisal.   We all know how 360Degree feedback looks like.  If I were to go by that data in most organizations all of them should be to stars of their industry.  Such is the “high level” of capability that can be seen with 360 datac) 30  Keep revisiting the model you develop by adding data you generated post the earlier time and that way your model is strengthened.

d3)   Use the right frame of understanding before you decide what to test in your model.  It is very important to remember that if we ever want to have new ways to look at a problem we have to have new ways of thinking.  The frame has to be different to view a situation differently.

These three simple focus elements actually will help build that credibility when HR works on analytics because at the end of the day it is important that the insights or  prescriptions lead to tangible actions.