As Heather said there has been lots of discussion (here, here, here, here, and here) recently on recruiting metrics and the quality of a hire. Andrew Marritt provides a very interesting perspective on why quality of hire is even only appropriate in a limited number of circumstances.
All in all a fantastic discussion. But what does it all mean?
A metric should not be used in isolation, not all metrics are suitable for all circumstances and the biggest you have to be able to compare apples with apples, not oranges. Let’s not forget that at a base level your data must be accurate.
But let’s also look at the whole discussion from a information management point of view. A while back I saw a breakdown from the Meta Group on Workforce Information Management that highlighted the typical 5 categories that organisations go through when implementing workforce analytics. These were:-
- Information distribution
- Metric delivery
- Contextual embedded analytics
- Correlated analytics
- Predictive modelling
The discussion to date on quality of hire seems to be focused around the need for correlated analytics and predictive modelling, while the tools and methods available today a more focused on metric delivery. Therefore a gap in what we want to deliver and what our tools and technique enable us to deliver.
Correlated analytics allows us to relate workplace information with overall enterprise data that then allows us to understand the impact of workforce investment. A typical example of this would be the impact of turnover within a job grouping on revenue. While predictive modelling would allow us to generate views of the future using time-sliced data in multi-scenario analysis. Such as a 1% increase in employee satisfaction will result in a reduction of time to fill by 10% and a 2% reduction in turnover within six months with the related linkages to revenue and bottom line profits.
Now with most organisations stuck at the bottom two levels when it comes to their analytics programs it is no wonder why we are unable to meet the demands on how to articulate the success of recruiting. The question we really need to ask is when will we have the tools and methods to move up the value change to deliver these more advanced reporting environments? We need vendors to deliver real solutions that are usable in a majority of organisations, not just in brand new installations or those on the bleeding edge.
Michael Specht 
Michael,
Thanks for moving this discussion on. The clarity of the Meta Group model is really useful.
Whilst steps 4 & 5 are the goal I am still not totally sure they can ever be achieved. I do, however believe that you need to take a dynamic modelling approach where you can integrate feedback loops, intangibles etc. There is a good overview of this at the Canadian Statistics site: http://www.hrma-agrh.gc.ca/hr-rh/psds-dfps/dafps_f_model1_e.asp
I am a really strong believer in this type of work as a way of understanding the dynamics of what is going on but I’m not sure that running real numbers through is that useful. Depending on how far in advance you predict you’ll get an error from acceptable to one that makes the whole process useless. The problem, which I mentioned in my post, is that organisations change, especially in non-process dominated work. These changes invalidate the model pretty quickly.
Andrew the Candadian link is very interesting, and looks like a great method to model HR processes and interventions. While changes in non-process nominated organisations can invalidate the model I would expect that within certain boundaries models could still function like and overall map.
I guess if you look at things like a coach of a football team, you can map out all of your moves and strategies before the game, understand the statistics from both sides and develop a plan. During the game your team tries as best as possible to implement the the strategy within an even changing dynamic environment. Once the game ends you then sit back and analysis the results and compare with the entry plan. The cycle then goes on. As you say you cannot predict too far into the future otherwsie the model does not work, for example a coach planning the strategy for the last game before they have even played the first will not work.
Taking this into an HR environment I feel you can use the types of tools to help predict certain events within certain boundaries. But like a game of football if the boundaries/circumstances change your team (management and employees) need to have the skills to make dynamic changes based on skills you as an HR professional have provided them. This is where the rubber hits the road.