There is a big interest in using predictive analytics within HR but many organizations don’t have someone on staff to do this research. What are some ways that companies can do this research and learning with limited resources/experience?
1 Like
Predictive analytics is talked about a lot, but there are a lot of hurdles to practically achieving this in an organization.
In my view, it’s best to focus on these hurdles first, and let that create the demand. This lets the case for Predictive Analytics build itself – Rather than jumping to Predictive Analytics and facing an uphill battle to get things done.
In practical terms this is:
-
Get executive buy-in for the problems you want to predict (“solve”). Ideally this should be questions the executive really want answers for. The more specific the better - “how do we have higher retention in sales?” This is really defining “success” for your organization. Sometimes you’ll feel this too narrow, but going broad is usually too big a risk.
-
Decide on a minimum maturity level of your HR practices and data. It’s a mistake to try and make things “perfect.” Data and processes in HR can rarely ever be definitive. However, it’s actually much more likely to go the opposite direction – e.g. You can predict the effectiveness of new hires if you don’t have structure with your onboarding process.
-
Make sure the processes and data you have will support the questions that you’ve got buy-in for. If it’s around retention that’s employee and exit data. See if you can leverage relevant case studies to build a case for solving your specific problem.
1 Like
Predictive analytics (PA) is a term that lots of people talk about but few really understand or need to achieve their goals.
What is Predictive Analytics?
PA is a form of advanced analytics which examine survey, demographic, and other data to answer the question “What is likely to happen if X is true?” PA is characterized by techniques such as regression analysis, forecasting, multivariate statistics, pattern matching, predictive modeling, and forecasting.
How does it differ from the stats we already use?
PA makes use of the same statistical methods already in use by HR researchers and practioners. What is new is that data scientists are now using their computer science skills to clean, format, and merge large, diverse datasets stored in HRIS, Sales, Inventory and other data systems that have not traditionally been used in HR analytics.
What is great about it?
When considering large groups of employees, PA can help you identify the drivers of things like sales performance and turnover. By combining data from areas outside HR you can see unexpected relationships between things like real estate budgets and turnover and or employee wellbeing and sales performance.
What is hard about it?
PA often turns up unexpected and potentially confusing results. For example, PA could find that everyone in your organization who eats pizza 2+ times a week are 50% more likely to leave. Does that mean pizza is bad for retention? Or is frequent pizza eating a sign of other retention issues (e.g. compensation or unattractive office locations)? PA might find a powerful relationship but it may still require an expert to interpret those results to identify useful actions.
What are the risks?
PA is used to anticipate future behavior among large groups of people. When applied to individuals it is less accurate and can result in mistaken predictions (especially since individuals can have lots of unique characteristics not be accounted for in the data sets or the analysis).
Given the high stakes of PA in HR, these predictions (especially incorrect ones) can have significant impacts on people’s lives. Given that few companies would share the details of their analyses with employees, those employees would never know the why behind what happens to them.
Should I pursue a PA solution?
That depends on a couple things:
Do you have easier options to get the answers you are looking for? For example, employees are often eager to tell their managers what would make them more engaged and successful. Often simple feedback mechanisms matched with a willingness to listen to and act on feedback will get you the same insights for a lot less effort.
Do you have enough data?
PA requires a lot of data and works best when you have lots of employees and other non-HR systems and outcome measures to plug into an analysis. If you don’t measure a thing you can’t include it in PA and you might miss a key driver without knowing it.
Are you willing to change?
All data analytics efforts do is identify problems and potential solutions. Leadership must still hear those insights and change how they manage to achieve new results. If your organization is pursuing PA to try to find ways around dealing with well-known issues (e.g. cutting back on pizza rather than addressing poor salaries) PA will be a lot of effort for very little ROI.