Predictive Modeling

Predictive models receive tons of attention for good reason: they have the potential to help organizations grow and people flourish.

But talk to almost any leader or HR professionals and you’ll see there is a sustained disconnect between potential and practice.

To understand why, let’s start with a basic definition.

A predictive model is just a mathematical tool that uses data to help predict future events or outcomes.

In some cases, we want to predict a categorical outcome (think staying v.leaving or promotion v. no promotion). In others, we want to predict a number (think percentage of first time resolutions at a regional call center or the number people in new roles in the next 12 months).

Regardless of your target, the basic steps are the same: gather historical and current data, develop your model, and see how well it predicts.

That’s it.
Fifteen or 20 years ago, developing predictive models for HR and workforce challenges was difficult and rare because the tools were cumbersome and data was
limited.

Today, the tools are better, the data more plentiful… and yet the results are wanting and the impact is limited.

What happened to the revolution?
We contend three key barriers persist.

Barrier 1: Lack of essential intent.

Someone creates a predictive model for you on say, voluntary employee departure, maybe even a highly accurate one. Great!

Now what?
More and more people are asking this same question with the increasing ease of predictive model generation.

As it turns out, this gap between model creation and business impact lies in the failure to ask what what you intend to do with the model.

Good model development start with key people questions, questions like

  • What result do you expect to get?
  • What will you do if you get that result?
  • What if you get the opposite result?
  • What decisions does this model impact?
  • Who is responsible for taking action?
  • Who cares?

A predictive model is only as good as the questions it seeks to answer.

From the outset, our predictive modeling approach stays focused on the actions and decisions supporting better business execution.

By asking you the right questions first, we help you effortlessly answer a thousand questions later.

Barrier 2: Prediction is not understanding

It might seem counterintuitive, but many times the most accurate predictive model tells us little about the processes underlying the outcomes we are predicting.

How can this be?
Models just detect patterns in the data. If we have a complex behavior and enough data, we might pick a complex model (or even an ensemble of models) to get the most accurate prediction. The upside of such modeling complexity is increased predictive accuracy.

The downside is that we’ll end up with a model that no human being can really understand or interpret.

And in some cases, an impenetrable but highly accurate model might be perfectly fine. If we are trying to forecast labor costs in the third quarter, for example, then all we really need to know from predictive turnover model is THAT some people are at higher risk of leaving, not necessarily WHY they might leave.

Now imagine generating a list of high potential employees at high risk of departure, sharing that list with HR leaders, but then telling them you have no idea why they have a higher risk because your “best” models can’t tell you that. That won’t work for them or you because it doesn’t support informed action.

Sometimes the “best” model is not the most accurate model but the one we can understand and explain.

By focusing on intent from the beginning, we maintain the distinction between prediction and understanding at every step. This helps you make the deliberate model- action choices that best serve your workforce needs.

Barrier 3: The skills gap

Yes, the analytics skills gap in HR and workforce development is real.

The good news? It’s closeable.

Any successful analytics project requires two pieces: domain knowledge and analytics skills.

But you already have that domain knowledge when it comes to HR.

Moreover, individuals with that HR domain knowledge have an ideal platform for developing HR analytics skills.

Why? Because starting with domain-specific problems like turnover, hiring, and promotion gives learners purchase on key analytics principles from the outset. It helps them map people questions to the techniques and this increases understanding and sustains motivation.

If you are looking to grow your internal HR analytics capabilities and impact for the long haul, you already have a head start and we offer the tools and training to help you the rest of the way.

For those further along the HR analytics journey and looking to close the loop on the just that final piece of predictive analytics, we can help you there too by providing both immediate predictive capability and a path to training your team for independent model development.

Like we said, the skills gap is real and addressing it takes time, patience, and vision.

But for forward-thinking organizations, we can help clear the path to the self-sustaining processes and predictive insights that drive decision-making, organizational performance, and human flourishing.

Contact Us

Yes, I would like to receive newsletters from HR Analytics 101.