What is R-Squared - YouTube

ConnectCubed: Understand Your People Through Data

ConnectCubed helps you predict candidate performance using statistical models.

Statistical models take data and predict an outcome.

Let's consider an example.

Here are all of the houses sold in your town last year.

If we look at all the houses on a graph, we can see that in general,

larger houses have a higher price,

and smaller houses have a lower price.

In fact, we can draw a line that approximates this relationship.

This line is our model.

With this model, we can pick a size for a house,

and the model will predict the price for that house.

But no model is perfect.

The model predicts that this size house should cost this much

but a house this size sold last year for more.

Because the model is not perfect, we want to know how accurate our model's predictions

are.

To measure accuracy, we use R-squared, which runs from zero to 100%.

R-squared tells us how much of the difference in outcome is explained by the model.

In this case, we want to know how much of the total difference in price is explained

by our model, which uses data on house size.

Some houses sold for a very low price, some sold for a very high price.

How much of that difference can be explained by differences in size?

If our model has an r squared of 70%, we can say that the size of the house explains 70%

of the difference in price.

But what about that other 30%?

The other 30% comes from all the factors we don't know about.

Remember, all we know is how big the house is.

We don't know about whether the house is in a nice neighborhood, if the house needs significant

repairs, the style of the house, or even the number of bedrooms.

So while the model tells us that bigger houses have a higher price, because of all the other

factors we don't know about, it's prediction will never be exactly correct.

Using models to understand your employees and candidates faces similar constraints.

There are many other factors about your candidates that may have an impact on work performance.

Things like…

Teamwork, creativity, previous work experience, and other skills.

So what does this mean for you?

Let's consider your company.

Assume that you have 2 employees, Bob and Jane.

Bob and Jane are customer service agents.

Bob received an interview score of 48.

Jane received an interview score of 85.

When we look at your company, we notice that employees in customer service with higher

interview scores tend to get higher performance reviews.

Here we have an r squared of 30%.

That means that if you know an employee's interview score, you can explain 30% of the

difference you observe in work performance.

Being able to explain 30% of the difference in performance is a huge help!

We have figured out 30% of what makes a good customer service employee.

But what about the other 70%?

Let's look at Bob and Jane.

Here we can see Bob on the graph

and here we can see Jane.

Jane's work performance is higher than we would predict using the interview score.

Why?

Maybe she's very experienced, maybe she's helpful and mentoring to colleagues.

And what about Bob?

He performs even worse than just the interview would predict.

Why?

Maybe he hates his boss.

Maybe he arrives late to work.

Those behaviors are not captured in the interview and they make up the 70% the interview cannot

explain.

How much of the total difference among the employees was explained by the interview?

Here we can see the total difference predicted by the model using interview scores.

When we compare what the model predicts to the total difference in performance in reality,

we see that the interview explained 30% of the total difference in work performance`.

That 30% is...r squared.

For most companies, interviews have an R-squared of about 30%.

When hiring, you select the candidates with the best interview scores.

Unfortunately, with an R-squared of 30% some candidates who interview well are low performers.

Even though you can avoid some low performing candidates with your interview process, many

low performing candidates will still sneak through.

But let's look at what happens when you combine an interview with a ConnectCubed assessment.

By using both an assessment AND an interview, your total R-squared will increase.

This means that when you now choose the top candidates, you will have fewer low performers

and more high performers.

In fact, large scale studies have shown that combining an assessment with a structured

interview can increase your R-squared by 24%.

Increasing your R-squared by collecting additional information through assessments means an improved

ability to distinguish top performers from low performers BEFORE making a hiring decision.