Why Workforce Analytics Shouldn’t Start with the Data

One of the most common models in HR analytics is shown in Figure 1. : the Data-Insight-Story model. The model starts with the data, then you draw insights, and finally tell a story. This isn’t entirely wrong; it’s just not the best way to approach analytics and it’s caused some heartaches for HR professionals trying to do the right thing.

Here is how this model usually plays out: An HR professional is given a report, usually data drawn from the HRIS, and is asked to look at the data, derive an insight, then tell a story. Normally, HR professionals find this very hard to do, not because they lack the right competencies, but because it’s genuinely difficult to find legitimate insights in rows of numbers and figures. Even if you were a data scientist with lots of data, tools, and time, it would be a challenge to find insights that would have an impact.

The secret to success is this: with workforce analytics, the story lies in the business issue, not in the data.

The story lies in the issue–not the data

Let’s imagine you have the following business issue: customers at a consumer electronics chain complain that the staff don’t know very much about the products. Everyone seems to have an idea why this is happening: Some think training is poor. Others think customers have unrealistic expectations. The regional head thinks it’s because turnover is high, which has left the stores with inexperienced staff who don’t know the products. The regional head is pushing HR to improve retention so that stores have more highly experienced staff.

We now have the right starting point for analytics: a business issue followed by several hypotheses about the cause of the issue.

Since the regional head is the senior leader in the area, we might test her hypothesis first. HR pros are asked “Is there any data showing that inexperienced staff generate more complaints than experienced ones?”

HR may be lucky and have complaint data for each associate so that they can directly test the hypothesis. However, it’s more likely that such fine-grained data will not be available. If this is the case, they might instead look at the number of complaints by store versus the average tenure at each location. This isn’t perfect, but let’s see how the insights and story might play out.

If the regional manager was correct, then we’d expect stores with fewer experienced staff to have more complaints. But imagine that the data shows little difference. Now HR should go to the regional manager and tell this kind of story:

• Issue: Customer complaints
• Hypothesis: Inexperienced staff are the source of complaints
• Decision we need to make: Whether to make a significant investment in retention
• Data: Complaints by store plotted against average tenure at the store
• Insight: Tenure doesn’t appear to have much impact on complaints
• Conclusion: We can’t be sure, but it appears likely that a big investment in retention would not solve the problem of customer complaints; something else is causing the problem.

This list of bullets is the story HR is telling. The regional head is sitting on the edge of their seat because they care a lot about the business issue of reducing customer complaints. HR had little trouble drawing an insight from the data because they knew exactly what hypothesis they were testing. The story flows from the business issue, which then guides the data collection and analysis.

Why is the Data-Insight-Story model hard to give up?

If it’s much easier for HR to start with an issue and derive the story from that, rather than staring blankly at a page of numbers hoping for insight, why is the Data-Insight-Story model so popular? I’m guessing here, but I think there are two reasons:

1. The Data-Insight-Story structure is a great storytelling model
It sounds cool for a food company to talk about how they combined big sales data with big weather data and, in looking for patterns, derived the insight that soup sales rose after a cold snap. It’s more likely that they started with the hypothesis that soup sales might be related to cold snaps, then looked at the data to see if that were true. The “we started with a hypothesis” version is more accurate; just not as exciting.

2. Analysts may be unaware of the first steps in their thinking
Ask a financial analyst about their work and they’ll say they look through the numbers for patterns to derive insights. What they don’t realize is that, based on their many years of experience, they have a whole lot of business issues and hypotheses in their mind as they start going through the data.

They are not just randomly looking for patterns in data, they are seeking answers to specific questions such as “Cash flow is an issue. I wonder if we’re having trouble collecting accounts payable?” They are actually starting with the issue, then seeking insight from the data, which leads to a compelling story about how to improve cash flow.

Whatever the real reason for the popularity of the Data-Insight-Story model, the truth is that it’s a hard way to go. In my work in analytics, I’d say this is the most useful insight I’ve had: When you are working on an issue people care about, then it makes the whole process much easier and plays closer to the HR professional’s strengths. When they are clear about the hypothesis they are testing,  you will find them much more capable of drawing insight from data.

Ready to share powerful insights? Don’t forget to download your Five Steps to Getting Started with Workforce Analytics white paper right here.