Don’t just look at data; look at data about data

According to our analytics maven, David Creelman, the best (and the cleanest) data is actually data about data:

Article main image
Jun 28, 2023

All HR professionals will know that most people analytics is done using data from the HRIS, other HR systems, and surveys.

However, what many may not realise is that there is a whole other world of data, often from communication tools, that offers exceedingly fresh insights.

To explain what I mean, let me tell you about two people I recently had the pleasure of meeting – Denys and Elena Grabchak.

They both worked at Google before starting their own analytics company Performetry.

Both specialize in metadata analysis.

Metadata: the data about data

In its simplest terms, metadata is data about data.

For instance, a Slack message contains a message (the data) but it also contains information about when the message was sent, who sent it, and to whom it was sent (the metadata).

There is potentially a lot of information about people and teams that can be gleaned from that metadata.

And best of all, Denys says downloading metadata usually isn’t that hard.

Most systems have interfaces where – with the right permissions – you can get to this data.

Meta data is clean data

But perhaps even better than it being easy to download is the fact that metadata is typically very clean (unlike much of the data in the HRIS – where users typically have to clean up they data they need).

Let me paint a picture of just how clean this is.

In his work at Google, Denys looked at metadata from communication tools (email, chat, video calls, etc.) and calendar data.

The analytical process was to look for patterns to understand what’s normal for each team or manager or generation, and then look for deviations from those patterns.

What Google was particularly interested in was the attrition of high performers and by detecting changes in patterns of behavior it was possible to estimate flight risk.

Another objective of the analysis was to understand what high performers did differently from average performers.

Again, it was a matter of looking for patterns – but this is all very clean work. It simply requires a deep dive into the data to look for insights, since what makes a high performer may vary from group to group.

Elena says that what makes metadata analysis attractive is that there is so much data.

An employee may generate hundreds of data points every day. This allows people analytics pros to use machine learning and other AI techniques to study the data.

The takeaway for people analytics pros is that if they are not already analyzing metadata, then there is a rich world of analysis remaining to be explored.

The only cautions are that the usual issues about data privacy need to be kept top of mind. Denys suggests that when you do use machine learning, be sure to have “explainability” – that is if you just say the insights came out of a black box then managers won’t believe you.

At times if feels like analytics has got bogged down in reporting and data management.

A look at metadata brings some excitement back to the work.