Machine learning and the future of employee feedback
Chief Scientist, Culture Amp
The basic premise of employee feedback is that understanding your people is crucial to making smart business decisions. It’s data-driven decision making, focused on using primary data from your organization as a direct line of communication to understand what’s really happening.
Our big picture aim at Culture Amp
At Culture Amp our aim is to create a symbiotic relationship between machines and people that helps companies ask and answer intelligent questions about their own people - rather than relying on outdated theories and generic prediction models. To do so, we’re looking to emerging technologies and experimenting with ways to make them work for us in service of this aim.
One of our long-term goals at Culture Amp is to free ourselves from preconceived ideas about what makes people tick in the workplace. Instead of reading a consultant’s book on best practice or relying on hearsay, we want to make it easy for anyone to go direct to the source (their own people) and collect the data they need to understand exactly what's happening in their organization.
Google arguably pioneered this approach in the first half of this decade, and its business success is at least in part thanks to how they’ve used employee feedback. What’s notable about their approach is they didn’t speak to external consultants, they captured feedback from their own people using surveys and other data collection tools. Once they had this data, they were able to challenge their own ideas and determine the most effective approaches for managing and growing their people.
That approach is more widely accepted now than it was in its infancy at Google. In fact, even today I think about employee feedback as an industry in its teens. Like any teen, we’re looking to the future and experimenting with different paths to determine the pros and cons. So as we continue to uncover the future of the field, the two areas I’m most excited about are deepening our understanding of workplace experience and using collective intelligence to help anyone take action.
Deepening our understanding of experience
In the bad old days, employee feedback was only able to measure a very narrow range of things because research was expensive and our tools were limited. Research and data analysis was a laborious process.
Today, employee feedback is substantially easier and more affordable. There are a number of platforms and tools available and that’s inspiring a lot of innovation. We can use employee feedback to understand more aspects of the workplace in much greater depth. For example, where most organizations once used to conduct a large annual survey giving a general sense of how happy people were at work, we can now get more granular and specific. We can regularly seek out people's thoughts on a wider range of issues, from where they're sitting to what their team relationships are like.
Not only are we understanding more of the subtle aspects of the experience that people are having in their workplace, employee feedback is also expanding to the extent that organizations are showing an interest in people's wellbeing outside the workplace. Employee feedback is progressively reaching much deeper to understand people on a more holistic level.
This is a key development in employee feedback, and it’s accelerating. Whilst on one side, this is empowered by better data collection platforms, on the other side, this is also empowered by machine assisted data discovery.
Machine learning is a huge buzzword in the employee feedback space, but the truth is it’s having a major impact in how people explore data more efficiently. As data collection gets more granular the amount of feedback is simply too large for any one person (or team) to deeply understand or process. Though sheer amount of data, all employee feedback gets to the point where we must at least somewhat rely on machines to tell us where to look.
As data collection expands, machine learning is only becoming more important in guiding people towards the most important things in their data. This is enabling employee feedback to move deeper, and at the same time consider a much wider range of issues. I don’t think many people in the industry appreciate just how constrained we’ve been by analyst processing time over past decades. As this limitation falls away over the next year or two, expect to see the applications and depth of employee feedback increase substantially.
Using collective intelligence to help anyone take action
At Culture Amp we define collective intelligence as:
A shared or group intelligence that emerges from the ideas, experiments, collaboration and/or competition of many individuals, leaders, managers, teams, or companies.
We see collective intelligence as something broader and richer than artificial intelligence (and the machine learning that powers it). Rather than just relying on algorithms to make predictions (or highly prescriptive suggestions), collective intelligence is a hybrid approach that collates data, questions, ideas and approaches generated by thousands of companies and practitioners (including our customers and community) and uses algorithms to connect you with the information that is the most relevant for you, your team or your organization.
Alongside machine learning, collective intelligence has huge potential to increase the impact of employee feedback. Collective intelligence is built on the idea of leveraging the experience of the smartest people analysts across thousands of companies. Whilst we’ll seldom be able to predict people on an individual level, in the short term we’ll be able to make powerful recommendations at a group level on the basis of a repository of millions of experiences and interventions.
This approach relies on hard and fast data, not hunches or hearsay. When coupled with source data from within your own organization, you can find out what’s most important to your people and focus in on ways other people have solved similar problem. That’s where we see the future of employee feedback - using real data to make better decisions that make the most impact for your organization.
Why survey data quickly leads to data overload in the absence of machine learning
Employee feedback surveys often contain around 50 rating questions (and at least two free text questions). People also filter the data using demographic groups such as tenure, age and gender groups and organizational groups such as locations, departments, teams, functions and managers.
Typically even with this minimal number of variables we are generating thousands of data points just for rating questions in a single survey. If we want to consider combinations of demographics and organizational groups then this number quickly rises into the millions.
As I’ve shown below, this means immense processing time for a person to work through an understand all the relationships and implications. This has all too often led to ‘analysis paralysis’ and people being either too exhausted or confused to get to the point of doing anything based on the data.
Even with just six data groups, the average survey has nearly one million combined data points - making comprehensive human analysis close to impossible.
Modern machine learning approaches to analytics can quickly sift through millions of data points in seconds and quickly show you the most divergent and important trends. If you’re using Culture Amp’s platform, then you’ll also see suggestions for focus via our Focus Agent.
Request a personalized demo Get started