Are you a HR or People professional who wants to really understand your employee feedback? Do you want to know how to increase team engagement, optimise new-hire onboarding, improve workplace culture, measure and track the motivation, well-being, and happiness of your employees?
Of course you do!
I’m not going to waste your time telling you how vitally important it is to listen to your workforce, how they’re the greatest asset any organisation has, how nobody knows your organisation as well as them, how if you’re not looking after your people you’re not looking after your business. You already know all that.
And I’m not going to tell you that now, more than ever, as there is a paradigm shift in working dynamic with an endpoint that nobody yet knows, the risk of not listening to your people is perhaps as great as it has ever been since offices became a thing. You definitely know that. Or at least have a strong inkling you need to bolster your employee engagement pretty damn quick. What I am going to tell you is something you might not know yet:
That there’s a new way to engage with your people and it’s all about story telling.
But like any good story, let’s set the scene first.
There is a lot of speculation by HR experts as to what the new challenges will be keeping our people happy, sane, and productive within this mid-Covid-19 evolving landscape. Nobody can be sure yet. But we can be sure about the best people to listen to for identifying existing and emerging problem points, or ‘opportunities’ as we should call them. Your people.
When you actively listen to your people you are tapping into the best information source available to help you make the necessary improvements and innovations. And when you make changes based on the actual feedback of your people, you are engaging with them and building trust. And they’ll realise it and positive momentum will build. Employees should be empowered to affect change by any organisation that wants to evolve in an informed way.
But how do you get at this bubbling reservoir of insight? You ask for it, like we’ve been doing for 100 years. Annual staff surveys, pulse surveys, virtual suggestion boxes, working groups. And for the last 2 decades you can use some increasingly smart employee engagement platforms that leverage emerging technologies to generate and harvest employee feedback.
Nowadays, leaving feedback is near frictionless. You don’t need to hunt down an HR manager, wait for a quarterly pulse survey, dig out the right form, or even find a pen. People can engage with survey portals and sophisticated chatbots (just like DIEMinnovation’s DUCHESS) on their phones and computers 24 hours a day. Had a less than stellar week? Open the app and leave some constructive, and often anonymous, feedback about your manager’s failure to grasp why the present budgeting process will never work. And whilst I’m here, if somebody doesn’t fix the blasted coffee machine it’s going out the window. (A good example of this is the FH Institution’s Happy App.)
Traditionally, analysing employee feedback has been limited by the capabilities of the technology used to understand it. Feedback on any useful scale could only be processed if the survey mechanism was highly structured and quantitative. You know the ones: “Rate your happiness out of 10”, with a small box for written feedback at the bottom of the form. This type of survey has been around since the 1920’s.
This style of quantitative survey continues to provide a useful service. It delivers a top-level metric that statistically quantifies, for example, the happiness level across the organisation. But what does a ‘level 7 Happy’ look like, feel like? And this type of survey tells you very little about what you need to do to make that person or department an ‘8’, nor what is making them fulfilled enough to even be an ‘7’ in the first place. It’s good for an impressive looking presentation to the board but what happens when they ask you what to do next?
“Thanks Sonia. So, the factory floor happiness level has dropped down to 6.75 this quarter. What changes do we implement to get it heading back up?”
That is hard to respond to when the quantitative data does not provide the answer. This is where text analytics and, more specifically, the qualitative survey enter our employee experience story. Qualitative text surveys are designed to analyse the experiences, emotions, attitudes, and behaviours of individuals at scale. Think of it as the survey equivalent of a ‘watercooler’ chat. This is where you are most likely to hear somebody’s real feelings and thoughts, expressed in an informal dialogue that tells a story.
These are the words and emotions that can provide the detailed insight your organisation needs to understand in order to fully and proactively engage with your people. It is a vital piece of the employee experience jigsaw that frequently companies are not aware is even missing. It can, for example, provide detailed evidence that explains why your factory-floor people are becoming increasingly unhappy. And this knowledge can be used to design informed improvements that result in the required outcomes. But only if you are can make sense of it.
The challenge with analysing written or spoken feedback is that it is highly unstructured and complex. People express themselves in a myriad of different ways, using different expressions, tones, metaphors, and languages. Historically, trying to make sense of human language had to be done by a human. It was impossible to use a machine to automate and scale meaningful analysis of human language as machines were too dumb. But imagine manually sifting through 10,000 feedback comments trying to populate a spreadsheet with common themes, topics, and the feelings being expressed towards them.
That is why the small box at the bottom of the traditional survey questionnaire was small and at the bottom. There was limited value in collecting it.
Nowadays there are a lot of marketing dollars spent on promoting the capability of certain Artificial Intelligence (AI) components to help us understand unstructured text at scale. Natural Language Processing (NLP) and Machine Learning (ML) are frequently touted as making a real difference in decoding human generated unstructured text. Unfortunately, and as Gartner and others confirm, at the moment this is mostly just hype. Skynet from The Terminator and Agent Smith from the Matrix have a lot to answer for.
The majority of text analytics platforms, that rely on statistical AI alone, provide what should be considered a very basic capability, that, in a nutshell, consists of identifying certain keywords and assigning a positive or negative sentiment score to them. This is known as Sentiment Analysis. For example, the machine will associate the words “lack of” and “communication” in the same sentence and give them a negative score. So, at the next board meeting you end up with this:
“Thanks Sonia. So, the factory floor happiness level has dropped even more and is now down to 5.82 and the most frequent negative keyword hits are ‘communication’ and ‘teamwork’. What changes do we implement to get it heading back up?”
Well, it’s obvious. You need to improve communication and teamwork. But if you’re Sonia, what do you base your improvement plan on? What areas of communication do you start with? Perhaps a company team-building day, or let’s revamp our intranet, maybe some cross-team workshops? All of those should help with our communication problem, right?
That is a lot of questions not to know the answers to. They highlight the limitations of Sentiment Analysis. It is a blunt tool unable to identify the individual emotions and themes in your text dataset that would pinpoint the exact problem points that are causing the drop in happiness
In our present reality, the ‘Machine’ is still too dumb to understand what we are talking about. It needs to be given the opportunity to learn, which is a long, labour intensive, and expensive process. Alternatively, rather than wait for it to figure it out for itself, we can give the machine certain rules about how human language works. This is like we are holding its hand and giving it a torch so it stops stumbling about in the dark, falling down stairs, bumping into things and generally falling over a lot.
We can define how to identify emotions in language, recognise reoccurring themes, the meaning of metaphors, the value of contradictions, and many other elements that our brains do automatically when we are listening to our colleagues talk. These rules are based on complex conditionalities and need to be manually coded by a human. But once these rules are created and combined with statistical NLP and ML, you have a Hybrid Text Analytics platform that can finally understand what we are talking about and be put to work on our qualitative survey feedback.
This is no longer Sentiment Analysis. This is a next-level text analytics capability that tags your unstructured text according to emotions.
Emotions like delight, excitement, frustration, anger and themes like morale motivation, autonomy, welfare support. It is not blunt, it is laser focused. It analyses the employee feedback down to the granular level. All of which means you can understand exactly what is causing the positive and negative emotions in your workforce that are driving particular behaviours. This capability is sometimes known as Emotion Analytics, or Psycho-Emotional Profiling, and it is a game-changer for anybody who wants to understand what their employees (and customers) are really saying.
Back to our story. Before the next board meeting Sonia gets her hands on an Employee Engagement Hybrid Text Analytics platform, for example Pansensic’s Employee Experience, and runs her unstructured employee feedback from the factory floor through it. It tags this sort of thing:
“….the new eating arrangement is disappointing as I never get to see my friends from the office”
“….not sure why they get the new coffee machines with the touch screens…”
Understanding that the organisation should be assessed as a whole, she then decides to take 5 minutes to run the employee feedback text from the office through it. The results highlight further problems but also provide insight as to what the solutions should be:
“…less time as I’ve been having more meetings with the shop floor teams recently…..”
“….like the idea of starting Factory Fridays again…..”
Sonia now has the information and evidence to devise and justify an improvement plan.
“So Sonia, what’s the plan?”
“We need to modify the new dual cafeteria set-up as it is causing a disconnect between the office and the factory floor.”
“But we’ve just invested £15k in creating the new cafeteria for our factory floor teams!”
“Yes, but let me show you the evidence that explains why it isn’t working, and why it will cost the company well over £15k this quarter alone for lost time and lower productivity. But don’t worry, this evidence also shows us where we need to make the changes, which we’ll monitor for effectiveness on a monthly basis.”
Sonia then considers performing a ‘microphone drop’ but thinks better of it.
Using Hybrid Text Analytics to understand unstructured text allows you to turn individual qualitative responses into quantitative data streams that plug the critical gaps in your existing employee data. It provides the most accurate and detailed assessment of your workforce.