You can’t create a great employee experience in one hit. Even if you’ve identified where you want to begin, it’s very hard to predict how your people will respond to new interventions. Instead of trying to design the ‘perfect’ employee experience, our suggestion is to take an iterative, data led approach that responds to user (employee) behaviour over time. If you’d like to know how to do that, read on…

Traditionally HR teams have spent a long time designing their frameworks and people processes – taking months (if not years) to develop the ‘complete’ answer. They’ve looked to the latest ‘best practice’ and HR expertise for guidance and typically rolled out one-size-fits-all solutions for the entire business. 

In the old HR world, measurement has been an afterthought. By the time we’ve got to implementing a solution, we’ve invested so much time and effort that we’re not overly keen to see evidence that our intervention might not be working. Where they have existed, metrics have usually been targeted at employee satisfaction, not tied to commercial outcomes.

This is a pretty unsatisfactory state of affairs. So how can we do a better job?

Here are three steps to taking a data led approach to employee experience…

1. Designing evidence based prototypes

Designing evidence based prototypes is about starting with good theory – the scientific evidence on what really motivates people, how we make decisions and how you can influence people’s behaviour. When it comes to designing new interventions to improve your employee experience, you should start with good science. If you’re looking for a good place to find well researched ideas, Science For Work is an excellent online resource. 

2. Taking an experiment based approach to implementation

There’s lots of exciting new research out there but there are very few replication studies and most of the research is conducted in “non-natural” environments. So even if you’re following the science, there’s no guarantee it will work in your context… You can’t predict how a complex human system will respond to a new intervention. 

So once you have good ideas on the table, you need to take an experimental approach to implementing them by setting up explicit tests.

To do this you need to agree what you’re going to measure in advance and build in a feedback loop before you release a new intervention. 

This helps you avoid two common human biases:

  • Confirmation bias – our tendency to look for data and information that confirms what we already believe. If you have people on the team or in the business that believe the intervention won’t work then they’ll look for evidence that backs up their view point. Agreeing the evidence in advance helps to mitigate this risk.
  • Sunk cost fallacy – our tendency to continue committing to something that we have already invested resources, time and effort into. If you don’t agree a feedback loop and what would constitute success in advance, you’re likely to want to continue with your new employee experience because of the effort you’ve put into developing it.

Agreeing what you’re going to measure starts with identifying a clear hypothesis: “If we do ‘x’, then we expect ‘y’ to happen.” Once you have your hypothesis you not only need to decide how you’re going to build ‘x’, you also need to decide how you’re going to measure ‘y’. This feedback loop on ‘y’ is an important part of your prototype.

Once you have your hypothesis, you then have two main options for setting up tests:

  1. Measuring pre and post intervention
  2. Setting up a randomised control or A/B test

Option 2 is only open to you if you can split your employees up. This will depend on things like the size of your company, your setting (it’s easier to split people up if they’re spread across multiple sites) and the type of intervention (it’s easier to randomise recruitment candidates than changes in an office environment).

If you’re small you’ll probably want to do a pre and post intervention test i.e. measure for a fixed period of time to get a benchmark and then implement the idea and continue to measure for same period of time so you can compare results during the implementation period with the benchmark.

If you have more employees, you may want to test interventions with a subset of your employee population. To make these kinds of tests fair, it’s important the subset is identified at random and as far as possible is representative of the wider population. Otherwise you risk learning things that work (or don’t work) for the subset but don’t have the same effect in the wider population. 

If possible (scale, setting and intervention dependent) you may want to set up A/B tests so you can learn which of your interventions are most effective. 

Exactly what and how you measure will depend on the size of your organisation, what you’re testing and what you’re aiming to improve e.g. engagement, retention, performance etc. 

However, the sorts of things you can measure include:

  • Impact on reported satisfaction i.e. how employees say it makes them feel
  • Changes in behaviour i.e. did or didn’t click on something; did or didn’t stay with the company)
  • Changes in other people’s perceptions of behaviour e.g. 360 feedback

The most important thing to remember is that you are measuring to reduce the uncertainty about whether your intervention is having the effect you want and to support better future decision making.

3. Commercially analysing your results 

You’ve started with a good evidence based idea and set up a test to measure its impact, now you need to review the results! Just having data doesn’t make you data led. You need to build in prompts to review the data and to use it to inform what you should build next to keep improving your employee experience. 

During your reviews you not only want to analyse the data you’re collecting but also compare your results with business outcomes. You’re looking to discover whether your interventions appear to correlate with or cause better business outcomes. This analysis is critical for making the business case for continuing to invest in the employee experience. 

To bring all this to life, let’s go through a specific example…

You develop the hypothesis: “If we give rejected candidates timely, personal feedback during the recruitment process, then it will improve their experience and they’ll be more likely to recommend us as an employer to their network despite being rejected.”

The hypothesis is based on good evidence – research has shown that people respond more positively to experiences where they feel recognised as an individual/personally significant and similar ideas have delivered good results when implemented in other businesses  

You now need to set up your experiment. To do this you need to:

  1. Decide on your intervention – how you’re going to give rejected candidates timely, personal feedback. Are hiring managers going to call every candidate to give feedback? Is it going to be done via email? When do you plan to deliver the feedback by? What’s the feedback going to cover? Will you give different feedback depending on how far they got in the recruitment process? Etc.
  2. Determine your metrics and measurement mechanism – how you’re going to measure the likelihood of rejected candidates recommending you as an employer. You could add an NPS question to your candidate feedback survey: “How likely are you to recommend us as an employer to your friends and family?” If you don’t have one, then you’ll need to introduce a candidate feedback survey. You’ll also need to agree how you’re going to distribute the survey to candidates.
  3. Set up your test group(s) – so you have different data sets to make comparisons between. You might want to roll out the candidate NPS survey first and allow this to run for an agreed period of time (or until a certain number of candidates have been rejected) to get a baseline NPS score. Then introduce the intervention to the business and, after another agreed period of time (or certain number of rejected candidates), compare the NPS score since the intervention was introduced with the baseline score. Alternatively you could assign a group of hiring managers to a test group (providing timely and personal feedback to rejected candidates) and keep the rest of the company as a control group (no change to current candidate management). All candidates would receive the feedback survey including the NPS question. This would then allow you to assess the impact of the intervention by comparing the responses to the NPS question between the test group and the rest of the company. 

Once your experiment is up and running, you need to build in reviews to discuss: 

  • Was our hypothesis correct? 
  • What else can we learn from the data?
  • How well was the intervention implemented?
  • What should we do next?

Wherever possible you want to be bringing commercial data into this discussion too. 

The more regular and frequent you make this test and review process the better. 

It is only by taking a data led approach to the design, build and testing of your employee experience that you’ll be able to iteratively improve. Without a good hypothesis, you won’t know what to measure. Without building the appropriate measurement mechanisms, you’ll never capture the data that will help you to discover what works. And without reviewing this data at the end of a pre-agreed test period, you’ll miss the opportunity to learn and inform your next steps based on the real experiences of people in your organisation. 

Finally, don’t abandon your judgement!

At this point it’s worth noting that data will only take you so far. We’re always going to have an incomplete picture of the whole system and we have most uncertainty at the start of developing new interventions. There’s a danger that we put too much weight on one piece of data and only see the problem through the narrow lens of what we’ve chosen to measure. It’s important we never forget to use our judgement too. Data isn’t truth, particularly in the people space. So even if you follow all this guidance around taking a data led approach to improving your employee experience, remember to exercise judgement throughout.

Click here to read about how we took a data led approach to improving the onboarding experience at Five Guys. Over the course of 5 months we reduced 90 day turnover by 20%.

If you’re already convinced this is an approach you should be taking in your business and want to talk through the challenges you’re trying to solve, please give me a call. I’m happy to help and share advice to get you started.