I’m at the London Data Festival today, attending sessions from the HR & Workforce Analytics Innovation Summit, as well as the other conferences taking place here - including the Predictive Analytics Summit and a session from Peter Grindrod - Professor in Mathematics at University of Oxford on the Social Customer.
Peter’s view is that a lot of the current focus on data and analytics is a red herring because what we really need to focus on is our decisions. Plus we often end up talking about different things without realising it. Eg social data includes lots of people, lots of behaviour but tends not to be structured. Unsupervised distribution (clustering for grown-ups) can be used for other pieces of data but this doesn’t sit well with social, behavioural, unstructured data.
The issue isn’t about needing better data or systems. That’s a bit like suggesting that if Leeds United had a better pitch if would win with European Champions. But actually you still need Bill Bremner to turn up and do his magic. (I grew up in Leeds too so this point resonated well for me.)
For social analytics this is about developing understanding from the data. You can either develop a model without theory - based on a summary of what you’ve seen, the empirical observations you’ve measured, which is fine as long as you don’t have to explain the model. Or you can use theory based models to describe the observations with theory in the background.
The advantage of the second of these is that when the model fails it tells us something - either there’s a fault in the data or a fault in the theory - which can lead to a new paradigm (though we’re often loathe to give these up) which may be useful to a business.
Eg Peter showed us some analysis of social communities in Leeds and Bristol. If you put Bristol’s communities in a bag and took them out could you create Leeds? - no, they’re different social networks. It’s a faulty idea - what will work in one place won’t work in another. Or many things are time dependent - something may work once but will fail the next time because of dynamic instability in the system.
(This is why I don’t think Gartner’s 4 types of analytics featured by Tibco in the previous session works for social analytics work. They talked about sensors working out what needs to happen and sending an alert or a requisition for a new part - but you don’t want to do this in Marketing or particularly in HR. It’s why there’s no difference between diagnostic / descriptive and predictive / prescriptive analytics in HR. For us, data and analytics provides intelligence but action needs human interpretation.)
Ethics is important too, beyond just privacy and security. There needs to be a balance between the benefits gained by the people providing the data and the firms using it. But often we cross a creepy line intruding on customers to a greater extent that they feel comfortable with. It’s about a perception of fairness - a feeling, you don’t need to do the arithmetic. But gaining trust requires a level of transparency, even if you’re using proprietary data and systems.
So analytics project are never just about throwing money at projects, we also need to integrate the people. That’s why HR people should be interested in analytics around the social customer, but I think Peter’s points apply well to social employees i.e. to HR work as well.
Ethics is important too, beyond just privacy and security. There needs to be a balance between the benefits gained by the people providing the data and the firms using it. But often we cross a creepy line intruding on customers to a greater extent that they feel comfortable with. It’s about a perception of fairness - a feeling, you don’t need to do the arithmetic. But gaining trust requires a level of transparency, even if you’re using proprietary data and systems.
So analytics project are never just about throwing money at projects, we also need to integrate the people. That’s why HR people should be interested in analytics around the social customer, but I think Peter’s points apply well to social employees i.e. to HR work as well.
Let's cut down on the confusion by being clearer about what we need to measure, and using this to identify how we're going to measure it (instead of thinking that everything can be populated through an engagement survey). And let's start analytics work with some theoretical models about what we're trying to achieve. Above all, let's remember that our data belongs to our people and use analytics as a means of helping people perform better, rather than as way to ratchet up already high levels of control.
- Consulting Research Speaking Training Writing
- Strategy - Talent - Engagement - Change and OD
- Contact me to create more value for your business
- jon [dot] ingham [at] strategic [dash] hcm [dot] com
