A lot is being called analytics today – so it’s time to remind ourselves what ‘true’ analytics really are, and use this definition to force the abusers of the term analytics to call their offerings something else.


Let’s remind ourselves what analytics are
  • Analytics take an action for the user.
  • In a business application context – analytics may also recommend (but force ranked) a number of actions.

‘True’ Analytics exist

Here are some examples of older and more recent ‘true’ analytics in action

  • When your ABS system in your car decides to kick in for you.
  • When the auto pilot in a plane takes action on course because of a deviation.
  • When Google Now decides to wake you up earlier because traffic is bad.
  • When FICO’s Falcon declines or approves a credit card transaction.
  • When a recruiting application force ranks which applicants to talk to for higher hiring success

Why do we seldomly see ‘true’ analytics?

‘True’ analytics are hard as they must use a predictive model to come to the action or at least to the recommendation. And the result of that model must be a better outcome for the user (e.g. the ABS breaking for you) or reasonable (the list of recommendations is trusted and followed). To find, validate and have the business user trust these result of these predictive models is a significant challenge for any software provider. Intrinsically business users – rightfully or not – think they know better than the ‘machine’ – so it needs some trust building before a business user will unleash a model.

Easy and tougher use cases

It’s easy to trust the model when humans don’t have a chance to make the call. Take the FICO Falcon fraud use case, the 0.2 seconds to make the call between fraud and not fraud cannot be done by a human, so if the analytical software works correctly – it saves big bucks.

It’s trickier when there is more time, more latitude and the user may be replaced by the model. Take recruiters who pride themselves to find the best candidates out of a sea of potential recruits like needles in the haystack. So for a recruiter to trust a model to the point of e.g. Google Now moving your wakeup call up – which would be inviting the candidates for calls – takes a big step.

There is value in other stuff

Analytics is not everything. Visualizations, Modelling, interactive data exploration – all have their place and benefits in the world of business. But let’s call them on what they are and not confuse them with analytics.

How to find out it’s not ‘true’ analytics

Here are some easy guidelines even for a non-technical business user to see what not analytics is:
  • If you can see colors and a chart – it’s not analytics. (Think your car asking you – do you want a bar chart of a pie chart on the reasons why the ABS should kick in).
  • If you hear talk of ‘actionable insights’ – it’s likely not analytics. Because if the vendor was sure what the right action would be – the software should take them – and then we would have ‘true’ analytics.
  • If you hear ‘what if analysis’ – it’s likely not analytics. There is a place for ‘what if analyses – but it does not take an action.
  • If you hear talk about ‘insights’ – it’s likely not analytics. Yes insights are important – but insight without action is useless. When a system helps business users to find insights – its fine – but not analytics. 


Do you agree? Time to call out the ‘false’ or ‘faux’ analytics.
2012, 2013 & 2014 (C) Holger Mueller - All Rights Reserved