Predictive Analytics

One of the key elements of the human nature is the anxiety of the future.

We strive majestically to anticipate the future instead of living and enjoying the moment that a huge industry has been existing right about since the beginning of time.
Thanks to our ancient Amygdalas (part of the brain) we are able to suffer anticipating what’s going to happen next, right from the moment we have acquired this mechanism.

Our predicting mechanisms (as in “will this tiger attack me or hug me?”) have carried the previous generations out of the desperations for survival into the are of the knowledge.

“Seeing” into the future and anticipating the development/flow/growth is the capability that a great number of animals possess.

Besides the very basic capabilities that we share with other animals, humanity started the business with the Oracles/Seers (which must have been existing since the very first humans) and which are still humongously popular between certain layers of earth population and incredibly close to those who are desperate.

Since ancient times, Humanity have been practicing Polytheisms in the religious life, with some specific gods/saints responsible for assuring some certain future(s).

Betting/Gaming (and for the modern times we have Online Parts), where one can try to predict some certain sports/political/cultural/any results, betting their own debt (aka money) on it.

The huge drive to be the early adapter of the technology (this is where I am definitely guilty) is also a kind of an anxiety to learn right away, to discover and then to be able to relax, be the first, to lead, the list definitely goes on.

In the digital era the selling point of the software right now is something that is called Predictive Analytics.
Most people won’t even care much on what it is – Data Mining, Machine Learning, Artificial Intelligence, Deep Learning – you name it, we’ve got it!
Important Disclaimer: I am not thinking/saying that all of those technologies are the same, I simply imply that the “marketing speak” has popularised them beyond any reasonable doubt and that in the terms of the Hype Cycle, it feels to me that we are right at the peak of the inflated expectations.

Artificial Intelligence is scary, but if you start thinking about the Artificial Stupidity – that’s where the things will get miraculously dangerous. :)
I wish so hard that more people would understand the basics of the statistics and the fact that a big analytical picture can’t be applied on small numbers or personalised without a huge effort on customization.

Predictive Analytics

The only constant in the nature is the change.
Assuming that all the same factors are equal or even similar is not very far-sighted in the modern ever-changing world where the pace of innovation and change has been increasing exponentially.

What are we predicting ?
Is this the metric of the future or is it the metric of the past ?
Do we want to predict the metrics of the past (to be comfortable) or are we predicting the prediction ?

I have questions that I want the clients to answer before adventuring into the predictive analytics:
How do we “predict what’s need to be predicted”? :)
What shall we do if while implementing the project, the metrics that we choose to predict shall become obsolete ?
How much time will it take to implement the project ? We all know how good are IT people at predicting the implementation time … Ha-ha-ha!
How and WHEN do we measure and evaluate our prediction ? At which precise point we decide about the moment of the decision and when is the right moment to do so ?
How much data should change in order for us to review and rethink our model and metrics?
This is ain’t far a complete list, but this a rather small and kind of important start.

Some metrics are declared to be pretty much constant and clear – such as cash flow (well, with market value variation they actually less than clear), the number of clients (with merges and bankruptcies one needs to predict the future existence of the client) and so on. Those are pretty much determined (or are they?) and can be discovered through learning or acquiring the consultants.

The spectacular angle is that the really important things, the so-called soft skills, relationships & culture are so much uncovered, and by the way is there some magic formula that can help one organisation to understand ? :)
If there is one – please share it with all the huge corporations of the past, with vast amounts of magically smart people who should have known or would love to learn now.

Some enviously intelligent people that I do admire have been saying for years that “future prediction” is actually a process of better discovering and understanding the past and not the vain attempts of uncovering the real future. I tend to align with this point of view.

A great colleague of mine reacts in the following way to many future snake-oil-kind-of-prove promising technologies – “Enhanced with AI ? With Python and API as a Service?” :)

Final Thoughts

At some certain point (and everyone should decide for themselves when) we need to stop predicting and start limiting and deciding. The model we are building will never be perfect and working more a lot of times will mean overfitting instead of improving.

I, for myself, predict that the future is unpredictable, but I am also predicting that I will be predicting what is predictable.

But very carefully.

Leave a Reply

Your email address will not be published. Required fields are marked *