Einblicke und News /

11 Oktober 2018

How to sort out Data Analytics and AI: Data Strategy (Planning)

David Blackwell

David Blackwell
Partner | Daten, Analytik und KI | London

So… Planning. In other words, writing down how you are going to deliver the data strategy. 

To explain further… and in a blatant rip-off of a former colleague, Jonathan Calascione, I believe data strategy execution needs to balance “capability” and “credibility”.  Essentially, a data strategy programme needs to earn the right to spend big on capability, by proving that it can deliver value – i.e. be credible.

“Ho hum” you say, and take a bite out of an egg sandwich.  That’s just “strategic initiatives” and “quick wins” by another name.  That may well be true, but there is a subtle difference to how I now see it – which also steals unashamedly from the world of agile delivery.

I have seen a lot of data strategy execution focus solely on capability development.  It’s all about the data warehouse, or the data governance rollout, or hiring loads of data scientists. Big left-to-right, multi month (or year) initiatives that, not unlike a snowball rolling downhill, grow in size, pick up bits of mud and stones and become quite dangerous as they gain momentum. They come to an end by knocking someone over and giving them a bloody nose, or hit a flat spot, come to a halt and then melt. Leaving a dirty puddle of misery. At the end of it all, nothing was really delivered that anyone in the “real world” cared about, and ten million, twenty million… even a hundred zillion has been completely wasted. 

The reaction from most organisations who have suffered the pain of a big, failing data programme is nearly always to resort to delivering “quick wins”, Proof of Concepts (PoCs), hackathons and a whole host of other short term sugar-hits that look great (especially if delivered wearing selvedge jeans) but ultimately leave you with no nutrition or useful calories. Just like a kid chinning a can of Coke, the sugar rush is followed by an equal and opposite collapse in blood sugar – which can only be remedied by guzzling a pack of Haribos… and so it goes on.  PoC after PoC after PoC… nothing ever goes anywhere as no lasting capability is developed by the initiatives. 

How to reconcile… Well… we have created the “saw model”, which I am going to attempt to explain without using a picture and before I get to the end of my 500 words.

In the saw model, the saw-teeth are bursts of activity (sprints) which deliver demonstrable value and thus build credibility – they create the “bite” with the business, they are what cut into the hard log of data strategy execution.  Now, think the body of the saw as your “capability”, which rises from the pointy tip to the handle.  The closer you get to the handle, the more value can be delivered by each of the little teeth, because they have more capability sitting underneath them. 
The key thing however, is that you don’t build credibility and capability independently.  Your “saw teeth” MUST leave behind capability as well as delivering value. 

So that’s it really – when planning the execution of your data – think about cutting through a log. 

Next – something more interesting.  Why software development is the key to getting value from data science.