Scroll

Insights and news /

28 August 2018 3 min read

How to sort out Data Analytics and AI: More on Data Governance

David Blackwell

David Blackwell
Partner - Data, analytics and AI

“So…that all makes sense and sounds very logical… but what’s the business case for data governance?”
“It’s an enabler.”
“Sounds like you are ducking the question.”
“No, it’s an enabler… like electricity, roads, currency… it makes other things easier.”
“Will it save us money?”
“Absolutely. It’s a case of calculating...”
“How much this year?”
“It’s difficult to quantify exactly, but if you were to have clean data you wouldn’t have spent…”
“Actually, you are boring me.  Please leave…”
“Would you like to keep the pack to read through?”
“No.”

This will sound familiar to anyone who has ever tried to make the case for data governance to a CFO (typically the person who ends up deciding whether or not to do it).  

How do you get round this? 

Sometimes you will get lucky and find an evangelist on the board who will make the case to the rest of the Exec, usually whilst waving a Harvard Business Review or McKinsey Quarterly article about the data-driven enterprise. 

Sometimes you will get lucky and there will be a clear, direct link between data integrity and cash dripping out of the organisation. This happened to me once at a Telco where poor quality data led them to paying their suppliers too much. Tens of £millions too much…

But if you are not willing to rely on luck, then the best approach is to build a portfolio of initiatives that collectively deliver recognisable benefit, and hope to smuggle data governance in through the back door with a bag over its head, even potentially calling it something completely different and more exciting (use “AI” as a prefix, that’s always good). 

If you have been able to secure funding (or are a believer and have given funding), then you need to realise that doing data governance well is really hard. Unless you have an enterprise-wide culture of fastidiousness, an organisation full of nit-pickers who care about accurate documentation, rules, taking minutes and making sure the correct font-size is being used, then data governance will always feel a bit like a wearing a woollen jumper on bare skin. 

It’s hard, because humans aren’t usually great at being neat and tidy.  Humans are like water, we take the path of least resistance. If you try and force water uphill, you’ve got issues.  But it is possible to find a way to allow water to flow almost naturally but in a way that supports your objectives (kind of like a canal I suppose)… so if you can build canals, then you have a chance of making it work.

If you do data governance properly:

  • You have a map of all your organisation’s data entities (e.g. customer data, employee data, sales data), and how / where they support your business 
  • You have someone responsible for each of the data entities, who actually cares
  • You write down everything you need to know about the data (how it is defined, which systems it is in, what rules it needs to follow to be “fit for purpose”)
  • You check it regularly, to make sure it isn’t getting too dirty / unfit for purpose
  • You fix it when it goes wrong.

When you put it like that, it seems amazing that anyone would ever argue about it? 

The challenge is, that if the business isn’t collapsing, it’s difficult for people to understand why they might need a step change in data governance capability. As I said, just drop the words “data is the fuel for AI” into the mix, and hopefully they will get it just as quickly as Amazon’s stock price is rising.
 
People often ask about how to land data ownership successfully.  The challenge is that data can span processes, organisational boundaries, systems, countries and regulatory regimes – so ownership of data never sits neatly into an existing construct.

As such, whoever “owns” the data will have to work in a matrix-like fashion with lots of other parts of the business to get stuff done. If they don’t get that, they won’t be a good data owner. To make ownership work, it’s more important to find someone who cares about data, is influential and is good at working with other people than it is to be slavish to a specific role.  

That’s it.  Next up I shall tackle data strategy (I know, I probably should have done that first.)