Building a Data Culture

Its a hot topic with posts from Harvard Business Review, Forbes, CIO, McKinsey and others.

There tends to be a focus on just the people side of creating a data culture, without recognizing that the end goal is to make better decisions faster. For that you must consider the whole. If the other components are not in place then the enterprises efforts to create a data culture will not succeed.

What happens when one of the ingredients is missing?

Knowledge Constrained

If the effort is knowledge constrained then there are likely to be a number of symptoms.

• Wont know what questions to ask or how to ask them

• Analytics will be in a small team of experts (who might have some great questions, but wont have the front line experience to solve many of the problems)

• Insights and improvements will be capacity constrained by the size of the expert team, if one exists.

Process Constrained

The likely symptoms when it is process constrained are:

• Inconsistency: done differently every time

• No Standards: root cause of inconsistency

• Continued reinvention of wheel

• Limited knowledge sharing, maybe some socialised knowledge sharing, but no formal methods..

• Need methods and guidance to help people get value from the data quicker

Technology Constrained

This is where many assume they are. All you need is a magic bullet and all your problems will be solved. So what are the symptoms?

• Don’t have the best technology

• Results in additional work to make use of the technology you have

• Additional cost to configure the technology to get what you want

• Use additional technologies and process steps

• May lock you in to current technology for longer due to customisations

• Not easy enough to deliver the results you want

Data Constrained

Symptoms of a data constrained effort are:

• No data, bad data, or wrong data.

• Significant data cleaning and adaption prior to use.

• Slows analytic efforts. Increases cost.

• More likely to fail.

• More likely to have misleading or incorrect results.

• Garbage In = Garbage Out

So each component needs to be in place. They don't have to be perfect, but do need to reasonably matched so that you get the results that you need.

The maturity of each of the components will be a topic for another day.