A typical question in a world of data-driven business: what do I need to do to start leveraging my data for decision making, for growth, to get ahead of competition?
While each customer’s situation is different, there are some basic truths that needs to be followed i order to set up a stage for the data-driven jump.
Usually what everyone does is they begin to capture and store massive amounts of data, and hope to make sense of it later. As with anything in life, putting a little thought to the process upfront pays off much quicker compared to brute force approach of throwing millions at your data science team to make sense of the data.
Let’s get down to basics.
The data has value if it’s fairly easy to understand and access. Then it can be used for decision making, new product development, analytics, machine learning and other needs.
Data needs to be easily understood
What does it mean to have data that’s easy to understand?
In plain words, it means the data needs to be well structured and described. And in technical terms it means the data model needs to follow industry nomenclature and have good data dictionary.
Well structured means it follows a standardized format of representing information, hopefully following industry nomenclature.
Why is it so important to follow industry nomenclature? Well, if you are hiring someone to work with the data, you want them to be able to be productive as soon as possible and not ask a hundred of question each time he or she is trying to interpret the meaning of each attribute. So this means cost of operating is small and your resources are easier to replace or scale when needed.
Additionally, if your data is being shuffled around in the organization, if it’s used to generate other data or if it’s being used as a reference – again, it’s an easy win and your projects are cheaper to deliver because you don’t rely on a data subject matter any more – the data has been standardized.
Granted, this requires initial investment when loading data into the database for the first time, but it pays off very quickly, the moment you have more than one consumer of the data.
Data needs to be well-described
What does it mean for the data to be well-described?
Well-described means anyone who wants to start using the data should be able to figure out what the whole record means, and what each element (attribute) means – unambiguously.
Not all attributes are very easy to understand and name – so if a person relies on the attribute names alone, very soon the reports and decisions based on the data will lead to wrong conclusions and eventually business losses. So you better read each attribute definition and pick the right one for your decision or analysis. And this implies someone writes that description for each attribute and publishes it for the company use.
Data quality
Next piece of the puzzle – data quality.
Of course if you were to make critical business decisions, the data needs to be trusted – so the data quality is of paramount importance. Otherwise – garbage in, garbage out.
Very often people try to make decisions based on the data hastily arranged from bunch of sources, but if the data is not clean enough – the decisions will now be correct.
Only if the data is cleansed, validated and relevant it becomes a source of your analysis, drives revenues and growth, allows trimming unprofitable parts of business or restructuring them to become profitable again.
Easy access
Sometimes data is collected – but it cannot be easily accessed – which makes it unusable. You have to plan how the data will be accessed, by whom, and when. Who will have access to read the data, and who will be able to modify it – so you need to strike a good balance between access control and ease of access.
Time decay
Data also needs to be fresh for most of the purposes – like anything in this world, things age, and very old data is good for statisticians but with time it becomes less and less relevant to the business.
So these are the basic needs that have to be met if one wants to run a data-driven operation, but there’s of course more to it – yet the above is a good start for a journey to make decisions based on data.
These form only part of the good data governance, which includes many more concerns that support decision making.
Data governance covers processes and systems that define how the data is handled in the organization, and on top of the above it also covers consistency, integrity, accountability etc.