The Key Message from Gartner’s Data Summit
In Kathy’s blog highlighting key takeaways form the Gartner Data & Analytics Summit in Dallas, Texas which took place at the beginning of March 2017, she started by sharing the insights of Gartner analyst Thomas Oestrich. In his view organisations need to:
- Define their “rules of the game”
- Seek agreement across the organization
- Appoint data & analytics stewards
- Avoid over-governing
There followed a debate where two analysts argued from the polarities of “chaos” and “control”, which ultimately came to the (not unexpected) conclusion that there needs to be a balance.
The Risks of Too Much Freedom
The risks inherent in allowing a free-for-all approach for users to carry out ad hoc analysis can be moderated by setting up a rules library, and educating users about the dangers of raw data extraction. Without reference to common standards and naming conventions the mirage-heaven of direct data access quickly turns into the reality of analytics-hell – maybe even worse than spreadsheet-hell.
Never before has it been easier for individual users to get at their data using free software downloads and transform it into eye-catching charts and reports. Never before have the frustrations that users traditionally experienced waiting weeks for a report to be written by a specialist programmer been so quickly overcome. But never before has there been so much data from so many sources. Never before has there been so many different field names, or the same field names, in different systems, that mean different things.
Users might, for the first time feel like they are unhandcuffed masters of data transformation, with abilities they had never dreamed of, to control, clean and join data sets together. But are they really “jacks” not masters, susceptible to all the pitfalls of data analysis that have been controlled up to now by more experienced technicians?
In many ways these new tools are a good thing – valuable data is no longer just sitting there wasting space on a server – users are able to gain insights that allow them to do their jobs much more efficiently. But, and this is a very big but, in an organisation with even the simplest data sets, there are always rules, and always anomalies, and always the chance to over or under filter, mismatch field names or misunderstand them.
Big opportunities can quickly become big mistakes. Even analytics experts with years of experience (not mentioning any names here but this is our business after all ….) must test every change, every filter, every calculation in the data transformation process and follow those changes through for as many scenarios as there are in the business before they can be sure that the resulting data model is accurate. And then some. Because it’s far too easy to miss something, and then it’s too late. Decisions have been taken on the basis of your information.
That is why the advice set out by Oestrich is so important.
Oestrich’s Advice for Balance
1. Define the “Rules of the Game”
It makes great sense to have a library of business rules and field name definitions so that all BI users can refer to this when transforming raw data and creating their reports. At Dimensional Insight they call this a “Measure Factory” (See more here). It solves all the issues of users making up new rules that bear no relationship to everyone else’s definitions.
2. Seek Agreement Across the Organisation
This one is obvious if you implement the first one. “Rules of the Game” should be agreed across the organisation. There is no point in making up different rules for different departments on the same data set. For example, the summary value definition for what constitutes a sales measure should mean the same wherever that value is used. If a department needs a different definition for some purpose it should have a different name, not the same name with a different meaning understood only by those “in the know”.
3. Appoint Data & Analytics Stewards
In an organisation with a lot of data as well as those with only simple data sets, it is enormously helpful to have data stewards – what used to be called Data Base Managers in the old days. These people were responsible for reviewing and signing off on data changes in the company databases. New tables, new fields and relationships had to be analysed for impact on the rest of the business before these changes were allowed. In the BI sphere, the issue now is that even if there is still a DBM for the back end system, there is often no-one who has been given responsibility for the shared output on spreadsheets, BI data models and reports. And these have quickly come to replace the standard reports coming out of the back end system. This is where mistakes and misunderstandings can replace consistency.
4. Avoid Over-Governing
Once you have the basic protections in place as defined above, here is where the beauty of ad hoc analysis can be left to do its job. In the past everyone saw the same information using the same ERP reports. But nowadays, as long as the grounding is solid and the data is sound, clever analysts and ordinary users alike can use the data models to come up with key insights, looking at the data in ways that no-one had ever thought of. Here is the value in letting people at it – let them find out what they need to know to make them more productive and better at their everyday tasks.
A last recommendation of my own
However simple and easy some of the newer BI tools appear to be, beware of them in the wrong, inexperienced hands – especially if you haven’t implemented the first three recommendations above.
It is so, so easy to transform the data and hide information or mistakes – more so even than with spreadsheets. At least you can see what is behind each cell in a spreadsheet – not the case with a finished data model.
My advice for any organisation is to choose a solid BI platform and make sure that you have the rules defined and in place. Only at the stage where you have common, agreed data models, should you let the users out of their cuffs to carry out their own analysis. Don’t make the mistake of allowing your key business information to be derived from a number of different individual data extracts, using multiple free or desktop-only tools that have limited audit control, logging facilities, batch & production facilities and access control levels. You will never be able to trace anything back or understand what went wrong.
With this in mind and by following the advice from Thomas Oestrich and you will hopefully find the right balance.
About Debbie Lonsdale
Debbie Lonsdale has been working with the Diver Solution as a BI Consultant at Dynamic Business Informatics since 2008. Her previous experience includes computer programming, analytical and technical roles, team management, account management, sales and marketing in a variety of market sectors, including the Travel Industry and Distribution. She combines this experience as an all round ICT professional in the BI sector.