Comments on a Peer Exchange on Shadow BI

Last October 18, I took part in a peer exchange with about 60 analytics professionals to reflect on three questions:

• ·        What are the top three reasons for Shadow BI?
• ·        What are the top three opportunities Shadow BI may bring to the organisation?
• ·        What are the top three solutions for the issues it brings about?

One of the ten peer exchange products

The group process produced some interesting insights as indicated in the previous post.

Some of these remarks triggered me to elaborate a bit more on them.

Some of them download open source data science tools like Weka and KNIME and take it a step further using fancier regression techniques as well as machine learning and deep learning to come up with new insights.

There we have it: the citizen data scientist. A another big promise, launched by Gartner a couple of years ago. The suggestion that anyone can be a data miner is simply pie in the sky. Would you like to be treated by citizen brain surgeon?  I will not dwell on this too much but let me wrap it up with the term “spurious correlations” and a nice pic that says it all from Tyler Vigen’s website 

A funny example of what happens when you mix up correlation with causation

Other, frequently mentioned reasons were the lack of business knowledge, changing requirements from the business  and the inadequate funding clearly indicate a troubled relationship between ICT and the business as the root cause for Shadow BI.

I wrote “Business Analysis for Business Intelligence” exactly for this reason. The people with affinity for and knowledge of both the business and the IT issues in BI are a rare breed.  And even if you find that rare species in your organisation, chances are you’re dealing with an IT profile that has done the BI trick a few times for a specific business function and then becomes a business analyst.  And worse, if this person come from application development, chances are high he or she will use what I call the “waiter’s waterfall method” . The term “waiter” meaning he or she will bring you exactly what you asked for. The term “waterfall” to describe the linear development path and by the time the “analytical product” is delivered, the business is already looking at new issues and complaining about obsolete information . Some participants at the peer exchange claimed that agile BI was the silver bullet but I beg to differ. The optimum solution is “infrastructural agility” which means two approaches. First you need complete insight in the data structure of minimally the business function impacted and preferably on an enterprise level. Only then can you challenge the requirements and indicate opportunities for better decision making by adding other data feeds. In a Big Data scenario you can add open data and other external data sources to that landscape. The second is about analysing the decision making processes your counterpart is involved in. The minimum scope is within his or her domain, the optimum analysis is the interactions of his or her domain with the enterprise domains.

Shadow BI can improve efficiency in decision making provided the data quality is fit for purpose. 

This is absolutely true: data quality in the sense of “fit for purpose” is a more agile approach to data quality than the often used “within specs” approach in data quality. Marketing will use a fuzzier definition of what a customer is but a very strict definition of who he is and where he is. Logistics will not even bother what a customer is as long as the package gets delivered on the right spot and someone signs for the goods reception. This means that enterprise master data strategies should manage the common denominator in data definitions and data quality but leave enough room for specific use of subsets with specific business and data quality rules.

This under-the-radar form of BI can also foster innovation as users are unrestrained in discovering new patterns, relationships and generate challenging insights.

Just as in any innovation process, not all shadow BI products may be valuable but the opportunity cost of a rigid, centralized BI infrastructure and process may be an order of magnitude greater than the cost of erroneous decision support material. On one condition:  if the innovation process is supported by A/B testing or iterative roll out of the newly inspired decision making support. I often use the metaphor of the boat and the rocket: if the boat leaks, we can still patch it and use a pump to keep the boat afloat but two rubber O-rings caused the death of the Challenger crew in 1986.

"Bet your company" decisions are better not based on shadow BI. 

The group came up with both technical and predominantly organizational and HRM solutions.

This proves for the nth time that Business Intelligence projects and processes are of a mixed nature between technical and psychological factors. It is no coincidence that I use concepts from Tversky and Kahneman and other psychologists who studied decision making in the business analysis process.

In conclusion

Strategy alignment and adopting operational systems and processes for analytical purpose were also mentioned in the peer exchange.  Exactly these two are the root causes of poor decision making support if poorly managed.

In the next post I will dig a bit deeper into these two major aspects. In the mean time, have a look at this sponsored message:

The full story on strategy alignment and tuning organisations for better analytics is within reach