What is actionable Business Intelligence — 3 levels of implementations
“Can we/ you do more?”
Hiring a good BI developer can be quite a difficult task. For the most part, one needs two opposite characteristics that are usually hard to find. A candidate needs to have a good level of technical and programing skills, as well as sound business understanding and communication skills. In the past few years BI and Data Analytics roles have come to be very popular as they offer the attractiveness of business and tech. As this is a hard position to be accepted to, the average candidate holds a Master’s degree with programing skills (SQL, Python), statistics and machine learning and a solid business understanding. Once a team of rock stars is in place, there is more than ETL (data integration), data warehouse maintenance and reports creation this kind of team can deliver. This is the place to expand the discussion about Actionable Business Intelligence and what it actually means in practice.
Actionable BI was born from the question “What’s Next?” — asked frequently by enthusiast BI developers and managers that do not limit the BI outcome only to reports. Managers already had enough reports and there was a constant need for improving the speed with which a company makes decisions to create true business impact. These days operational and business users do not have the capacity to look at the dozens of daily dashboards and search for the single point that actually requires a decision or an action. The approach had to be changed. This is where Actionable BI was born. Most of the topics listed below are not new as a concept but a refresher on how the BI team can take an active role in the implementation. Here are three levels of implementations of actionable BI which can make a true business impact:
Level 1 — Alerting and push notification
Alerting system is probably the 4th component any BI team should have after a solid data warehouse, Integration tool and Reporting (dashboards). The alerting function “push notification” can be built in-house, be part of a BI tool (reporting), integration tool (ETL/ELT) or be a stand-alone solution. The approach is to notify an end user (business, analyst or BI) on any event or data item that is considered relevant for them. Asking the end user to connect to more than 2 systems a day will result in low adoption of the BI development effort. Therefore, the suggestion is to integrate the system with already used solutions such as Slack or Email. A good starting point of push notification would be to check what is considered to be a drastic change in performance or behavior (positive or negative) that requires one’s immediate attention or action.
Level 2 — Recommendation engine
One will be surprised, but every company could use a variety of recommendation engines. Some examples of recommendation engines can be related to lead generation and lead qualification, traffic controlling and, obviously, product suggestions to external users, which can be clients or partners. Those recommendation engines, can be established using heuristics or statistical methods.
The heuristic approach (also considered as human learning) can be quickly implemented as they mostly relayed on knowledge of the business and human observation.
The statistical approach, mostly based on Machine Learning, takes longer to implement in order to become effective. The statistical approach, however, gives the option to search beyond a local maximum and constantly improve and adapt over time.
70% of time spent on recommendation systems dedicated to data collection, data cleaning and preparation, results distribution, measure performance and effectiveness. Only 30% is focused on the actual recommendation algorithm improvements. The BI developers have easy access to the 70% and a very good business understanding as well as statistical knowledge to decide on a hybrid approach that will yield the best results in terms of time to market and business impact.
Level 3 — Automation of optimization
The third level and maybe most complex, is the implementation of actionable BI, what I like to refer to as the “BI Self Driving Car.” BI (and Data Engineering) teams have worked hard on integrating external and internal data sources, some of them updated periodically while others are constantly being streamed. They have access to entire data to start with decision-making in an autonomous way. A BI developer with integration skills can easily pull data from any system as well as feed it to other systems to take further actions. Those actions can be related to any operational function a human could make. The ability of doing it automatically can free the time of an operations team in exploring new ways of optimization and also cover 100% of all operational activity which may be missed otherwise. The BI team with their very strong analytics and performance checking skills, can easily monitor with set of dashboard the accuracy of those automated decision making and their quality of those decisions.
Summary
It is strongly recommended to constantly asking the question “What’s Next?” and use the full potential of those talented intelligence developers and analysts in making a true business impact.
