APPLYING THE ANALYTICAL FACTORY APPROACH
The analytical factory approach is about standardizing procedures to drastically reduce the development time of analytics, and this is especially important if digitalized processes are making decisions. The main idea is to "force" all analysts to reuse data sets, various templates, and documentation practices in order to save money and time and to bring campaigns to market closer to real time. An example could be a telecom operator that frequently needs to send relevant offers to customers in real time.
If the telecom operator has 400 product combinations, 200 pieces of service advice, and 100 product-specific messages, in addition to ad hoc communications, it becomes a complex task. Also consider that there are multiple customer segments in multiple countries. We will also have to put some business rules on top, so that only the most relevant messages are sent, and never too often. If we don't, we will create what is known as "monkey with a gun"—a dumb marketing and communication system that simply spams the receivers.
Now imagine a traditional analytical department where the development of a model to support a campaign is done from scratch from time to time on local laptops and on data sets developed by individual analysts. It will take weeks or month for a model to be developed, signed off by the business. And as we need to develop at least 700 models and business rules for the system to send the right communication to the right receivers at the right time, the analytical development time will take hundreds of analyst years. Not good news for those in a rapidly changing business, so strong procedures seem to be the answer.
By using the analytical factory approach, all analysts working on different campaigns would simply start out with the same data sets, documentation procedures, and real-time communication engine. Where the communication real-time engine is the server that is responsible for sending out communication to customers the same second a rule is triggered. For example, if a customer changes their address, we would like to inform him or her of the location of the company's nearest physical outlet. This data will typically be on a customer level/subscription level, which also means that we easily can map a target variable on to the data set. Therefore, if we would like to model which customers are likely to buy a certain kind of phone, we would identify those who just bought it versus those who did not chose to buy it, and assign a buying probability to the rest.
The analytical factory approach improves upon the company's efficiency and readiness for change, as it drives down the development time of models from months to days or even hours.
In this chapter, we looked at how the activities of the BA model can be carried out via a business analytics competency center, or BACC.
A BACC is a forum that includes analytical and business competencies as well as IT competencies and works to ensure that the needs of the business drive all technical initiatives, thereby making sure that the business does not get a data warehouse with a life of its own. One of the most important tasks of a BACC is to coordinate information wheels in order to create synergy on the data side, as well as synergies across analysts and IT professionals.
If the objective of a BACC is to achieve a closer integration between the BA function and the company's strategy, we recommended the establishment of the BACC as a formal organizational entity. However, if the objective of a BACC is to optimize the company's performance, we recommended the establishment of the BACC as a virtual organizational unit.