It’s all about recommendations
Business departments are interested in concrete recommendations and decisions, and not in complex tables and graphs. This is the lesson, most of data scientists learn after a couple of real industrial projects. Very often, data scientists train models, create predictions, build fancy visualizations, and get at the end a question like: “So what, how should I use it in my daily work?”
Also, in customer/marketing analytics, most of software products apply various advanced algorithms and produce dashboards, which are unfortunately useable only by analytical people with scientifical background.
Our approach is different: clients get concrete recommendations for action. These recommendations simplify their daily job, improve the decision-making process, and save their time. The process of calculating these recommendations is complex and involves data mining, machine learning and statistics. But let it be our job, business users shouldn’t bother about the algorithms. At the end, the results are explicit, understandable, and useable.
Let us show a couple of examples from our customer segmentation service CusaaS. In the background, more than 30 algorithms calculate the KPIs (incl. predictive) per customer like lifecycle segment, average order value (AOV), RFM-segment, ABC-group, customer lifetime value (CLV) and others. However, the result contains recommendations, which can be further executed by the people from marketing and sales departments.
These recommendations are understandable and customizable, e.g.:
If a client is dormant, AOV and CLV ranks are high, lifetime is long and dormant time is short then, the recommendation is “WAKEUP”.
If a client is active, belongs to the ABC-group “A”, AOV and CLV ranks are high, lifetime is long and dormant time is 0 then, the recommendations is “UP-SELL”.
So, dear business people, don’t be afraid of data science, it will make your life easier. And, dear data scientists, don’t forget about the final goal, even if the algorithms are so cool and interesting.