Data science teams and businesses at large can greatly benefit from the introduction of modern platforms for data science and machine learning. At the same time, the introduction of such a platform represents an investment of a considerable amount of time and money and bears the risk of a lock-in effect. Hence, a diligent evaluation is crucial. However, it is a challenging task. There are many technical dependencies. In addition, many potential benefits can be realized only after a phase of implementation and learning. While a POC style of evaluation can efficiently reduce uncertainties around the technical dependencies, it cannot easily be translated into monetary benefits justifying the investment and the risk. We present an approach based on a breakdown of the data science process that allows a benefit estimation depending on the kind of data science projects an analytics unit performs. Once the process steps are defined, the platform-driven efficiency gains per step can be estimated based on the POC experience and projected to whole project work of the analytics unit.
Dr. Björn Höfer, Manager Advanced Analytics & Data Science, Telefónica Germany
Frederik Ström, Senior Business Developer EMEA, dataiku