Predictive Analytics is slowly but steadily garnering attention from business stakeholders in India. As a means of adoption clients are experimenting with different approaches ranging from bring in consulting companies like Mckinsey, partnering with analytics software vendors like SAS or engaging with specialized Data Analytics players.
Let’s assume that output of any such engagement would be a decision framework, a.k.a Predictive model in most cases. This framework will have to be incorporated in existing IT stack for regular usage by business teams of client which will require designing ETL framework for extracting fields required by the model and scoring or processing of records by the model hosted as a batch process before finally writing the scores back into the database systems.
Such customizations in IT stack are not always cost effective and may not happen in time bound manner. Additionally one needs to have a team in place that is responsible for regular performance monitoring of decisions system and tuning in case any adjustment is required. In nutshell you are looking at these steps before you could actually start using your model on regular basis
•Changes in IT systems.
•Presence of Data science teams
•Collaboration between Datascience and IT teams for tuning and performance monitoring.
Keeping in mind above bareers lot many players put analytics at back burner. An alternative framework can be a solution to this problem. Clients can ask their vendor for cloud based offering with subscription based pricing model. This will in effect transfer responsiblity of monitoring,tuning and maintaining infrastructure for decision system back to the vendor. Such a delivery model would also bring down TCO by not requiring in house data science teams.
Over 100,000 people subscribe to our newsletter.
See stories of Analytics and AI in your inbox.
Even clients having well established analytics competency centers can benefit from migrating standard analytics frameworks to such a model and have analytics team focus on newer work areas. Notwithstanding data security challenges such a framework makes a strong case.