Why decision automation is key in big data analysis

Uwe Weiss (Blue Yonder)
Business & Industry
Location: 120-121
Average rating: ****.
(4.00, 8 ratings)

Conventional wisdom maps a progression from descriptive analytics (what has happened), via predictive analytics (what will happen) to prescriptive analytics (what should we do). Ostensibly this development is mirroring the increasing creation and availability of data, through the waves of computerized and networked business computing, following the money (CRM and ERP), to business software dealing with behavioral data (digital marketing software and HR software) to the next wave of autonomous devices, connected sensors and the internet of things. Despite the amount of data that needs to be processed growing by multiple orders of magnitude from wave to wave, the success of approaches in democratizing data access and visualization has been timid. If our strategy in dealing with big data is concluded in giving everyone access to visualizations and simplifying visualizations so much that they can be consumed by everybody, we are going to fail.

We present an example of european companies that are leaders in their industry that realize this problem and unlock the potential of big data in an entirely different way, by automating some of their most critical, most decision-heavy and most impactful business processes like pricing, replenishment or staffing based on data that is being extracted, processed, turned into predictions, which are turned into decisions that are put into action in minimal roundtrip time. We call these businesses predictive enterprises. Is there an analyst in this process? Only to provide oversight and make sure the machine is running smoothly? Are there data scientists in this process? Plenty. Understanding the core business, modeling the data and the decisions to prepare is one of the most challenging tasks for data scientists in the world. With each small decision being automated, an aggregate of more intelligent, data-driven decisions leads to tremendous impact on the operational efficiency of the business.

For companies bent on following this example to become predictive enterprises, a set of tough challenges needs to be addressed:
- Cultural challenges: How can we trust our core business to a black-box algorithm? What if something goes wrong? How do you reflect the implicit (or biases) of our team in the model?
- Technical challenges: How to get access to the data in the first place – and how to execute decisions? How can we turn data scientists from system explorers into system builders, crafting fine-tuned automation processes?
- Scientific challenges: How can we communicate decision quality objectively and understandably? How can we keep models learning, accurate and flexible at the same time?

This presentation is based on our experience in running a data science practice and operating customer’s predictive applications on our predictive application platform as a service. In the final section of the presentation, we will show what a predictive application platform as a service looks like and how applications are built and operated.

Photo of Uwe Weiss

Uwe Weiss

Blue Yonder

Uwe Weiss is Blue Yonder’s visionary leader. His objective is to bring together world-class data science and world-class standard enterprise software. He is a passionate software entrepreneur with a noteable and successful track record.