Q&A: Machine Learning, Predictive Analytics, and the Modern Enterprise
Are you interested in bringing machine learning and predictive analytics to your enterprise? Many prospective users don't know where to start.
- By James E. Powell
- October 18, 2017
Are you interested in bringing machine learning and predictive analytics to your enterprise? Many prospective users don't know where to start. Ramon Chen, chief product officer for Reltio, recently spoke with us about this challenge.
Upside: What's the importance of machine learning and predictive analytics for the average enterprise?
Ramon Chen: Predictive analytics helps enterprises plan ahead and make better, more data-driven business decisions based on insights that forecast future outcomes. Predictions require historical and transactional data to identify risks and opportunities. Machine learning (ML) algorithms can automate many of the interpretation and evaluation processes that would otherwise need to be done manually, thereby improving efficiency and allowing you to execute repeatedly at scale.
What's driving interest in these technologies from both IT professionals and business/end users?
These technologies can improve the efficiency of otherwise manual tasks involving the interpretation of data, and the resulting actions that should be executed based on the results of the analysis. For IT, such analytics can support data management initiatives that focus and prioritize data quality efforts on data that is most valuable to the company. For business users, recommended actions derived from predictive analytics and machine learning can guide and inform users, allowing them to improve their outcomes.
What struggles or challenges do organizations face when trying to get business value from advanced analytics initiatives?
Organizations typically find it tough to match and correlate data from different sources -- and of different types -- to form the foundation upon which machine learning can effectively operate. Traditionally, master data management (MDM) has been used to solve this problem for master profile data. However, combining the transaction information that highlights the behavior of each master profile, such as a customer or organization, has never been solved through MDM.
How does data quality play a role in gaining the greatest benefits from these technologies? What else may enterprises need for successful implementations?
A reliable data foundation through MDM that continuously manages and ensures data quality is essential to such initiatives. No ML algorithm in the world can be expected to take poor quality data and create accurate, meaningful insight. Garbage in will result in garbage out. However, MDM is insufficient on its own. Interaction and transaction data must also be correlated to get a complete 360-degree view of the behaviors and patterns that support ML initiatives.
Why is it important for enterprises to create a closed loop between analytics and master data?
A closed loop supports continuous measurement of outcomes from generated insights and recommended actions, leading to quantifiable ROI. Outcomes, good or bad, are required for ML algorithms to continuously evolve and improve future results.
What machine learning and predictive analytics innovations is Reltio providing to enterprises and how are they different?
Reltio is a modern data management platform-as-a-service (PaaS) that has MDM built into its foundation. Reltio delivers reliable data through cloud MDM and brings together interactions and transactions reconciled with those profiles for analysis.
Rather than just sending mastered profiles and dimensions to downstream data lakes or data warehouses, Reltio allows all of the data to be analyzed in place through Apache Spark. This eliminates any need to keep data models synchronized across environments. Reltio also provides aggregate attributes to be written back to master profiles, enriching them with additional segmentation criteria that can be delivered to end business users in the context of their goals via data-driven applications. The use of Apache Spark allows in-memory, flexible, and cost-efficient at-scale processing on demand.
Why is it important for machine learning and predictive analytics to be embedded into today's data management strategies and implementations?
With data volumes continuing to increase and the velocity of change of data, decisions have to be made in an automated, data-driven fashion for enterprises to remain competitive. ML can predict and recommend actions, as well as form the backbone of improved customer experience through personalization.
Where are these technologies headed? Will ML/predictive analytics be "must have" technologies in a year? Five years? Are we there now?
Companies are in the early stages of leveraging ML and predictive analytics. There are no shortages of vendors providing such capabilities, and many are confused by the landscape. For many companies, their data is of too poor quality to even contemplate using ML to generate business outcomes. They first need to ensure that they can form a reliable data foundation before attempting to generate insights and recommended actions that they might bet their business on.
Of course pioneers such as Amazon, Facebook, LinkedIn, and others are leading the way. They can do so because they've already put in place the data management backbone that allows them to take the next steps to get the most out of ML.
For most, the journey has only just begun, and they would be wise to get their data in order before jumping into the fray.
James E. Powell is the editorial director of the Business Intelligence Journal and BI This Week newsletter.