This article was originally published at http://www.cmswire.com/digital-experience/putting-machine-learning-to-work-on-customer-data/ by Ajay Khanna
Putting Machine Learning to Work on Customer Data
Businesses large and small are being lured in by the potential of artificial intelligence (AI), machine learning (ML), deep learning and cognitive computing, while others are still trying to figure out how to tell them apart.
It's easy to fall under the spell of technology that promises to solve all of your business challenges, just for the asking.
Ask any business user and their objective seems simple: offer the right product, to the right person, at the right time, via the channel of their choice.
But delivering on this magical formula is hard if customer information is scattered across systems and channels. Companies collect large volumes of data, but often have no single source of truth across functional groups, like sales, marketing and support. This results in inaccurate or unactionable customer insights and disconnected customer experiences.
Companies are starting to realize that machine learning not only doesn't offer a silver bullet, but we still have much to learn about how to use these technologies to understand our customers.
The key to using ML and advanced analytics is having a focused set of benefits for each user’s role that’s accurately measured, to avoid being labeled just another data science experiment.
Use Machine Learning to Improve Data Quality
Most companies aren’t ready for any form of AI, deep learning or cognitive computing because their data is in such poor shape. However, businesses can deploy machine learning to help with data consistency and accuracy to improve data quality, uncover patterns, detect anomalies and assist people such as data stewards be more successful and productive.
If your organization is trying to achieve a 360 degree view of your customers, products or suppliers, you would have to bring together data from all internal, external and third-party sources. Blending requires careful matching and merging of the data. Defining matching rule sets is a challenge because it takes time and a deep understanding of data profiles.
As the number of sources increases, and the format and data types grow, defining rules grows in complexity, as simple rules-based matching may not be sufficient. Data matching accuracy is also often questionable. It’s not comprehensive and requires backup processes and manual interventions.
The intricacy of data matching adds to the time to value for most master data management tools. Organizations can spend months iteratively defining the matching, testing and fine-tuning of match rule sets.
Utilizing machine learning within a data management platform can help generate match rules automatically from data, and provide active learning training for data stewards. Data stewards can take a set or a sample of data and run it through the matching rule sets. They can then evaluate the data matching quality to indicate to the system which matches were good, and which were inadequate or inaccurate.
Those evaluations instruct the ML system on how to treat the data and determine new match rules. Thus, the system adapts to the customer data and user behavior. ML can provide recommendations that improve data quality by suggesting better matching rules, finding potential matches as new data sources are onboarded and determining profiles with poor data quality and wrong addresses.
Machine Learning Uncovers and Understands Relationships
The next important step is revealing relationships between data entities. Graph technology helps us understand relationships. A commercial graph similar to LinkedIn or Facebook can relate customer profiles with products, accounts, family members and locations. You can establish many-to-many relationships between these data entities to understand where customers shop, what their preferred products are and who influences their decisions.
A data management platform can then provide intelligent recommendations based on data and behavior. When applied to a graph that connects people to people, products, stores or locations, you can learn a lot about your customers. You can use relationships for identity resolution, and page rank and node connectivity to understand the influence a customer has within a network, or find out which influencer affects the customer. You can group people into households and market to them based on the needs of the household, rather than marketing to individual customer records.
Deliver Relevant and Timely Offers
Once there’s reliable customer data in place and an understanding of the relationships, machine learning can help recommend the next best offer to send to a customer, at the right time, using their preferred channel. It can also help identify the key influencers to contact within an account, and which product(s) to offer.
A unified platform that brings together master data, interaction data and advanced analytics gives users access to various algorithms, like single value decomposition, clustering or collaborative filtering to learn more and improve the relevance of the offers you send your customers.
Modern Data Management systems should have the machine learning built inside the platform. Bolting AI or ML into legacy data warehouses or siloed systems, or using legacy MDM tools to feed a ML tool downstream will never deliver the desired outcomes.
Free Data Scientists From the Busywork
By some estimates, data scientists spend more than 80 percent of their time in data cleanup and extract, transform and load activities, which is a poor use of their time and skills. Make clean and reliable data readily available to data science teams.
Combining reliable data, relevant insights and intelligent recommendations into one, single platform helps deliver deeper understanding into customer behavior and needs. Successful execution requires a closed-loop of all data, insights and actions, to ensure accurate measurement so you can continuously improve your outcomes.
About the Author
Ajay Khanna is the vice president, Product Marketing at Reltio, the creator of data-driven applications. Prior to joining Reltio he held senior positions at Veeva Systems, Oracle and other software companies including KANA, Progress and Amdocs.