CIOREVIEW >> Customer Relationship Management >>

Real-Time Personalization: Everything All the Time

Mike Berry, Senior Director, CRM Technology, Shutterfly
Mike Berry, Senior Director, CRM Technology, Shutterfly

Mike Berry, Senior Director, CRM Technology, Shutterfly

As the stream of marketing messages in all channels becomes a flood, the messages which are the most arresting, most interesting, most relevant to the recipient are the ones that are going to stand out. The ability to determine what these messages should contain has begun to exceed the capacity of humans to identify. We are now turning towards the algorithms of machine learning to best evaluate what to show to whom. By utilizing this approach we are able to maximize the return on each and every touch of the customer.

1,000,000 offers? Maybe. It depends on how you define an offer. For example, in the world of credit card insert-statement marketing an offer can have many different definitions: How much was the discount? How much is the line of credit? How big was the font of the offer? Did you use red or blue? What shade of red? Once all of these combinations are added up, you could create hundreds or thousands or millions of offers. These many different possible combinations used to be hard to generate and display. In the world of the Internet and Machine learning that is no longer the case.

You must have a platform that allows you to display an offer to a person. If you are lacking this fundamental plumbing then no other step makes sense. Usually this is found within and the capabilities of an ecommerce platform, be it web or mobile. It can span the spectrum of marketing channels or can be as simple as a banner in a purchase confirmation email. But you need to have a platform into which you can place your offer.

 The decision engine also needs access to what the customer is doing right now 

Today it requires a database to keep track of all of your offers. Each offer will have its own unique metadata. Some offers will be live while others have been removed and new ones are being added. You need to store them in a container that can hold millions of them, can be quickly accessed, and is always available. An offer management database is needed.

Marketing Strategy and Machine Learning: Living side-by-side

You’ve got your one million offers running in your marketing publishing channels and your machine learning solution is identifying the ones that are doing the best and predicting which ones that will do even better…so, a marketing technologist might ask, do you still need humans to generate marketing material? ABSOLUTELY! There is no machine out there that is going to tell you what to try next: should it be a black square with golden lining highlighting your biggest seller? Should it be a full multichannel blitz announcing your newest product? What about a new brand voice? How do your corporate initiatives play into this? You are going to need humans to drive the strategy for developing new marketing messages. Once these have been created, run for a long enough period to fully evaluate performance, let the machine learning kick in and make decisions allowing the newcomers to run alongside the tried-and-true.

Decision Engine

The decision engine is the tool that will utilize algorithmic machine learning and human-defined business rules to decide what offer should ultimately be displayed. All in less than a second. Webpages cannot wait on an offer to load. The decision engine has to utilize the rubric of machine learning scoring and layer on top of that any rules-based prioritization to decide what to show immediately. This requires access to the relevant offer management storage solution as well as information about what the customer has done in the past and what they are doing now.

Most companies store a lot of customer data. This can include data like a person’s first name, when they created their account, how much money they have spent with the company, what email they opened last, whether they have downloaded the company’s app. But a lot of this is useless information when trying to decide what to offer that person right now. To that end, the marketing department needs to decide what critical information is needed to decide which offer to display. That information needs to be placed into a small, speedy customer profile database that the decision engine can access.

The decision engine also needs access to what the customer is doing right now. Prior to coming to our site the decision engine might have determined that a person would be interested in getting an offer for a bicycle. However, after the user began browsing basketball memorabilia, the engine should change its offer selection to a Steph Curry jersey for display. Once they buy that jersey, we need to immediately decide what’s the next best thing to offer them on the purchase confirmation page… Basketball shoes?

Offer response data needs to be housed in a reporting environment such that easy-to-understand performance visualizations can be rendered to provide insights into customer behavior to guide future marketing strategy. Frequently referred to as Business Intelligence solutions, there are many out there to choose from to help leadership decide what to keep investing in, what areas to explore, and what simply isn’t working.

Some Assembly Required

All of these elements of real-time personalization must come together in under a second to deliver the right offer to the right person. Integration of your ecommerce platform must allow for the display of the offer. The offer needs to be housed in an offer management solution which is fast and easy to access. Machine learning and marketing strategy need to work together to decide what to use from the existing inventory of offers to anything new. An extremely fast decision engine needs to have access to the most relevant customer behavior data as well as what the customer is doing right now. And finally, we need to have a reporting environment which allows leadership to quickly and easily decide what to pursue and what to avoid. This is not an easy task and should be undertaken with the goals, plans, processes, and people secured well in advance.

Read Also

Machine Intelligence In Insurance: Insights For End-To--nd Enterprise Transformation

Machine Intelligence In Insurance: Insights For End-To--nd Enterprise...

Yannick Even, Director, Head of Digital & Smart Analytics APAC,
Buy, Build Or Both?

Buy, Build Or Both?

Alex carries,Evp chief Digita after at simmons Bank
Digital Transformation: Challenge The  Status Quo, Be Disruptive.

Digital Transformation: Challenge The Status Quo, Be Disruptive.

Michael Shanno, Head of Digital Transformation, Global Quality, Sanofi
Vehicle IoT, the Connected Car

Vehicle IoT, the Connected Car

Stefan Gudmundsson, Chief Innovation Officer, KARMA AUTOMOTIVE LLC
Bolstering the Structural Heart Paradigm

Bolstering the Structural Heart Paradigm

Douglas Gregory, RN, MBA, BSN, NEA-BC, CCRN-K, FACHE Kevin Wall, MBA, BSN, RN, EMT Amar Patel, MD, FACC, FSCAI, Wellstar Health System
Driving Digital Transformation In The Pharma Industry

Driving Digital Transformation In The Pharma Industry

Wayne Hull – Chief Digital & Technology Officer, PCI Pharma Services