Photo by / Marketing Analytics Summit Website

2020 Highlights from Kishen Jayanti, Tim Wilson, and Kent Lewis

Why Building End-to-End AI is More Important than Models — Kishen Jayanti

According to Kishen Jayanti, a Data Science Lead at Anthem, AI incorporates machine learning combined with data connections and real decisions. He explains that the life cycle of AI products consist of four stages — Ideation, ML Build, Engineering, and Learning. Jayanti tells us to constantly be asking ourselves, “Is it worth the effort?” and “What is the expected ROI?”

Think about the larger impact of your solution. You should always be testing and experimenting. Take your tests and learn from them. Additionally, when it comes to machine learning builds, scalability is your best friend. Jayanti suggests you consider if your solution can be used to solve multiple problems. “Start Small and Scale Big”

Jayanti offers great insight into how we can use the solutions from one problem to solve the next one. He tells us not to look for answers, but solutions in which you can scale, to focus on the products you are creating, opposed to the pinpoint focus of a single project. Finally, he emphasizes the importance of starting small and scaling big with your findings. Reduce Cognitive Load in Charts and Tables

Tim Wilson’s presentation, Data Visualization and Neuroscience: Why it matters to the Analyst offers great insight into how we can reduce cognitive load through the visuals used in presentations.

My favorite part of Wilson’s presentation was the notion that “pie charts are evil.” Wilson’s comical way of explaining that when viewing a pie chart, our viewer will spend unnecessary amounts of time trying to identify what information they should be pulling from the data, as opposed to understanding its significance. This is due to the amount of cognitive load pie charts, and similarly, doughnut charts, placed on the viewer. It is difficult to mentally process the difference in arcs and slices while simultaneously reading a legend. The viewer’s eyes will bounce between the two, and the result is a lack of comprehension.

Data visualization is huge, but how you choose to showcase your findings can be just as important. Strategic data visualization techniques can effectively reduce the amount of brainpower used in order to understand your findings. Wilson suggests incorporating color for impact. For instance, when focusing on bar graphs, use gray for each category and a single accent color for the one you drawing attention to. For highlighting key findings within an extensive numerical table, use heat maps. This tactic can draw the eye to the most significant result.

Put bar charts side by side when comparing metrics.

Sparklines can show text size variation.

And if you want to show change, use a slope graph.

For additional insight into the topic, Wilson suggests reading Story Telling with Data: A Data Visualization Guide for Business Professionals by Cole Nussbaumer Knaflic.

Building a Winning Omni-Channel Attribution Program — Kent Lewis

In Kent Lewis’ presentation, we were offered a consumer-centered view of marketing. As a result of the pandemic, Kent Lewis tells us about the rising importance of chat-bots and an increase in virtual and online channels. Current brands who are employing successful omni-channels include REI, Bank of America, Starbucks, even Disney as demonstrated through the company’s use of RFID bracelets in their parks.

Lewis tells us that one of the most important aspects of a winning omni-channel attribution program is the company’s ability to reduce friction throughout the entire customer journey. Think of the different channels you can use to provide the best overall experience. Channels could include email, social media, even corporate websites. “Understand your customer.”

Focus on gaining insights into major milestones and constantly be looking for new opportunities to make the customer journey seamless. Integration between channels is key.

Lewis offers five steps for building a successful program:

Measure. Always track your data, and use it to identify new trends. Lewis suggests creating a repository for all data. You can’t make data-based decisions without measuring your data.

Segments. Lewis tells us that we should always consider the context behind each touch-point, and further, we should be creating customized content for each segment.

Create. You need to consistently deliver compelling content. The trick for this step is empathy. What are your customer’s needs, and how can you meet them?

Listen. Learn from your data, and be willing to adapt. Your findings can tell you which touch-points are working — and which are not. Once identified, you should focus your time on the touch-points that do work. For the best results, Lewis says to monitor feedback across all channels and segments.

Expand. Don’t limit yourself, always be on the lookout for new ways to collaborate outside of your team. A fresh perspective can make all the difference, so try to break the mold and think outside of the “marketing box.”

This article was originally published on Medium