by Thor Olavsrud

How Adobe uses machine learning to drive marketing success

Aug 17, 2016
AnalyticsBig DataMarketing

Adobe's new data science tool, Segment Comparison for Analysis Workspace, leverages machine learning to help marketers uncover the key characteristics of the audience segments driving KPIs.

marketing IT bullseye
Credit: Thinkstock

Earlier this year, Adobe took the wraps off its new Adobe Marketing Cloud, touting new data science capabilities like Adobe Analytics’ Segment IQ, which uses machine learning to help marketers gain deep insight into audience segments. On Wednesday, Adobe advanced Segment IQ another step with the release of Segment Comparison for Analysis Workspace.

Segment Comparison for Analysis Workspace is the first in what Adobe promises will be a series of audience analysis and discovery tools within Segment IQ. It uses machine learning techniques to perform automated analysis on every metric and dimension to which you have access. Nate Smith, senior product marketing manager, Adobe Analytics, says this allows Segment Comparison to uncover the key characteristics of the audience segments that are driving your company’s KPIs.

segment comparison

Adobe’s Segment Comparison for Analysis Workspace 

“Segmenting is a core strategy that is crucial to any marketer’s success,” Trevor Paulsen, product manager at Adobe, wrote in a blog post Wednesday. “As not all customers have the same characteristics or behave in the same manner, it’s increasingly important to employ different marketing tactics for each distinct group. While traditionally segmenting has been thought about quite simply in regards to age, gender and even geography — as our threshold for data-driven marketing continues to increase, the definition of segmenting has shifted as well.”

Traditionally, audience segmentation has been based on broad segments and simple clustering because sifting through mountains of data can be a nearly insurmountable task for humans, especially at the speed with which marketers need to operate.

“Segments often have overlap with each other,” Smith says. “There are non-obvious differences lurking deep within the data. Finding insights is more and more like picking the needle out of the haystack. We, as humans, just can’t process all the data that’s being collected now.”

Segment Comparison can ease that pain, Smith says, helping brands determine which type of customer buys a small TV vs. a big TV, what type of person visits a brand’s Facebook page vs. its Twitter page, or what type of person watches one show vs. another show.

“It goes through and intelligently discovers the differences though an automated machine learning analysis of metrics and dimensions,” Smith says. “It saves marketers and analysts a ridiculous amount of time.”

Paulsen points to Pixar as a perfect example of what can be done with advanced audience segmentation.

“While one might think that their cartoons are just geared to children, in reality their movies appeal to a variety of different groups, including kids, parents, couples, teenagers,” Paulsen wrote. “The messaging and movie promotion go far beyond just simply getting a five-year-old to laugh. Pixar is smart in their approach: aside from traditional advertisements on children’s programming, the movie is also positioned on shows geared towards adults such as Ellen (as they did with Finding Dory), and promoted on social channels with specific storylines that are geared towards a group.”

By identifying the key characteristics of the audience segments that are most significant to a brand, Paulsen says marketers can better understand the behavior that drives positive interaction, sharing and conversions.

Once Segment Comparison is trained on your data — which takes virtually no time at all for existing Adobe customers because it can access historical data — Smith says brands can use it to complete a comprehensive segment analysis within minutes with just a few mouse clicks, comparing every dimension, metric or data point between any two segments and automatically discovering the most significant differences between them.

“This is going to be probably our fastest-adopted feature once it rolls out and hits the market,” Smith says.