Adobe this week introduced three features to its Target application in an effort to improve customer experiences through increased personalization:
- New Personalization Insights reports;
- Propensity score model comparisons; and
- Real-time customization models.
These new features give users enhanced transparency and customization of Adobe Sensei-powered artificial intelligence for personalization in Target.
“Although not all firms will have the data science expertise to manage their own algorithms, these additions by Adobe give customers more visibility and confidence into how AI-based results are delivered,” noted Rebecca Wettemann, vice president of research at Nucleus Research.
“There’s been a big push for embedding AI in apps such as marketing, but also concerns about user confidence with black box recommendations,” she told CRM Buyer.
Personalization Insights will launch in beta this spring. The propensity score model comparison tool is scheduled for release in June. Real-time customization of customer-owned models in Adobe Recommendations is generally available as part of the Target Standard/Premium 18.3.1 product Adobe released last week.
“There are a lot of niche vendors providing best-in-class marketing personalization, but no one provides this across all elements of retail personalization — price, product, promotion, content and buyer journey,” observed Michael Phelan, principal at Go-to-Market Pros.
Getting Close and Personal
Personalization is critical, suggests Segment’s 2017 State of Personalization Report. Among its findings:
- 70 percent of 1,006 adults surveyed in the United States were frustrated by impersonal shopping experiences;
- 32 percent of shoppers expected to receive a personalized discount one hour after identifying themselves;
- 54 percent expected to receive a personalized discount within 24 hours of identifying themselves; and
- 22 percent of consumers found their retail shopping experience to be highly personalized.
Adobe’s new Personalization Insights reports show the basis for the models Adobe Target builds — that is, the visitor attributes considered most influential in the model built by Adobe Target; how it grouped customers together into the audience segments used; and how it decided what offers would resonate most with those customers.
The goal of this approach is to let marketers create experiences specifically for any group of visitors.
“Providing marketers with greater transparency empowers them to benchmark their results with others, and find patterns on how to improve engagement with customers in various segments,” noted Cindy Zhou, principal analyst at Constellation Research.
“Marketers want transparency,” Go-to-Market Pros’ Phelan told CRM Buyer. “Smart companies like Adobe try to demystify AI and not hide it in a black box. I support that.”
Reading the Customer’s Mind
Propensity score model comparisons let marketers, developers, product owners and data scientists bring their own models with custom propensity scores into Adobe Target.
This feature lets Target compare multiple propensity scores for visitors on the fly; place those visitors into an audience based on the highest propensity score; and deliver the most relevant experience.
It would be a useful tool for companies of all sizes, because “advanced marketing personalization is really 1:1 marketing at scale,” Phelan pointed out. “Propensity scores are tangible ways to evaluate next steps to convert and what’s needed to convert.”
Organizations “have a treasure trove of first-party data on their customers, and the propensity score comparison solution lets them add their own data models to provide better next-best offers,” Constellation’s Zhou told CRM Buyer.
Customizing the Model
Real-time customization to customer-owned models can be accessed through the Custom criteria option in theRecommendations activities of Adobe Target Premium.
This feature lets brands pass the results of their own models to Adobe Target for use in the Recommendations algorithms. Users can execute real-time filtering rules, weightings and customizations on top of custom criteria. For example, if something changes at runtime — say, an item is out of stock — users can account for it in the recommendations Target delivers.
“Your actual real time on-site behavior is a much better predictor” of customer behavior,” Phelan pointed out. “A woman who was expecting a year ago and looking at buying an online baby item is looking for something different today.”