Artificial intelligence (AI) is a hot topic and many marketers have high expectations for the technology. There are lofty predictions for AI in the future, however, there are already a lot of applications for the technology within marketing automation that retailers have yet to take advantage of. What does it look like in practice?
First a word on data. Machine learning (ML) algorithms are smart enough to identify patterns and important variables on which a visitor gets the right (contextual) recommendation. You probably already have lots of it going unused.
With first-party data, you could think of:
Product recommendations aren’t new — Amazon.com has been doing them for over a decade. And before Amazon popularized the tactic, salespeople were recommending products to customers one-to-one for centuries. What’s exciting is AI and ML are taking product recommendations to a new level.
Before recent technology advances, we thought of recommendations in terms of personalized merchandising on the page level. While this method of recommending products has great potential, there are plenty of improvements that can be made. Take this example from Disney’s online store. A shopper who is looking at a sleeping mask triggers the “mask” tag and without knowing more context, is presented with inaccurate product recommendations.
Availability of the right training data is therefore critical. Brands can make a combination with another type of AI that’s able to automatically tag images and text because the AI understands the meaning and can group products that go well together (for alternatives, cross-sell and upsell).
In physical locations, the endless isles (i.e., screens with complete product assortment), POS checkout and tablets held by store associates help bring smarts to the shopping floor — like what you would see at a Tiffany’s. Once people have identified themselves with a check-in or loyalty card, for instance, the recommendations can even be more specific.
In-store behavioral data analysis is where possible optimizations can be guided by AI, while the divide between online and offline narrows. For example, buy online, pick up in-store (BOPIS) customers could be targeted with relevant product cross-sells once in-store.
An example of how to improve the performance of existing email marketing campaigns comes from AI technology vendor Amplero with PetSmart. The below example shows a message designed for driving more sales and net margin dollars vs. current campaigns.
We know that automated pricing optimization is another form of AI, but in this case PetSmart has personalized the subject lines, sub-headers, messages, products and offers based on micro-segmentation and automated optimization with multivariate testing.
This is where AI and marketing work together. AI can identify thousands of new customer micro-segments (e.g., high spenders, people about to replenish, people about to churn, etc.) across all products, types of pets and life stages, while marketing contributes creative expertise. Variants are tested on the autogenerated segments to see which perform best.
For PetSmart, this reportedly resulted in an increase in email open rates, sales and more purchases of featured products. Key performance indicators are crucial when using AI because the algorithms need to know what to optimize for.
Marketing automation today goes way beyond if-this-then-that logic. AI can match faster and better than the smartest marketer ever could on gut feeling.
Zalando, for example, has an “Algorithmic Fashion Companion.” This is a digital outfit combination tool, based on ML and AI, which provides customers with (almost) unlimited outfit suggestions.
The tool makes it possible to automatically generate a “complete the look” feature on product pages.
Predictive analytics tools use data to forecast future trends, but also on a one-to-one scale to improve customer service and customer experience.
Thanks to AI’s learning capacity, predictive analytics can effectively model and analyze customers’ purchase behavior much better than before. AI can add incremental value over other analytical techniques — 87 percent of cases in retail, according to McKinsey.
Marketers can use predictive analytics to “reverse engineer” customer segments, actions, offers, content and touchpoints to determine which marketing strategies had the most positive results. The outcome works by ingesting all types of marketing segmentation data.
Companies like FedEx identify “Flight Risk Factors” and are able to predict which of its customers will defect to a competitor with 60 percent to 90 percent accuracy. Telecom provider Sprint is able to identify a segment of customers that are 10 times more likely to cancel compared to other customers, then automate the campaigns that follow to reduce risk and improve retention.
Predictive analytics can also help brands anticipate volume of customer service calls or staff needed in-store. Measured staffing saves companies money by not overstaffing, while having enough people on hand to ensure a good customer experience during peak times.
We’ve seen that the applications of AI in marketing automation for retail are already there. McKinsey suggests that the majority of applications are evolutionary. AI allows marketers to do what they already do — only better and faster.
Jordie van Rijn is an independent email and eCRM marketing consultant.