When is milk not milk? This is no trick question — it’s a distinction that artificial intelligence (AI) is going to have to learn to make in order for eMerchants to fully leverage the potential of machine learning.
To one customer, “buy milk” means buy a gallon of whole milk; to another, a 1.4-liter jug of unsweetened vanilla almond milk. Digital shopping lists, apps and virtual assistants must understand this and not force the customer to spell it out each time before these platforms can successfully become the new normal.
“We think about things in shorthand, not in terms of specifics,” Dave Barrowman, Skava VP of Innovation, told PYMNTS’ Karen Webster in a recent webinar. “I don’t want to have to have a long, drawn-out conversation with Alexa to order my milk. I just want her to know.”
While it’s true that eMerchants leverage machine learning to recommend products, target ads and optimize sort order and search results, Barrowman says retailers who only use AI for these basic functions are selling the technology (and themselves) short.
“Making products and ads better is just the tip of the iceberg,” he said.
Visual search engine technology is scary accurate. Barrowman gave the incredible example of a visual search engine pulling up pictures of his eight-year-old daughter based on the input of a recent photo he’d taken of her at a garden in Amsterdam.
The engine turned up a snapshot of his daughter swimming a few months earlier — hair plastered to her head, goggles and a hand obscuring much of her face, yet still the engine picked her out of the crowd. Even more amazing, it found photos that were six or more years old, somehow making the connection (without any tags) between the eight-year-old girl surrounded by flowers and the crying infant born just minutes before the photo was taken.
Machine learning is what makes these connections possible, and retailers could be putting it to use in much less personal scenarios, Barrowman said. One day, it may be possible to use an image from a smart TV to drive a product search so viewers can purchase an item they liked from a commercial or show. Or, as Webster suggested, you could take a picture of someone passing you on the sidewalk and use visual search to find out where they got their handbag and order one just like it.
In fact, Amazon has already released some early technology in that vein, using machine learning and visual search recognition to pull trends from Pinterest, Instagram and fashion influencers’ blogs. In the hands of apparel retailers, said Barrowman, this tool could affect the entire process, from design all the way through retailer purchase and distribution. Which brings us to…
Barrowman says machine learning has potential in the realms of pricing, categorization and strategic store allocation and inventory. In the latter case, machine learning could also pull in data such as weather to help retailers stock the right products in the right places — for example, in February, the swimsuits should be at a Miami store and not at one in Minnesota.
That doesn’t, however, mean dynamic pricing is coming to the regular retail world, said Barrowman. Rather, machine learning could be leveraged to determine the correct starting price for a product as well as appropriate discounts or even personalized pricing offers.
How would a personalized pricing offer work? Imagine a store hangs a 40 percent off sign in its window. For some customers, 20 percent off would have been enough to win the buy; the other 20 percent is just icing. Yet other customers won’t move until the markdown is 50 percent or more.
It would be in the retailer’s best interest to save the 20 percent it didn’t need to offer the first customer and reallocate it to the customer who will only shop if the discount is higher. That way, said Barrowman, retailers can attract shoppers at both ends of the spectrum and enjoy sales from both of those customers instead of just one.
Webster noted that it’s difficult to overcome consumers’ perceptions that everything is on sale all the time, and if it’s not today, it will be later this week. It’s a valid perception that retailers created themselves, she said. In that climate, how much good can a personalized offer really do?
Barrowman agreed that price integrity is hard to build and easy to lose, and the perpetual discount singularity is a difficult one to escape. He believes that the key will be for retailers to better understand their customers on an individual level — again, using data and with machine learning to translate it.
That data could be culled from browsing and social behaviors as well as the retailer’s own historical data on customers and segments. Then, brands can use that data to power ads and experiences that are…
Efficient and Engaging
Creating a grid of products that go together in the retailer’s mind is easy. More challenging, said Barrowman, is having the perfect piece of studio photography and marketing to reach every customer, but he believes the industry is moving in a direction where it could eventually generate bespoke marketing based on micro-segments.
The world of eCommerce promises to make shopping more efficient, but Skava’s Barrowman said it doesn’t always deliver on that promise, and sub-optimal usage of machine learning technologies can contribute to that shortfall. Machine learning could be a critical tool in the effort to improve the customer’s experience and, therefore, drive profits, he said.
Most companies have at least some data on their customers’ shopping history and other contextual information. The more data they have, the better machine learning can help them, Barrowman said. Companies with less data can augment it through third-party partnerships that look at social sources where customers are generating and sharing their own data.
He sees using that data to develop smart, personalized shopping lists as an area of opportunity. Paving the path of least resistance is another. Barrowman explained that online retailers can do this by implementing a user interface that “nudges” customers toward the appropriate next step.
But it’s not always about churning customers through from inception to purchase. Even though efficiency is important, Barrowman said retailers lose something if all they do is move people in and out as quickly as possible. That’s where bespoke marketing could come in (for instance, showing baby imagery to a new parent, but not to a recent college grad who is not yet thinking about starting a family — and that’s all based on contextual data gathered by AI).
It’s also where retailers need to go above and beyond in what Barrowman called “extended commerce.” Virtual assistants like Amazon’s Alexa could be powerful tools for merchants in replenishment use cases, such as buying laundry detergent or dog food, or perhaps to notify customers when something new is available that may interest them. As screens become more important with these smart home devices, there are use cases to be leveraged there as well.
“This is about retailers extending their relationship with the customer via assistants,” Barrowman said. “It’s not explicitly transactional. But the customer is always on a transactional journey, even if she’s not at the point of purchase at this moment.”