“Any sufficiently advanced technology is indistinguishable from magic,” wrote futurist, Alfred C Clarke. And as digital commerce leaders evaluated how AI could impact the role of merchandising at the recent Activate Now, Lucidworks digital commerce conference in NYC, an off-shoot of their annual Activate search and AI event, it became clear that that magic has a new name, hyperpersonalization.

Will Hayes, Lucidworks CEO, introduced the event by saying that expectations of users are changing and retailers must understand better how to convert those expectations into something productive or meaningful.

The million-dollar question? “How do we make the user feel like one in a million rather than one of a million?”

Humans versus data

The first step is to move away from a datacentric approach towards a humancentric one as increasingly users are starting to value the personalized experience. We can take a lesson from the short lifespan of 2017’s Facebook robots, Bob and Alice, which despite being programmed to interact with humans, they managed to communicate only nonsense so Facebook had to cancel them. Data alone might lead retailers to suffer the same fate. The fact that we can anticipate needs, understand intent, and deliver on it is vital, but a nuanced understanding of individual users’ habits could allow smaller ecommerce sites to triumph over the giants.

According to Hayes, companies currently succeeding are big on data but lack soul. Nobody cares about your data, he says, and most of those data collecting initiatives fail to move the needle on how business is done, and indeed can even be detrimental. “We have to shift our mentality, move away from boosting, blocking, ranking,” he said. Siloed experiences of online and instore applications are dangerous; the expectation should be to provide a seamless connected experience across every channel, to move from pushing out product to pulling in people.

Many brands use AI but not to help the customer experience but to be predictive behind the scenes. There is great opportunity to be exploited by extracting from all different silos of information and uniting it for the user. “You may have intelligent analysts doing a great job of poring over the data acquired from a program and they take all their learnings from that,” said Diane Burley, VP Content, Lucidworks. “Unfortunately it’s not a complete picture, even though it is correct. Insights may be both correct and deceptive. Single-sourced data may reflect the perfect sunglasses but only for that one moment in time, that does not mean they will become the sunglasses of the summer.”

Amazon’s absence of taste

We are able to go into a store and ask a salesperson, “Do you have anything like this?” The next step in digital will allow the user to upload an image to determine if the retailer has something similar. Or if the user wants a colored blouse the exact match of a skirt, they can upload a swatch of fabric to find the item to complete the look.

We’ve all been there, say shopping for running shoes and all of a sudden a lawnmower pops up. It’s clear the search engine doesn’t understand the nuances of human desire. Artificial is going beyond that to augmented during which the system moves from Search and Browse to Curate. It is essential to have people internally who can craft specific experiences which help brands fulfill the expectations of the individual user on a more intimate level.

When Search and Discovery turns to Search and Destroy

If a person types something into the search bar and it brings up a null, it’s the kiss of death in retail. “But whose fault is it?” challenges Burley. “Is it the search architect? Those implementing product description?” Most brands need to do a more robust job of SEO, but can AI help eliminate the horrible null? Burley says, “There are merchandising tools out there that have machine learning and don’t necessarily have a strong search component, and there are search engines where the search logic is very robust but they don’t necessarily have the AI component. An AI powered search, mixing the two, is the answer.”

“When we receive a pair of jeans from the manufacturer usually no data outside of fabric content comes in, so merchandising teams must flesh that out,” says Liz O’Neill, Sr. Digital Commerce Manager, Lucidworks. But from the architecture side, is it tagged correctly? And do brand aspirations line up with users’ aspirations?

Katharine McKee, Founder, Digital Consultancy, advises, “Be truthful and honest about your product and what the market thinks it is. Be firm in your brand equity but how you wish to view your product can be unhelpful and lead to missed opportunity. For example, if no one calls your lip gloss a lip glaze, it’s pointless to continue using that term in search engine description.”

The element of surprise

Retailers are under two assumptions: the first is that users are browsing and will click on random tabs to look at other things, the other is that they’re specifically searching, typing in clearly the item and size they want. But are they being supported at the next level? Here location-based merchandising will be key, says O’Neill, “Ecommerce doesn’t always have the ability to offer merchandise based on location. I don’t think etailers are taking advantage of that. Word of mouth is still required, a human conversation.”

“If the user has ten different windows open, they’re maybe making some notes, can we trace their journey?” asks O’Neill. “There isn’t a merchandising platform to digest all that behavior, one dashboard of information could open up endless possibility for us.”

Some surprises are unwelcome

Product suggestions that initially satisfy the user who then clicks through only to find the product out of stock because the suggestions although set correctly at one time have been forgotten is inexcusable. The digital shopping environment should be one in which the user feels welcome, recognized. If they search for a white ankle strapped shoe, in size 11, and the result is out of stock, the engine shouldn’t send other white or ankle strapped shoe options only to further disappoint with none of them available in size 11. If the user has given size information once before, all results should be in the correct size. AI capabilities should be completely personalizing individual shopping needs.

Other ways to trip up

Compound noun queries can lead to ambiguous results. If the user types in Burberry bag, do all bags and all Burberry show up or does the system recognize that it’s a compound noun? A good search engine will manage it. Adjective noun overlap such as Red Valentino can add similar confusion. Similarly, Michael Michael Kors.

Often the user doesn't ask the right question so the system needs to be intuitive, and there can be customer vernacular disconnect. Misspellings can become logged but if the machine can predict what the user meant to write, the retailer runs less risk of losing them to the competition.

Fashion editor Jackie Mallon is also an educator and author of Silk for the Feed Dogs, a novel set in the international fashion industry.

Photos courtesy of Amazon