When building a fashion site designed to recommend personalized styles for customers, Armoire knew it had to deliver a customized digital experience, but with a human touch. With the wide range of available machine learning and AI tech, brands can risk coming off as creepy by going too far with the details they use to build customer experiences. Armoire didn’t want to cross this line.
“We want machine learning to be more human, but then we’re freaked out when it is,” said Katrina Taylor, Armoire’s head of user experience and product design, during a presentation on machine learning and human empathy at DX Summit in Chicago this week.
To avoid creeping out customers by crossing unnecessary boundaries when using their data, Taylor said her brand doesn’t use everything it may know about a customer. “It’s better to start off with a light touch and fill in the details as we go,” said Taylor.
She outlined the six rules that Armoire follows to create digital experiences built on machine learning and human empathy.
1. Determine the ongoing value
Taylor believes that the purpose of machine learning-enabled technology should be to advance the human experience, not the other way around. “Just because something is code smart, does not mean it is people smart,” said Taylor, “Great AI and machine learning solves customer problems before solving business problems.”
Once a company can solve their customers’ problems — and create ongoing value — the business problems will likely dissipate. As an example, Armoire knew women spent, on average, 219 hours shopping online per year and that 91% of online apparel returns were due to poor fit. Armoire addressed the issue by showing just 50 items from a customer’s search to reduce the amount of time spent browsing and show products the customer will likely “love” based on an algorithm that surfaces what will fit best.
2. Be sticky enough to build loyalty
Taylor believes the best way to create sticky experiences that drive return visits is to facilitate meaningful customer exchanges. If a brand is asking for customer data, there should be an equal exchange — or offer made to the customer — for that data.
Armoire gamifies this exchange by offering “sneak peaks” at fashion items that the customer is then asked to choose what they like and don’t like. The company also requires customers to review every single item they rent. (Armoire is rental clothing company that customers can use by paying a monthly fee). Both of these efforts provide a continual loop of customer data that results in sticky digital experiences.
3. Add behavior and emotion metrics to your data sets
Machine learning takes time, especially for a company like Armoire that relies solely on first party data curated from customers’ experiences. Behavior tracking metrics — such as scroll depth, heat mapping, click mapping, content viewed, time on page and landing exit page — can provide valuable insight. Emotion tracking is also beneficial, with tools now available to measure sentiment analysis, facial coding, voice analytics, gesture tracking and more.
“There’s some privacy concerns and elements of the creep factor here,” said Taylor. She recommends combining natural language process analytics with sentiment analysis for brands wanting to go deeper into emotional tracking measures.
There’s also something to be said for actual conversations with your customers. “The 360-degree view of the customer sometimes forgets the emotion,” said Taylor, “Each of us is better at interpreting human emotions than the most sophisticated technology.”
4. Build a relationship with your customer, but tread lightly
When integrating machine learning and AI into your marketing efforts, Taylor points out that sometimes technology allows brands to know more about an individual than their friends or family may know.
“We don’t tell people every single detail of our lives,” said Taylor. The same should be considered when using sophisticated technology to build digital experiences — this is why Taylor’s brand doesn’t necessarily curate an experience using every single data point they may have access to. Being transparent and keeping a dialogue about how customer information is used builds trust with a customer that can lead to a valuable relationship. “When they know how we’re using their data, it motivates them to give us more data,” said Taylor.
Another tactic Taylor’s team implements is spontaneity, recommending products outside of the algorithm to, in her words, surprise and delight customers and to learn more about them. Taylor said the “surprise and delight” approach is a great way to make machine learning more human.
5. Keep actual humans in the loop
“Because we are using an automated approach, we don’t want to follow it up with more AI,” said Taylor about the company’s business model, “That’s why we keep human interaction with customers to answer style questions.” Armoire also has set up physical locations where customers can schedule appointments with a stylist. When a stylist does meet with a customer — either in person or over the phone — there are protocols in place for the stylist to input the customer data into their systems.
“The human interactions improve our AI and our AI improves our human interactions,” said Taylor.
6. Machine learning + user design = exceptional digital experiences
As the head of user experience and product design, Taylor believes data scientists should be integrated within the user design team. By bringing data scientists into the fold, the UX team can build relevant experiences based on quantifiable data. “Machine learning and UX should go hand-in-hand,” said Taylor.
Taylor realizes that not all brands may have the resources or bandwidth to practice all six rules she outlined during her presentation. If forced to choose, she says the first and last rule are the most crucial: Determine the ongoing value you’re creating and make sure your machine learning systems are integrated into your user design processes.