It’s always interesting to hear how AI is revolutionising specific industries, and few companies are more qualified to comment on the impact on the retail industry than Shopify.
AI News caught up with Vincent Chio, Data Science Lead at Shopify, to hear what the company is doing in the space and how AI is improving the end-to-end retail experience.
AI News: How has AI changed the shopping experience in recent years for both sellers and buyers?
Vincent Chio: From automated marketing, to smart fulfilment, and predictive analytics, AI has made it easier for retailers to manage and grow their business. But not every retailer has the resources to build and implement AI in their business.
That’s why at Shopify, we really focus on bringing the power of AI—which has historically been reserved for enterprise businesses—into the hands of our merchants, who are businesses of all sizes. To provide a few examples of what that looks like, we use machine learning in our fulfilment network to predict the closest fulfilment centres and optimal inventory quantities per location to ensure fast, low-cost delivery of our merchants’ products. Our business chat app, Shopify Inbox, is built on a natural language processing foundation that helps merchants convert conversations into sales. And, through state-of-the-art machine learning models that predict merchant business success, our product Shopify Capital automatically sends our merchants funding offers without them having to apply.
How this shows up for and benefits buyers is in the form of more accessible, seamless and personalized shopping experiences. From targeted product recommendations, to faster shipping and better ways to connect, buyers expect more from retailers, and AI is helping retailers meet those expectations.
AN: What are some of the hurdles you’ve faced in bringing an AI model into production and how did you overcome them?
VC: When it comes to shipping anything, there are always hurdles you’re bound to face. At Shopify, we follow a few guiding principles for implementing and scaling AI that ensure easy adoption from the get-go.
At Shopify, we take a merchant-first approach to identifying problem areas. So our first principle for implementing and scaling AI is to make sure that what we’re building is solving a merchant problem, and that we have enough data to create a solution.
Second, we start simple. If a regression model will solve our merchant problem, that’s where we start. This doesn’t mean we avoid building complex models, it just means that we first prove that we can use AI/ML to solve the problem, and then we iterate by building complex models.
These two steps are key for getting stakeholder buy-in which, if you don’t get, can stop your project before it even gets off the ground. We’ve got more principles and tips that you can check out here.
AN: An increasing number of third-party integrations are available for Shopify that harness AI—what are some of the most unique and/or interesting ones in your view?
VC: We’ve got a ton of third-party developers creating innovative apps that help extend the capabilities of our merchants’ stores. Some of the most interesting ones, in my opinion, are the ones using AI to translate those authentic, in-person retail experiences to online stores. Like I mentioned earlier, AI is changing the shopping experience by using data to bring more personalization to retail.
Our third-party apps that use algorithms to help merchants optimize the full buyer journey, from marketing and conversational automation, to recommendations and cart abandonment, not only have the power to help merchants convert, but they’re creating better shopping experiences for buyers.
AN: Shopify introduced LinNét earlier this year, its new product categorisation model. How does the new model differ from its predecessor and why was it deemed necessary?
VC: Shopify has seen amazing merchant growth in the past 2 years, hitting over 1.7 million merchants across the world. New merchants means new products, so we decided to reevaluate our existing product categorization model. We wanted to evaluate our old model’s performance because it’s important that we understand what our merchants are selling, so we can build the best products that help power their business growth.
After evaluating key metrics like how often our predictions were correct and how often do we provide a prediction, while also taking a look at our product road map and how our model might support new products, we decided to build a new model to improve our performance.
Compared to our old model, LinNét not only uses text features for prediction but also images. On top of this, LinNét has the ability to understand products in multiple languages. LinNét was also part of a larger effort to modernize our machine learning systems and we can now do things like real-time prediction, which we couldn’t do with the previous model.
With these new features, LinNét has increased our leaf precision by 8% while doubling our coverage. If you’re interested in learning more, read our blog!
AN: What’s next for AI at Shopify?
VC: We’re excited to further leverage the scale of our data to not only empower Shopify but to create new experiences for our merchants that are impossible without data.
Some of the ways we’re doing that is by exploring how to better support merchant workflows through product understanding and creating experiences for merchants that suggest best actions for their workflows, while foregrounding merchant autonomy.
How that will show up is through eliciting merchant feedback through accepting or rejecting our recommendations, and through education around our machine learning approaches to workflow optimization.
AN: Shopify is sponsoring, speaking, and exhibiting at this year’s AI & Big Data Expo Europe. What can attendees expect from your presence at the event?
VC: Attendees can expect the chance to really get to know the Shopify Data Science & Engineering team, and the kind of work we do.
On day one of the expo, you’ll get to hear more from me! Diving into Shopify Inbox and the natural language processing foundation behind the product, I’ll illustrate how we accelerate product development with AI at Shopify, providing takeaways that can be used at any organization. I’ll also cover how to build a data foundation to establish trust and identify opportunities only AI can solve at scale.
Then, in the afternoon, you’ll have the chance to hear from Shopify’s Yizhar Toren, Senior Data Scientist. Yizhar will join a panel conversation on ramping up AI projects, discussing key tips like how to move your project from experimentation to production, and turning AI into ROI.
Attendees will also get the chance to meet our speakers and recruiters at our booth on the exhibition floor. Come say hi at booth #301!
Shopify will be sharing its invaluable insights during this year’s AI & Big Data Expo Europe which runs from 23-24 November 2021. Shopify’s booth number is 301. Find out more about the event here.