Why we’re using Google Assistant to make shopping with ASOS more intuitive
As humans, we learn to speak before we can type. Now it’s common practice to talk to technology, what will the rise of conversational interfaces mean for online fashion? This is the story behind our new action on Google Assistant.
Testing and validating hypotheses is something we’re big advocates of on the digital product team at ASOS. Our ‘always in beta’ mindset helps us test our riskiest assumptions early on, so we’re co-creating amazing experiences with customers and making sure we’re building the right thing, right.
One area we’ve been exploring over the last year is Conversational Commerce — using natural language processing and understanding to power messaging and voice interfaces with customers. We believe Conversational Commerce could help us address one of our customers’ biggest challenges: how to find something you love, that’s truly relevant to you, quickly amongst our catalogue of 87,000 products.
At the centre of this is intuition. The more we enable customers to browse and navigate as naturally as they do through conversations with friends (and increasingly, interact with things like smart devices), the more delightful we can make shopping with ASOS.
Hey Google, Talk to ASOS
That’s why we’ve launched a new way to browse ASOS using your voice on Google Assistant.
We know our 20-something customers are early adopters when it comes to technology and voice search and voice platforms like Google Assistant. Our large product catalogue, with 5,000 new products added each week, means discovering that perfect pair of jeans or staying on top of what’s new on ASOS is not always easy. The first iteration of our ASOS action on Google Assistant is intended to help change that, by helping customers browse a curated set of the latest drops in a more natural way.
We settled on this initial use case because the desire to stay on top of ASOS is something that’s a common experience for a lot of our customers. But in order to satisfy their needs, we have intentionally streamlined our offering, limiting the number of product categories that are available at launch.
While enabling us to offer a more optimal MVP, one of the main reasons why we want to start simple and develop our experience into a more sophisticated shopping guide over time is because designing for voice and messaging is fundamentally different to the other user experiences we offer.
Before we’re taught to type, we learn how to speak. It’s something we’re naturally very good at — and consequently, when we’re given the opportunity to talk to things, we have high expectations of what the technology will deliver. Any early voice adopters out there will know that it’s not always easy to meet those expectations.
With that in mind, we think it’s more important than ever to bring our customers on the journey, to focus on solving one of their needs with our Action before trying to resolve the rest. And publishing our Action means we’re also learning about this new form of interaction every day.
Powering our MVP
We went with Google’s recommendations for the technology stack used for our Action, as they’re well documented and allowed us to release our MVP in less than eight weeks.
We used Dialogflow as our NLU engine, a web-based service provided by Google that uses an agent to process user input into Intents. An Intent is an underlying goal or task the user wants to do. For example, asking for ‘show me blue jeans’ would be mapped to a SelectCategory Intent.Dialogflow transforms speech to text but with much more additional information in it.
These intents are forwarded to our fulfilment middleware, which defines our Action’s functionality, and is responsible for mapping intents to business logic. This either communicates with some of our ASOS services or generates and sends back to Google Assistant conversational elements.
In voice, users can be hard to predict, so we’re continuing to build out our dictionary and grammar library, while also ensuring we make the experience feel relevant and personalised to each user — for example, avoiding the repetition of questions they’ve answered previously. As more devices come into the market, it’s also important to consider and adjust to the different behaviours users might have on a device with a screen vs. when they’re voice only.
The future
We’re just at the start of our journey to unlock the true potential of these interfaces — whether that be on Google Assistant, apps like Messenger, which allow customers to chat to Enki to discover personalised recommendations, or within our own ASOS experience.
With any new technology, it’s hard to predict how conversational commerce will develop. That’s why we’re approaching it by experimetning with different ideas that we believe will make our customers’ lives easier and our shopping experience more intuitive.
As we understand more about the context and intent behind what our customers are browsing for, and how they are doing so, we’ll discover new areas in the conversational space in which to work. Some might be opportunities we know about. Others might come from learnings from our ASOS Action on Google Assistant. We can talk about those another time.
Our ASOS Action on Google Assistant is available in the UK and the US now. You can start the conversation by saying: ‘Hey Google, Talk to ASOS’.
—
Jason Gregory is a Senior Product Manager at ASOS exploring the opportunities of using voice and messaging to improve the digital shopping experience. Outside of ASOS, his main passions are the unusual combination of music, golf and gardening, although rarely all at the same time.
This article was co-authored with Chris Oikonomou, a Senior Android Engineer at ASOS who has been experimenting with Conversational UI for the past 18 months and is fascinated by how powerful and intuitive it can be. Aside from ASOS, Chris travels quite regularly and uses photography a creative outlet.