Splendid Unlimited

You talking to me, cause I don't see anyone else standing here?

Voice UIs and what they mean to your business Splendid’s Benedict Ireland and Edward Hasting-Evans

November 28th 2016

So Amazon launched Echo, and it’s really wowed the world… or not. Since then, we’ve seen the launch of Google Home, and Apple’s HomeKit. But, like the wearables market, the sales haven’t matched the hype. Well, at least this seems to be the prevailing opinion, but is it true?

Well kind of. But that’s not necessarily a bad thing. According to New York-based consulting firm Activate Inc, sales of Echo are ‘in-line with those of the original iPhone’. The key to iPhone was the success of Apps. Sure, we went through that terrible phase of trying every App we found, and that resulted in some truly awful experiences. But the market matured incredibly quickly, and Apple sorted out their App development and approval process.

While Echo and Google Home both have obvious service bias, that doesn’t have to stay the case, and indeed both companies are encouraging developers to see what they can do with the platforms in order to create compelling use cases.

At Splendid, we’ve been getting down ‘n dirty with Alexa for a while, and recently put together a demo for our Group’s Tech Expo, hosted in our London HQ. The Expo was themed ‘The Future of Retail’; therefore, we needed to think retail, but Echo already does that, with arguably the largest retailer in the world. So what better challenge?

The Institute of Grocery Distribution tells us that the UK grocery market will be worth £203bn by 2019, and we believe that the grocers who provide not only best value but best experience and choice are the ones that will triumph in the war for our baskets. We know that customers want choice, and people don’t like to feel they are ‘forced’ to buy everything through one retailer, which is likely to be a perception Amazon face with Echo.

We decided to create a fictional Grocery store, Splendid Grocery, to see what challenges there are with voice interaction in everyday situations, and what we could do about them.

So, what did we do? We’re not an international supermarket with access to a vast product catalogue; luckily Tesco has an open API giving us access to their whole product database, complete with product shots and pricing.

From there the development was run in three streams. With only one developer in each, one concentrated on building the VUI (the Voice User Interface because that’s what it amounts to) and ensured that Alexa (Amazon’s AI) stood the best chance of understanding what the user was asking for, and what they expected her to do about it. This involved creating utterances (lots of versions of phrases) and intents, which is what the user is expecting Alexa to do. These form your app interaction model – and in keeping with the AI nature of Alexa, apps are called Skills.

Just think, a user who wanted to know where their delivery is might say ‘Alexa, ask Splendid where’s my stuff’, or ‘Alexa, ask Splendid where my order is’ – you get the picture.

The second work stream concentrated on building a service that Alexa would communicate with, which integrated with Tesco’s API providing product information, such as pricing and product photography. Alexa would pass across the product information the user had said, and would also let the service know what the user’s intent was (for example, add {product} to my order).

We now had a working demo of a voice UI, which would handle key user intents, such as adding products to a basket, place a fictional order, check delivery dates and check stock in their local store.

However, we felt we were missing something. We knew voice would not replace other interfaces, but augment them, and that wasn’t getting across. So, we had a third developer build a responsive basket for the shop which showed the orders being added to the basket as the user spoke to Alexa. When a user placed an order – ‘Alexa, ask Splendid to order my products’ – the basket page became empty, and the user could see their order appear in their order history page.

Alexa demo video still Play

Through our explorations we quickly realised that a user, unless they are particularly determined, is unlikely to build up a weekly shop one product at a time. But they might add items ad-hoc throughout the day such as running out of toothpaste in the morning and saying, ‘Alexa, ask Splendid to add toothpaste to my order’. Or they may have a weekly shop set-up and simply say, ‘Alexa, ask Splendid to re-order my weekly shop’.

Throughout the Expo the prototype generated a lot of interest from both retailers and industry experts; the wariness was palpable. So what is it that’s making people uncomfortable? Well, for a start, choice. It’s hard to pick exactly the right packet of pasta when having them read out to you by a robot, and it’s more time consuming than tapping on a screen. Data can be a huge help here — if you know they always order Pilgrims Choice Mature Cheddar, next time they ask for cheddar you can make an educated guess as to what they want.

However, the fact is that a taxonomy which works for devices doesn’t necessarily work for voice. So we need to work hard on creating a taxonomy layer which effectively translates hierarchical choices we’ve become trained to make, and convert them back into much more natural, conversational experiences.

There’s also the privacy issue. As Echo has the concept of a single account, you might not want your kids/wife/husband/flatmates to know you’ve ordered Imodium. And the individuality issue goes a little deeper: think about prescriptions. Placing an order requires authentication of who the person is, to ensure we don’t make potentially deadly mistakes (although we have a hypothesis we’ll be testing out around accounts and security).

But none of these problems are insurmountable. And in fact, some of them go away when we apply voice as an augmentation of existing experiences. Within your Ocado app, imagine if you could ask Siri to ‘buy me apples, cheddar, and some water’ (Apple, in iOS 10, have opened Siri up to app developers through their SiriKit) and through being intrinsically linked to your account it knows your preferred brands, sizes and flavours. So, no need to filter. Again, single account linking, as is the model of almost all online shopping experiences, gets us around the prescription issue by inherently knowing who we are talking to. There are definitely solutions out there, we just need to be open to voice interactions to explore them.

Consumers are becoming more and more comfortable with voice interaction, and according to a very interesting lady we met at the Expo who had carried out some research, the vast majority of people surveyed who have Amazon Echos found that their use of voice interactions with other tools, such as Siri, increased (which many had not used before).

There are many other areas where voice could help — consider people who struggle to use today’s many buttoned TV or DVD controllers. How much easier to say ‘Alexa, ask Sony to put Channel 4 on’. Or in pharma, ‘Dave, you are due to re-order your repeat prescription, would you like me to order it for you?’

There are many sectors which could benefit from exploring what they can do with voice to better serve their customers. Our voice prototype work in the energy sector is also gathering steam (not literally), and currently allows home control apps to connect to Echo without the need for additional proprietary hardware, and again it works through an API from a very large UK energy provider... a demo that will be forthcoming in the next couple of weeks, stay tuned.

And in the meantime, get in touch and chat to us about what Voice could mean for you.

Benedict Ireland


Benedict Ireland

Head of UX


0207 395 4800

Edward Hasting-Evans


Edward Hasting-Evans

Head of UI Development


0207 395 4800


Get in touch with us