Voice tech in travel, part 3: The Skyscanner story

The days of typing search terms into a browser or clicking
through drop-down menus may be fading. Rising in their place is a much more
natural interface: voice.

Market research firm Gartner predicts consumer demand for
voice devices such as Amazon Echo and Google Home will generate $3.5 billion by

Statistics regarding consumer adoption of these products are

In January, Google announced it sold “more than one Google
Home device every second since Google Home Mini started shipping in October.”
Google Assistant – which powers all of the Google Home products (original, Mini
and Max) and also works on Android phones and tablets, iPhones, TVs and watches
– is now available on more than 400 million devices.

But Amazon still dominates the smart-speaker market, with
third-party research estimating the Alexa products control 76% of the total
user base. In December, Amazon said “tens of millions of Alexa-enabled devices
sold worldwide” during the holiday season, and the Echo Dot was the top-selling
product from any manufacturer in any category across all of Amazon.

As consumers become more comfortable conversing with a
device for information and shopping, brands are investing more resources –
human and financial – to develop voice-enabled solutions.

In part three of our series on voice, we talk to one travel’s
earliest entrants into voice technology – Skyscanner.


Metasearch engine Skyscanner was one of the first travel
sector brands to introduce conversational search interfaces.

Its first iteration came in 2015, when it created a hotel
search bot for the Telegram messaging service. The bot was able to understand some
natural language input such as searching with the term “next weekend.” 

In March 2016, Skyscanner made headlines when it released a voice
skill for Amazon’s Alexa, making it the first travel search engine on that
platform. Users talk to Skyscanner on the Alexa device to find the cheapest
flight options for their preferred routes and dates, but booking still takes
place through Skyscanner’s app or website.

A few months after the Alexa release, Skyscanner debuted a
bot for Facebook Messenger, followed by a Skype chatbot in August 2016.

While it is similar machine learning that powers all of the
interfaces, the Alexa connection is the only one that’s fully based on voice
input – the others are primarily text-based, although users may opt to dictate
their query into their device.


Skyscanner has now drawn more than one million unique
interactions across its chat interfaces [VERIFYING THIS IS ACROSS ALL AND NOT
. Vice president for product management Filip Filipov says one thing it has learned is that pure voice communication is very challenging.

“It works really well if you are doing utility functions –
what’s the weather going to be tomorrow? How far is New York from where I am?
Those work really well with voice,” he says.

But conversations do not always follow a predictable
question-and-answer structure. Filipov says Skyscanner quickly learned it needed to
create hundreds of paths so users would not “end up on a dead-end street where
the conversation just stops.”

And even with that complex database of information, Filipov
says it has still concluded that – at least for the next few years – voice
interfaces are best suited to assist customers with discovery and inspiration
rather than purchasing.

“I think we as travel sites – because that’s how we
function, that’s how we make money, that’s how we fulfill a need for the
customer – we always want to lead to that transaction,” he says.

“What we’ve learned over time is if people start using it as
a utility, you will still provide a lot of value and then potentially later on
those people will come back and complete a transaction.”

Pros and cons

And by testing various conversational interfaces, Skyscanner
is learning the strengths and limitations of each. For example, Filipov says
the visual output available via platforms such as Facebook Messenger and Skype
makes the experience easier for travelers.

Responses can be presented with a list of choices that include
photos, so users simply click on the option they want rather than having to
type or speak an answer.

“For us as a comparison site we need to get to a stage where
people can truly compare. Right now we can provide an answer, but the
comparison, which is at the core of what we do, is not there purely on voice.
With a chatbot it’s a lot more present.”

One voice solution that creates more options for brands and
easier interaction for consumers: Amazon’s Echo Show, which combines a voice
interface for input and a visual interface for output. Particularly for complex
queries that can be involved with booking flights and hotels, the visual elements
can put the consumer at ease and may lead to a more frictionless path to

Future outlook

Filipov says one of the limitations in the current Alexa
structure is the fact that it requires users to identify a specific brand before
submitting their question.

“For example to get to Skyscanner, you need to say ‘Hey Alexa, open Skyscanner,’ and then you can start the conversation,” he says.

“But you and I both know you can’t keep all the brands in
your mind all the time. So it takes quite a bit of effort.”

One way he suggests to eliminate that would be to make the
Alexa interaction function more like a search on Amazon: a user would simply speak
a question to their Alexa device and allow the system to pick the best option
to provide the answer.

And for brands that are just starting to explore voice
technology, Filipov has this advice: if software engineering is not at the core
of what you do, don’t go it alone.

“We are a technology company. And we have decided that this
is an important channel,” he says.

“We’ll learn, we’ll build it, we’ll continue building it,
we’ll pause it at certain times, we’ll double down on the investment. But if
you are a brand that doesn’t have the capabilities in terms of technology, don’t
do it. Get someone else who can build it for you.”

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *