00:04KAI HALEY: Thank you all
for joining us today.
00:07And thank you for joining
us on the Livestream.
00:09We're very excited to be here
to share with you prototyping
00:14voice experiences, design
sprints for Google Assistant.
00:18WALLY BRILL: You know what, Kai.
00:19They sent all the hippest,
coolest, and best looking
00:23people at I/O right here.
00:28You guys are looking good.
00:33I lead the design relations
team here at Google.
00:36And I'm also one of the founding
members of our Sprint Master
00:40So I'm pretty passionate
about design sprints.
00:43WALLY BRILL: And
I'm Wally Brill.
00:44I'm the head of conversation,
design, advocacy, and education
00:47here in beautiful
downtown Mountain View.
00:50And I'm really stoked
because I spend
00:52my life helping people create
conversations with robots.
00:57And that's a part
of what we're going
00:59to be talking about today.
01:01KAI HALEY: So today
we're going to talk
01:04about how you can apply our
Voice Action Design Sprint
01:08model to prototyping experiences
for the Google Assistant.
01:14For those of you who are not
familiar with design sprints,
01:17we have some great
resources out there for you.
01:21And we recommend you
check out our website,
01:23Designsprintkit.withgoogle.com.
01:26At a high level, design
sprints were created at Google
01:30to help teams shortcut the
traditional product development
01:34process, to get to
learnings faster
01:37so they could make better
decisions about the direction
01:41The framework is
organized into six phases
01:45to support divergent
and convergent thinking.
01:49And this is really founded
on traditional UX practice.
01:53In the understand phase,
a cross functional team
01:56comes together to
explore the problem space
02:00from 360-degree angles, and
really develop a deep empathy
02:07In the define phase,
the team will then
02:10determine which area they
want to focus the sprint on
02:13and where they're going to have
the most value for their users.
02:16Then the entire team sketches
a broad range of ideas
02:19in the sketch phase, basically
creating a hypothesis of what
02:23will meet their user's needs.
02:25Then, as a team, we decide what
to prototype for the user test
02:31So that's a high level
look at design sprints.
02:35The framework is very flexible.
02:37And on our website
we have the core set
02:39of methods, which
you can see here.
02:43But today what we're
going to be talking about
02:45is the Voice Action
Sprint model,
02:47which is really a combination
of the traditional design sprint
02:51with a role playing workshop.
02:53And that's where the
conversation design
02:57The Voice-Action
Sprint framework
02:59that we've created and
prototyped over the last year
03:04And we've done this in order
to include role playing
03:08And we're going to go
through the methods in detail
03:11today so that you can run a
Voice Action Sprint yourselves.
03:16Here is the list of them.
03:17As we mentioned,
there are resources
03:19online for you to
learn more about them.
03:22WALLY BRILL: So when do we want
to use a Voice Action Design
03:25What's a reason to use it?
03:27So in a situation
in which you're
03:29trying to determine a
direction for a product,
03:33for a voice product, in a
situation where you are trying
03:37to determine the brand
persona, the actual personality
03:40of a brand, or you're
trying to create and develop
03:44dialog for the action,
these are all great reasons
03:47to use a Voice
Action Design Sprint.
03:50On the other hand, we don't
need to do it if we already
03:54have our dialogue, if we already
have our persona nailed down,
03:58and we know what we're
actually going to be building.
04:00At that point, we can go
straight into the prototyping
04:05KAI HALEY: So as with
any design sprint,
04:07planning is really critical
to the success of the sprint.
04:10You're going to have
a group of people
04:12together for a dedicated
period of time.
04:14And you want to use
their time effectively.
04:16So the first thing to do is
define the challenge and scope
04:20the sprint appropriately.
04:23We have some examples
that can illustrate how
04:25you might scope your sprint.
04:27Earlier this year,
we ran a sprint
04:29with Filtrete, which is
3M's air filtration product.
04:34And with them we were
developing a unified vision
04:37for the voice experience of
their suite of air filtration
04:41So you can see, this is
a really broad sprint.
04:44We're including
visioning work, as well
04:46as a number of products.
04:48And the deliverables
that this sprint yielded
04:50were key use cases, brand
persona, and validated voice
04:57Another sprint we ran
this year was with H&M.
05:00And we scoped it this way.
05:01Identify opportunities to better
meet the needs of H&Ms existing
05:06customers through
voice assistants.
05:09And you can see
here, we're looking
05:10for a couple of opportunities.
05:13So it's a little
bit more narrow.
05:14And we're also focusing on a
specific audience, the existing
05:19In this instance, we also
created key use cases,
05:22brand persona, and
validated voice prototypes.
05:27You can also narrow your
sprint down in other ways.
05:30For example, you might
look at a specific use
05:33case or a specific context.
05:35For example, you could
scope the sprint this way.
05:38Define the voice
experience for daily tasks
05:42that could integrate
into routines.
05:44And routines is a feature
of Google Assistant,
05:46where you can set up daily
things that you want done.
05:50You will always want to
create a brand persona.
05:52And you might want to identify
a specific number of tasks
05:56that you'd like to prototype.
05:58There's a couple other
things that you'll want
06:00to do in the planning phase.
06:03As in a normal
design sprint, you'll
06:05want to schedule your
lightning talks in advance.
06:07And these are very important
because they are the inputs
06:11for the ideation portion.
06:13We'll talk a little bit more
about those in a minute.
06:15You're going to want to
design your sprint team.
06:17It's very important that you
have the right people there.
06:20If you don't have a conversation
designer on your team,
06:23you're going to want to
recruit one for the sprint.
06:26And a developer familiar
with Dialog Flow
06:28is very valuable if you're going
to want to create a Dialog Flow
06:32Of course, you're going to
want to recruit five user test
06:35participants and secure
any devices or recording
06:39equipment that will be needed.
06:42Then you'll be ready to move
into the understand phase,
06:45where we're going to go
through lightning talks.
06:48We're going to use how might
we sharing and affinity
06:50mapping to identify
opportunities,
06:52and an experience map
to look for pain points
06:56and build more empathy
with our users.
07:00Now, I mentioned
lightning talks.
07:02These are the inputs
for your brainstorm.
07:05So they're really important.
07:07They help set the
context for the problem
07:09solving in the sprint.
07:11With a Voice Action
Sprint, because we're
07:14working on a relatively
new platform in potentially
07:18conversation design might
be new for your team--
07:20there's going to be an
educational component.
07:22And you'll want to bring
lightning talks that
07:24cover conversation design
principles, the idea
07:27of open-ended versus
guided conversations,
07:31and great conversational
task examples
07:33so you can inspire people
on potential tasks you
07:36might want to design for.
07:38And Wally's going to
talk us through the ideas
07:41behind these lightning talks.
07:43WALLY BRILL: Thank you.
07:45So a couple of things
we want to keep in mind.
07:47When we're talking about
these lightning talks,
07:50we're really talking about
principles of conversation
07:53and things we need
to know before we
07:56go into a design phase.
07:57So one of the things that's
important to keep in mind
08:00is that visual design
and conversation design
08:03are radically different.
08:05Visual design persists.
08:07I can look up the page to
see what's come before.
08:11I can look down the page
to see what comes after.
08:14I can't do that with
conversation design.
08:16Conversation design
is inherently
08:20linear and ephemeral.
08:21It goes by in real time.
08:23So I have to capture the
things that I need in real time
08:26so that I can hold
them in memory.
08:30We also like to
talk about, in terms
08:32of the principles of
conversation design, Paul
08:36Paul Grice was a linguist
philosopher who came up
08:39with the cooperative principle.
08:40And the cooperative
principle is really the idea
08:43that when two people
have a conversation,
08:46they want their contributions
to be truthful, informative,
08:50relevant, and clear.
08:53And the reason is, they want
the conversation to succeed.
08:56So we help each other
in a conversation.
08:59We know when to give the turn.
09:01We know when to take the turn.
09:03We're propelling the
conversation forward
09:08Now, let me tell you a
little bit about open-ended
09:10versus guided conversations.
09:12It's perfectly normal in
a situation in which I
09:16know the domain that I'm dealing
with for me to say something
09:19like, how can I help you?
09:21Let's listen to this.
09:22GOOGLE ASSISTANT:
Welcome to Acme Widget.
09:26WALLY BRILL: It's a totally
open-ended question, right?
09:28Any number of answers.
09:31I want to paint my Camaro gold.
09:34All kinds of possible
things I could say there.
09:36So I need to constrain
the area that's
09:40available in terms of
possible responses.
09:43So let's take a look at
what United Airlines did
09:47to help people understand
what their action could do.
09:51UNITED AIRLINES ASSISTANT:
Welcome to United Airlines.
09:53Please begin by saying, "What
is the status of my flight?"
09:56or "Check me into my flight."
10:01WALLY BRILL: What's the
status of my flight?
10:03Check me into my flight.
10:06So now let's move on.
10:09When was the last time you
reached into the refrigerator
10:13and you pulled out
that chicken you
10:14were going to cook for dinner?
10:16And you're holding it.
10:18And it's a little
yucky, we have to admit.
10:20But we forgot the recipe we
were going to use to cook.
10:26This is what we call
the raw chicken moment.
10:29There is nothing you
can do with your hands.
10:31You're not going to
reach into your pocket
10:32and pull out your
phone and start
10:34tapping through apps, right?
10:35You're not going
to touch a touch
10:37screen, because your hands
are covered in chicken.
10:43And we use a voice
forward application.
10:46Let me show you what Tasty does.
10:49Tasty helps us with recipes.
10:52Here's what they sound like.
10:54SPEAKER 1: OK, Google.
10:59GOOGLE ASSISTANT: Sure.
11:01TASTY ASSISTANT: Welcome back.
11:03Would you like to
continue making chocolate
11:04fudge box brownie cheesecake?
11:07SPEAKER 1: No, thank you.
11:11TASTY ASSISTANT: OK.
11:12What are you in the mood for?
11:15SPEAKER 1: Cheesy
asparagus quiche.
11:19TASTY ASSISTANT: Sure.
11:21Does this look good?
11:24WALLY BRILL: Cheesy asparagus
quiche always looks good, OK?
11:29Put that in your notes.
11:30Cheesy asparagus quiche
always looks good.
11:34Let me give you some examples
of some other conversational
11:37So when you're selecting your
task to create the action,
11:42remember, something
people can do
11:44in conversation, something you
can do with your eyes and hands
11:48busy, and something
that has value.
11:53Headspace provide meditations.
11:56When you had a
jangly kind of day
11:58and things are all kind of
crazy, you contact Headspace,
12:02and you get some
meditation action.
12:05HEADSPACE ASSISTANT:
Welcome back to Headspace.
12:07You can repeat
today's meditation,
12:09try a sleep exercise, or
unlock more meditations.
12:13What would you like to do?
12:15SPEAKER 2: Today's meditation.
12:19HEADSPACE ASSITANT:
Today's meditation.
12:22ANDY PUDDICOMBE:
Beginner's mind is
12:23something to bring to
every moment in life,
12:26not only to our meditation.
12:28So whether you're watching the
breath, or eating some food,
12:32be curious, as if
for the first time.
12:37WALLY BRILL: So a
little meditation.
12:39And Sephora brings us
beauty and skincare advice.
12:43It will tell me what
kind of skin I have.
12:45Beautiful, of course.
12:48And it will tell me how to take
care of my beautiful and supple
12:53SEPHORA ASSISTANT: Welcome to
the Sephora skin care advisor.
12:56Great skin starts here.
12:58Would you rather
discover your skin type
13:00or hear a skincare tip?
13:02WALLY BRILL: I should
probably hear a skincare tip.
13:06And Filtrete that Kai
was just talking about.
13:08So Filtrete creates smart
filters for air purifiers.
13:13And those smart filters
are really smart.
13:16They'll tell you when
they need replacing.
13:18And the action will
actually give you
13:21information about air quality.
13:22Let's have a listen.
13:23FILTRETE ASSISTANT: Your
home filter has 5% left
13:26and needs to be replaced soon.
13:28What else can I help you with?
13:30WALLY BRILL: Well,
that's enough.
13:31I better replace my filter.
13:34KAI HALEY: So in addition
to these lightning talks,
13:38you'll want to also bring
research about your audience.
13:42Who is it that
you're designing for?
13:44How do you think about them?
13:45What are their needs?
13:46And in this context,
you'll want to understand
13:49how familiar are they
with voice assistants?
13:51Are they currently using
voice-based products?
13:55And there are a
number of sources
13:57of data that you can go
to to get information
14:01You can go to your
contact call center
14:04and see what they're trying to
accomplish with voice today.
14:07That's a great resource.
14:08Or you can do more traditional
user research, like diary
14:11studies, interviews, or
participant observation.
14:15But the important thing
here is to understand
14:17the context in which somebody
is going to be completing
14:20a task with your product.
14:22And what are the concerns that
you might have in that context?
14:25Are they in a public space?
14:27Do you need to be
worried about privacy?
14:29Really think about
where you're positioning
14:31these people when they're
using your product.
14:35After you've built
shared knowledge
14:37as a team in the understand
phase with the lightning talks,
14:41you're going to
do a how might we
14:42sharing and affinity
mapping exercise.
14:45And this really helps the
team to identify opportunities
14:49and really understand the
problem space together.
14:52At this point, you'll
choose a specific area
14:55that you want to focus on,
potentially a good idea
14:59for your conversational tasks.
15:02You can also use
experience mapping.
15:05This is a method from
traditional sprints
15:06as well, which is really
helpful for putting yourself
15:09into the shoes of your user and
identifying more pain points
15:14and opportunities, ways
that you can help them.
15:18The nice thing about
an experience map
15:20is it helps you step above
your current product experience
15:24and look at the context
in which a user is
15:27trying to solve a problem.
15:29And this provides
more opportunities
15:31for how you can integrate
meaningfully into their lives.
15:37And now you're ready to move
into the define phase, where
15:39in a voice action sprint, you
want to create a brand persona.
15:43We're also going
to look at what's
15:45the user value that we
can offer to our users
15:48and agree upon success metrics.
15:52WALLY BRILL: Let's talk
about persona for a second.
15:54This is something that's
near and dear to my heart.
15:56We talk about brand persona.
16:01Now, how many of you people
brought a persona with you
16:07Actually, you all did.
16:08It's a trick question.
16:12Here's the thing
about brand persona.
16:14A brand has an identity.
16:19Just as with a person,
within a second and a half
16:22of hearing a voice,
we intuit all kinds
16:26of things about who
we're listening to.
16:29So too when the
brand has a voice,
16:32we'll intuit all kinds of
things about that brand.
16:36So we have to do a design job
to create the right persona
16:39to match the brand identity.
16:41Now, in that first second and
a half of hearing a voice,
16:45you get a sense
of register, which
16:47is the social status of that
entity in relation to me.
16:54Where does it sit in
the pecking order?
16:56We get a sense of gender.
16:58We get a sense of age.
16:59We get a level of education.
17:02And we also, if they have an
accent, get a sense of locale.
17:05But even more, we intuit
intelligence, trustworthiness,
17:11and likability in
the first second
17:14and a half of hearing a voice.
17:15It's something we unconsciously
and automatically do.
17:18We create an image in our own
mind of who we're listening to.
17:22So it's absolutely vital that
brands bring their essence
17:26to the table, and define and
create a brand persona that
17:30matches that essence.
17:34Let me give you an example
of a really cool persona.
17:37So if you've got kids--
anybody here have kids?
17:40You know who the Wiggles are?
17:43Everybody loves the Wiggles.
17:44So Emma Wiggle is the persona
of the Wiggles' choose your own
17:49And she's out to find her
friends who are missing.
17:52She's got to go to
a concert tonight.
17:54Let's have a listen.
17:57EMMA WIGGLE: Hi, there.
17:59You're here just in time.
18:01There's a Wiggles concert today.
18:03But I can't find
any of my friends.
18:06Why don't you help
me look for them?
18:09I'll help look for them.
18:12Now, H&M. As Kai mentioned,
we did a sprint with them.
18:16They had a really
interesting idea.
18:19They wanted to create a fashion
and style maven, somebody
18:24who was a friend of yours
but who also knew everything
18:27there was to know about style,
everything there was to know
18:30about fashion,
and could help you
18:32find products and choose gifts.
18:35H&M ASSISTANT: Hi, again.
18:36Do you want to find
a new gift, or have
18:38a look at the things
you've saved last time?
18:41SPEAKER 1: Find a new gift.
18:43H&M ASSISTANT: Do you know
what you are looking for?
18:46Or would you like help
with some suggestions?
18:49WALLY BRILL: I usually
need help with suggestions.
18:53KAI HALEY: So in addition to
creating that brand persona,
18:56you're going to want to
align it to the user value
18:59that you're offering.
19:01A user value matrix is
a really helpful method
19:05for mapping where your
conversational task is
19:08going to fit onto this matrix.
19:10It can really range from
something very simple,
19:13to something complex, or
utilitarian, to entertaining.
19:17And we saw some of those
examples earlier today.
19:20They fell in various
areas on this spectrum.
19:24And you'll really be
thinking about how
19:26does my product fit in here?
19:28And where am I going
to want to be focusing
19:30my conversational task?
19:33And then it's
always good to align
19:35the team on what does success
look like for this product?
19:38And how are we
going to measure it?
19:40So you can identify, what are
the success metrics that you're
19:43going to be collecting
later, once you've
19:45launched this product?
19:48Now, you're ready to move into
the sketch and decide phase.
19:51And this is where we use
traditional design sprint
19:54methods like crazy 8's
sketching to create
19:57a broad range of ideas,
assumptions, and sprint
20:00questions to make sure that
we're testing the right thing,
20:04and deciding on what the
conversational task will be.
20:10So for those of you who are not
familiar with crazy 8's, this
20:13is a fun method
from gamestorming
20:16where the entire team
sketches, and tries to generate
20:20eight ideas in eight minutes.
20:22It's a very fast exercise.
20:24But that added
pressure helps people
20:26to push beyond their first
idea and get more creative.
20:33Then the team will each
individually present
20:36their ideas to help
everyone understand
20:39what are the ideas that
they might want to pursue.
20:43Each person gets about three
minutes to present their ideas.
20:46And then they move into the
assumptions exercise, which we
20:52highly recommend you do this.
20:53This is where you take a look
at all the ideas generated
20:56and identify, what
are the unknowns?
20:58What are the
assumptions that you're
21:00making around what
people will do,
21:02what behavior people
will engage with?
21:05And translate those
into questions.
21:06Because a design
sprint is really
21:08a tool to answer a
question and get learnings.
21:11And this will then
help to determine
21:12which ones do you want
to prototype in order
21:18Then, as a team, you'll decide
which one of these tasks
21:21are we going to test?
21:23And you're really
limited by the length
21:25of time of the user test.
21:28A traditional usability
test is about 45 minutes.
21:31So you'll look at,
how long is this task?
21:34You might be able to
prototype three tasks
21:36if they're relatively short.
21:38But you'll work back
from that limitation.
21:42And now we move
into the role play,
21:44iterate, and decide
phases, where
21:46you're really going to be
doing the conversation design.
21:49We use improv warmups,
multiple rounds of role play,
21:54and then critique and
conversation repair at the end.
22:00WALLY BRILL: Let me talk
to you about some of this.
22:02So we use methods from improv
to workshop the dialogue.
22:07What we're trying
to do, what we want
22:09to come out with
out of this exercise
22:12is what we call a sample dialog.
22:15A sample dialog is like a
play with two characters.
22:20And there's an action.
22:22The robot and the user.
22:24And so what we're
going to do is we're
22:26going to have people
play those parts.
22:29And we're going to
have the conversation
22:32and try and get to that
dialog that works best.
22:37So what we'll do is we'll
sit people in chairs
22:39back to back because
we don't want
22:41them to use any of the micro
expression cues that we have
22:45when we're having a
conversation face to face.
22:48You know when it's
your turn to talk
22:49because you can see that I'm
finishing my sentence, right?
22:54So we don't want those cues.
22:56So we sit people back to
back to simulate a voice
23:01What we need to know is that
the brand persona comes through
23:06clearly, that character
that we've created.
23:09We want to know that we're
creating a clear path forward
23:13at every dialogue
turn for the user.
23:16We want to be natural.
23:17And we want to be brief.
23:19But we want to make sure
the persona flavor still
23:23We want to ask easy questions.
23:25We don't want to
overload anybody.
23:27And what's really
fun is, say you
23:30were creating an action for
somebody to use in the car.
23:34You might pick a book up and
turn it into a steering wheel,
23:37and be driving
while you're having
23:39the conversation with
the persona that's
23:45The next thing we'll do is
we'll critique with the team.
23:49So we're looking at whether
the introduction works.
23:52We're looking at if the brand
persona comes across exactly
23:58Is the conversational
natural and realistic?
24:01We know from research
that the more natural
24:04the conversation is,
the more successful
24:06the user is going to be,
because they don't have
24:08to learn anything new, right?
24:10They know how to
have a conversation.
24:13And we want to know that
there's a clear call to action
24:15or a question at the end
of every dialog turn.
24:17Every time the Assistant
says something,
24:21there'll be a call to action
at the end of what it says.
24:25Now, we'll do that again.
24:27Once we've had
that first review,
24:29we'll give the notes
to those folks who
24:32are performing the conversation,
and they'll do it again.
24:35And this time they'll
capture it on Post-It notes,
24:38and they'll put
it up on a chart.
24:41Because by the time
they've done that, we
24:43should have a pretty cogent
and complete conversation
24:46that we could prototype from.
24:50The next piece we
want to talk about--
24:52and we'll give a
lightning talk around this
24:55generally-- is about
contextual repair.
24:58I like to call it when bad
things happen to good people.
25:03You're having a conversation
with the Assistant.
25:05Everything's hunky dory.
25:06Everybody's
understanding everybody.
25:08And all of a sudden
the dog barks.
25:12Oh, I got your attention
with the dog bark.
25:17When the dog barks,
the system doesn't
25:19know what it just heard.
25:20It has no idea what
dog bark means.
25:23And so it throws what
we would call an error.
25:26And we need to repair that.
25:28We need to get back on track.
25:29So there are mechanisms
to do that with.
25:34And that's what we
talk about in terms
25:35of the conversational
contextual repair.
25:40KAI HALEY: So now we're ready
to move into prototyping.
25:43We've got a very close to
finish conversation flow.
25:47And you're going to decide
what kind of prototyping
25:50you want to use, whether you're
doing voice recording/Wizard
25:52of Oz or Dialog Flow.
25:54And then you'll prepare your
final script for the interview.
25:59So at this point,
you're going to want
26:01to incorporate that
contextual repair
26:04piece into the conversation,
document it very clearly so you
26:09have something to hand
off to begin prototyping,
26:12and really think about the
task that you're asking users
26:15to complete from an
end-to-end perspective,
26:18focusing on the user test.
26:20What are you going to bring
into that usability test?
26:24And here is a little example of
how you might be editing along
26:30This is from the
Filtrete sprint.
26:32And they were really
looking at what's
26:34the right level of
technical detail
26:37we should be providing
to people when they're
26:39asking about air quality?
26:41And can we get them interested
in indoor air quality?
26:45If you decide to do a
dialog flow prototype,
26:49you'll be creating
this with the text
26:52to speech computer
generated voice.
26:54And you're able to
prototype something
26:56very quickly in real code
that you can play on a device.
27:00So there are some
benefits there.
27:02You're a little closer to an
actual launchable product,
27:05but it is harder to adjust
on the fly in the user test.
27:10The alternative is to
use voice recording
27:13and create a Wizard
of Oz prototype.
27:16If, when you defined
your brand persona,
27:18you decided you wanted
to use a human voice,
27:21then you'll use this
type of prototyping.
27:24You'll have to record
all the clips that you've
27:26laid out in that flow.
27:28So you'll have your voice
actor there with you
27:30and your recording equipment.
27:32And you'll want to consider,
what are the edge cases?
27:34What are the potential
responses your user test
27:37participant might give you?
27:39You'll number and
arrange them very easily
27:41for retrieval in accordance
with your user interview script.
27:46And then you're going
to want to think about,
27:48how do I set this up so you have
the person behind the curtain
27:52playing these recordings.
27:55And along the way, with both
the dialog flow and the Wizard
27:58of Oz prototype, you're going
to write your user interview
28:02You want to set the
context for the user,
28:04so they understand
what is the task
28:06that they're going to complete.
28:08Write open-ended
interview questions
28:10so you get really
good, useful feedback.
28:13And then align the
prototype to this script.
28:15Make sure there isn't
a point at which you
28:17don't have a voice clip for
a specific point in the task.
28:21And review it with the
team for any potential gaps
28:24before you go into
that usability test.
28:28And now, you're ready for
the most exciting part
28:30of the sprint, where you get to
actually see people interacting
28:34You're going to set up
the test environment,
28:36run the usability test,
and gain your learnings,
28:39synthesize and evaluate
those learnings as a team,
28:42and then identify next steps.
28:46So setting up the
environment for good feedback
28:50means that you really want to
make it as real as possible.
28:53And in a lot of
these situations,
28:55you want people to actually
act out completing a task.
28:59And if it's a
hands-free type of task,
29:02you want to make them
feel like they're
29:03comfortable acting this out.
29:05So you're really trying to
create a real experience.
29:09Make them comfortable, but
also iterate as you go.
29:12If you find that it isn't
working quite right,
29:17And while you're
running this usability
29:18test in your usability
lab, the team
29:21is in another room observing
and collecting the feedback.
29:25A user feedback matrix like
this is really helpful.
29:28You can align the
participants' feedback
29:31to the specific questions that
you wanted to have answered.
29:35In a voice action sprint
it's really important
29:37to decouple the
feedback that you're
29:39getting about,
say, the platform,
29:41or interacting in a voice format
from the actual conversational
29:46Because here what you're
trying to do is find out,
29:49is this task going to
be useful for them?
29:52Is it going to provide value?
29:53Are they going to engage
with it and do it?
29:56So you'll want to review
against your questions.
29:59And you'll want to synthesize
those findings as a team
30:03so that you can decide
what to prototype,
30:06or what to move
forward in the future.
30:09A good example is from
the Filtrete sprint.
30:12Through the user test,
they were able to learn
30:15what the right
level of information
30:17is to address indoor air quality
questions, outdoor air quality
30:23And they were able
to engage people
30:25in a conversation about
indoor air quality
30:29through the outdoor
air quality test.
30:33And they defined a vision
for their connected home air
30:37filtration products.
30:41So what happens after the
sprint is equally as important
30:44as what happens in the sprint.
30:47I highly recommend that people
document all the learnings
30:50and everything that's
created in the sprint
30:52during the sprint itself
so you can walk away
30:54with a deck for
stakeholders' approval.
30:57You might need to schedule
follow-up testing.
30:59You'll have learned
something in your test,
31:01and you might want to
iterate and test again.
31:03And, of course,
you're going to want
31:05to identify, what's your
timeline for development,
31:07and create a product roadmap
to help you get to launch.
31:11And so we hope
you'll try this out.
31:13WALLY BRILL: Do
try this at home.
31:18KAI HALEY: We have
some resources for you.
31:19There's some lightning
talks on YouTube.
31:21You can go to the Google
Developers channel
31:23and look for those
there, and bring them
31:25into the understand
phase of the sprint.
31:28And we have our website,
Designsprintkit.withgoogle.com,
31:32where there are
templates and decks
31:34that you can download to use.
31:36And if you'd like to
learn more about building
31:39your first action, there's a
talk this afternoon at 5:00 PM.
31:43And tomorrow there'll
be a talk on 10
31:46best practices for
high quality actions.
31:49And you won't want to miss the
Designing Quality Conversations
31:53on Thursday afternoon.
31:55That's particularly a great
one, cause our own Cathy
31:58Pearl is doing that one.
31:59KAI HALEY: Fantastic.
32:02WALLY BRILL: And with
that, thank you very much.
32:05You've been wonderful.
32:06KAI HALEY: Thank you.
