>> Welcome to the HIVE.
My name is Andrew Woods.
I'm the manager of this new facility.
The HIVE stands for the Hub for Immersive Visualisation and eResearch
and it's a new facility in the university intended to support and encourage visualisation
and virtualization, simulation aspects related to all things visual.
The primary aim is to encourage research outputs in this field and it's seen as a field
of great opportunity and great potential for enabling us to do things a lot better.
So as I mentioned it's a new facility.
We launched on the 27th of November and we're still in the process
of commissioning so there's still a lot to do.
I might just provide you with a quick rundown of the four displays we have here.
These are sort of the most obvious part of the HIVE the four visualisation systems.
Each of them have their own characteristics and allow visualisation
to be done in a range of different ways.
The first of the displays is what is known as the Tiled Display.
It's a media wall.
It's another way of describing it.
It's a ten square metre area of LCD panels bigger than some people's swimming pools
and over that area we have 24 million pixels.
The wall is made up of 12 full-HD LCD panels, media grey panels with a very small border
around the sides so they can be shown together very closely.
There's a whole manner of different types of content that can be shown on here.
We don't have enough time to go through all of those different options right at this moment.
The next display on your right is called the Cylinder.
It's a three metre high screen, eight metre diameter curved screen when you're standing
in the central point it fills 180 degree of your field of view.
This can run in stereoscopic 3D using the provided 3D glasses over there.
The screens are filled with three projectors which are mounted from the ceiling there
and this can be used for a range of different topics and tasks including virtual environments
which we've got here which we've got here illustrating a project
on the HMAS Sydney which we're working on.
The screen behind you is called the Wedge.
It's two rear projected screens mounted at 90 degrees to each other.
Those screens can be angled outwards to form a flat screen as well, 2.8 metre diagonal,
full-HD resolution on each display and can also be run in stereoscopic 3D so we're anticipating
that screen there would be used for visualisation of volumetric data
or business data for example or viewing of stereoscopic content.
There's a demo loop there.
There's a set of a whole heap of glasses sitting on the small podium there so please,
once this session's finished please come through and have a look at that screen as well.
Each of the screens the glasses only work
on that particular screen so these two are stereoscopic.
These two aren't.
The screen on your left here is known as the Dome or it's actually a half Dome,
4 metre diameter and when you stand at the apex of the screen it fills your full peripheral
and primary vision so when you come in here don't stand too close.
We don't want you walking off the screen for example.
The purpose of today's session is to talk
about data visualisation primarily looking at business data sets.
We've invited ThoughtWorks on the campus to present this topic.
I want to be clear that we're not endorsing their work.
They're just providing an illustration of what is possible.
They have some very good illustrations of visualisation and what it is capable of.
They also have done some range of work with some groups on campus as well.
So I will now hand over to David and also give you my microphone.
>> Thank you Andrew and first baton pass of the afternoon.
So as Andrew said we're here today to talk about some organisations we've worked with
and the benefits that they've realised through data visualisation and we'll also talk a bit
about the approach that we've tended to use there which is quite a lightweight
and rapid approach to reaching a visualisation.
So before we go any further I'll just make sure I introduce the two of us properly.
I'm David Colls and my colleague--
>> Ray Grasso.
>> And ThoughtWorks is a software product development consultancy so we're based here
in Perth but ThoughtWorks is a global group.
So I guess the key thing is visualisation is used for a purpose and what sort
of problems do people solve with visualisation?
Well one of the earliest examples takes us back to Victorian London and the middle
of the cholera outbreak and some of you might be familiar with this.
In the 1850s over 600 people died in the cholera outbreak and John Snow was seeking
to understand why and to do that he drew a picture from the data.
He collected data about where the deaths had occurred and for each house
or each location along the street where there was a death he drew a black line to indicate
where people had been dying and as he put this picture together it quickly became clear
that there was a problem near the Broad Street water pump
and using this visualisation he was able to convince the authorities to remove the handle
of that water pump and was able to convince them
that that was the source of the cholera infection.
And this was before there was even a mechanism that could be understood
for transmitting cholera from a water pump.
It would be another seven years before Louis Pasteur introduced the theory of germs
and then the prevailing wisdom of the time was spread by miasma or bad air.
So this was a pretty good outcome for a drawing based on some numbers
and he was subsequently able to convince the authorities that dumping raw sewage
into the public water supply was a bad idea, a great legacy of public hygiene
from that particular data visualisation.
But if we come back to the present day then we see that all sorts
of organisations are using visualisation for all sorts of purposes, corporations, governments,
NGOs, formal and informal alliances of all of those then individuals participating
in hackathons and similar activities trying to combine data sets and produce meaning
from them especially where they are complex and difficult to understand.
And why are people trying to visualise this data?
Well, it's because it has a number of advantages for us as human beings.
Visualised data is easy to understand because it engages our innate cognitive mechanisms.
It's also the case that once you've understood a data visualisation you can use it
to pursue further investigation, so once you know what it's showing you can use it
to interrogate whether it's showing you what you expect.
It's a shared view.
So with a visualisation that's shared among a group the entire group can focus its energies
on making that picture better rather than disagreeing
on what the picture should look like.
It's a holistic summary you might say that nothing gives you the big picture
like the AV picture and it also provides new insight so you might set
out to create a visualisation that will show you certain things
but you can be pretty much guaranteed that you'll see new things in the process
of producing that visualisation and for those reasons,
for the reasons of visualisation it's good for people.
It's good for business because business runs on people.
But in particular we've seen with our clients that they're looking at two major outcomes
from data visualisation and one of those is to increase engagement
to produce compelling communications or build brand awareness with an audience
that might be external or it might be internal or it might be a combination of both.
And their also using visualisation to try and gain insight into their operations
to have a very lightweight approach to drilling into complex data and seeing what leaps
out as opportunities for improvement.
But to talk about engagements we'll hand the baton over to Ray.
>> So as Dave mentioned kind of one of the themes I guess that we've found with some
of our clients reasons they're using data visualisation is around increasing engagement.
In the first case we'll look at is increasing engagement by telling a story over a complex set
of data what is inherently a complex story trying to tell that in a really simple fashion
and so the client we're speaking about are the Independent Market Operator or the IMO
and for those who don't know who they are, the IMO, one of their responsibilities is to operate
and develop the wholesale electricity market of Western Australia so big producers
of energy and big purchases of energy.
They operate in these various markets trading sums of energy
and the IMO job is to facilitate that.
When we first spent time with the IMO sort of discussing the data in their world,
I mean they send in a lot of different data across these different markets
and they have a lot of different stakeholders both in the government and in the public
and within the industry itself and the kind of approach they had taken up to
that point presenting this data was kind of this style of presentation.
This is something we pulled straight from their website when we first got there.
It's sort of a static graph showing a bunch of figures and forecast lines and things like that
for the people sort of within the market there's probably a certain level engagement
with the raw numbers but for a lot of the average people they sort of see this
and eyes kind of glaze over and they sort of move on.
So I'd just like to cover the approach that we went through with the IMO
and one of the specific pieces of the project that we that we did with them
in taking this data and trying to get the engagement up and increasing it
with their audience and this is sort of the broad sweeps of the process and it was sort
of a broad question that started the whole project really then there was a data discovery
portion where you really dig into the numbers to actually see if it backs up the proposition
and then really rapidly refining a solution to communicate a visualisation over this data.
And the broad question that started the overall project really was just a hunch
and it was actually a conversation with the CEO of the IMO
and he basically effectively said some of the big players
in the market are doing some weird things with their trades.
It's unusual.
See if we can tell that story and show that story to people so you kind of take a step back
from that and go, well that's pretty broad like what does that mean?
And so the first reaction and the first step that we would generally recommend is to sort
of dig into the data to try and see what this story is if this is actually a story.
So that's what we did.
We grabbed a lot of the information straight from their core systems,
threw it out into really simple files like CSV files
that we could look at in Excel and stuff like that.
This is sort of where we started so we kind of graphed out all the different participants.
Each of these coloured lines are a different generator or retailer.
The top there it shows their total volume of trading in this particular market
and this one here shows who sold energy into the market and the bottom one shows
who bought energy from the market.
So you look at that and it's still really noisy and really doesn't say much
so the exploration kind of continues and so these are sort
of snapshots along the process that we took.
So here you can sort of see now we've broken it out for an individual participant
so there's two participants there so that the green bit
above is how much energy they're selling into the market.
The red below the line is how much they're buying.
These are quantities and this is sort of over time since the inception of the market.
So you start to get a bit more of a sense of who's doing what.
It's a bit easier to sort of dissect but the story still wasn't quite there so we sort
of split that graph out even further.
So now we've got it broken down by the hour so you can sort of see each trading interval
as sort of roughly an hour and the volumes that they're trading at that time
and the top there is the largest generator so particularly you would expect the largest seller
into the market and here is the largest retailer so effectively here we expect
to buy the most energy out of the market.
And there's something interesting about this.
Now it won't be immediately interesting to you guys generally but you can sort of see here
from about midway through 2012 the largest generator of power started buying lots and lots
of power and the largest retailer of power started selling lots and lots of power
and they we're doing it overnight.
So basically from 10 o'clock in the evening until 7 in the morning and if you look
at these two they effectively reflect each other so this sort
of high level visualisation you can get a sense of what was going on
and so at this point we knew that okay there is some interesting trading behaviour here
that would be useful to actually illustrate.
We kind of worked out that there is a narrative behind this and so at that point is
where we really switched gears into that final sort of solution part of the process.
Here it's really about shifting into the communication of realm rather than the data sort
of digging around, so you have a story so how do you effectively communicate that story?
So you start with paper, lots of brainstorming and sketching and ideas
like how can we show these different players and the volumes
that they're contributing or taking from the market?
Once we sort of had a general direction we wanted to go
in then we really went into sort of solution realm.
So the channel here that we're working with was the Web so via the public website
so we were working in HTML, Java script, CSS, typical web technologies and our part was just
to really with the IMO was to build a solution and integrate it as we go rather than kind
of Photoshop it up and imagine it as an idea and then come to implementation work
out that it wouldn't actually be implementable.
In fact that's the approach we took and this is a sort of snapshot of different milestones
or points along the evolution so we kind of had this sense of the market in the centre
and the different participants sort of around the edges and the different volumes of energy
that they were contributing or taking from the market as sort of flowing out from there.
And so as we started to evolve that that kind of resonated with a lot
of the folks we were talking to but we realised that we needed to kind
of anchor it in time a little bit more.
Just having a single snapshot was difficult for people to know where they were
so we added this timeline type presentation on the left which you sort of saw from before.
So here you can sort of get a high level snapshot view over time and you can sort of drag
in and see the individual trading volumes and so what we finally end up with come
over to the right side over here was this presentation piece in the middle,
so here we have this kind of selector or this little window, this monthly window
over the duration of the market and on the right you can see that the market in the middle
and the different volumes from the different participants around the edge, so for instance,
in September, you can see the largest generator there
at the top buying a really large amount of energy out of the market.
And the largest retailer there, that was almost magical, wasn't it-- selling it.
Let's see if I can scroll this up a bit.
There we go.
So there's a lot going on.
There's a lot of the information within this single visualisation and it's a theme
that you'll see again when we get into the portion that Dave was speaking about.
Once you get sort of, once you can navigate this, once you understand the layout
of what you're seeing you can actually assume a lot of data really quickly and you can sort
of interact with it and explore it in a way that just a table of figures
and a static graph just doesn't allow you to do.
And so if we come back that was again an example where one
of their clients were using data visualisation to try to get this information
out to their different stakeholders, getting more engagement with people.
Second case is really around amplifying engagement so again,
this whole idea of using data visualisation to get people closer
to the data and to the organisation.
And this client is basically the folks from the Desert Fireball Network,
the fireballs in the sky if you haven't heard of them based here in Curtin.
Phil is right there from the DFN.
Hi Phil. [Inaudible] So for those of you who don't know what the DFN is about
and Phil don't throw stuff at me if I butcher this.
Meteors are flying through the sky and they sometimes explode into fireballs
and the Desert Fireball Network is basically a network
of cameras throughout the Australian Outback poised
on the sky taking long exposure photographs continuously and then Phil
and his team afterwards take those photographs and use the fireballs in them and the positions
of the cameras to basically get the trajectory of these fireballs to work
out if they maybe hit the ground and maybe where they've come from and that can lead
to expeditions where you're actually trying to recover the meteorites or calculating
where they come from and the reason geologists will do this is because this kind
of information can help make us understand more about the origins of the universe.
It's pretty grand stuff.
It's pretty inspirational.
You can see-- is that running with the live one?
But there was different shots there so it's like a time lapse of some of the camera images
in this spherical screen over there, so it's really interesting stuff.
So when we first sat down with the Fireballs team it was really to talk about a brief
for a Smartphone app so really grand ideas and now we're talking about a Smartphone app.
The Smartphone app the main idea of it was really about bringing the general public sort
of closer to the science and sort of an outreach engagement awareness piece
and there's a really interesting piece to this where it was
about actually engaging citizens in the science itself.
So one of the core features of the application you can see as described here
and it really was this case where someone's out say, somewhere in Perth,
and in the night sky they see a fireball.
They pull out their phone with the app.
They tap where it started.
They tap where it ended.
They record a bit of information about it and that information goes off to the DFN folks
and you get enough of these folks seeing the same fireball it could help actually contribute
data to basically complement what was already in the DFN.
So this core sort of flow is sort of pretty simple and you can sort
of see there fireballs are a pretty spectacular thing but the information
that describes it is actually fairly mundane.
I mean you've got elevation.
You have azimuth.
You have with the duration in seconds, really a lot of numbers effectively
and so as we were sort of exploring this when we were working it
out if someone actually sees this how are they going to describe it?
Are they going to be saying oh yeah, it's like a minus four magnitude of brightness?
It was to 12 degree elevation here and it was about 5.2 seconds long.
Like that just wasn't realistic.
What you're more likely to get is someone that will say something like that.
Like it broke up and it was kind of green and it went in this direction and so that core idea
of can we put something together where people are interacting with this information?
This is really about authoring information.
Authoring data kind of coalesced in this first version
where you would basically build your own fireball.
So this would be part of the capturing so once you've tracked where it was you would come
in there and you would set the colour of it.
You would set the shape and how many pieces and all this kind of good stuff
and this animation would update, does update.
I'll go to my phone, I'll show you afterwards, in place and you can see that
and it was a much more engaging way to get people
to actually give you the information rather than going through a long boring form.
So after that first milestone, that first release we kind
of asked the question can we go better in other part of the application?
In two other parts of the application that we looked at was really
that capture piece and the sightings or viewings.
So the capture piece as you can see is really just a big button so it was
like it started there and it ended there.
It was quite static and the sighting piece
like showing a sighting was really just showing the textural information.
It was just like a static map of where they saw it.
So we did do better with the great help of the folks at the DFN and fireballs in the sky
and so the capture process became this sort of live, heads up display
and then an actual star map behind it, so you can see there these dots here are a star map
and based on where the person is sitting and the orientation of the device it would overlay
in this sort of augmented reality way the stars that they should be seeing behind there.
It became a much more immersive experience.
It's something that was actually just a stand-alone part of the app.
I'll sort of wave it in front of you right now so folks if you can see that you can sort
of see it's sort of live updating.
You can sort of see the horizon.
There's all the stars there and you've got the heads up display and as you sort of go
to capture it draws out the path like that and then you get into your building of the fireball.
So it's a much more immersive experience than just tapping those two dots.
And then on the sighting screen itself you know rather
than just showing the aesthetic attributes on the right there's also this animated fireball
over the star map as it would have been captured, so again, here you've sort of both
on the consumption side and almost the production side you're really trying
to up the engagement and get people to engage with the app
in a way that they otherwise might not.
So that's the two examples sort of within the whole sort of business goal
of organisational goal of increasing engagement.
I'll hand it back over to Dave to talk about operational insight.
>> Thanks Ray.
So I guess the next main case that we'll be looking at is gaining operational insight
and I guess here you might summarise that as trying to pick out something that sticks
out like a sore thumb as a starting point.
And something that stuck out like a sore thumb to John Snow was the fact that nobody died
in the brewery which was just down the street from the water pump but that was
because all the monks in there spent all day drinking beer.
How times have changed.
This case is about when you need a big picture view to kind of spot what doesn't look right.
And for this case we go to a call centre and a call centre
where there's a big improvement programme and the performance of the call centre is measured
in terms of customer satisfaction which is assessed by NPS or net provider score
and by how much it costs to run the call centre or operational expense.
So those are the two key measures of the performance of the call centre
from the organisation's perspective.
The improvement programme is looking to achieve a balanced improvement in both of those.
But it's a big call centre.
It's really big.
It's 200 thousand calls a day.
They're dealt with by 10,000 agents who are across multiple countries and time zones
and there are 500 or more products.
No one is really sure how many products are actually supported by the call centre.
It's a bit like that.
It operates 24-hours a day and seven days a week and this presents the challenge.
We're trying to improve this enormous call centre
but we have trouble picturing what it looks like now because it's so big and so diverse
that we can't really get a good handle on how to start the improvement process
or even what looks particularly wrong about it that needs to be improved.
We have some metrics around Q sizes and around wait times and we have some levers
that we can pull and push but it's not really going to--
that sort of small scale fiddling isn't going to really achieve the objectives of this programme.
So in this case we set out to draw a picture of the call centre and much like Ray showed
with the evolution of the short-term energy market this was an evolving process
and it's quite fascinating looking into that.
But we're going to skip right to the end product in this case and this is the end product.
This is the picture we drew of a call centre.
It might not immediately strike you as a call centre but maybe when I talk you
through it you might see the reasoning behind it.
This is actually a point in time in the call centre, so this is about ten o'clock
in the morning and it shows all of the calls that are active in the call centre at this time.
So there are about 3,000 calls active in the call centre and each one of those is represented
by a character on the screen, one of the catatonic characters.
The calls can be in basically one of two states.
They can be in a queue, so you've just called up.
You've entered a few things with your voice or with the phone keypad
and now you're listening to hold music.
In that case then you're in queue and that would mean you're
in the orange section at the top of the screen.
The longer you stay in the queue the further down that orange section your call progresses.
When the call is answered and it's a customer talking to an agent and it's
in the green section lower down the screen
and again the longer the call has progressed the further down the screen the call moves.
Different types of calls or inquiries by customers showing from left to right
across the screen so that the horizontal position determines the nature of the call.
So on top of this there are some things that we can also show that we know both upset customers
and lead to increased operational expense.
The thing that upsets customers is and they hang up,
the one thing that upsets customers is hanging up in the queue so we're showing calls that hang
up in the queue with an exploding bubble.
So those are where customers have hung up and another thing
that upsets customers is being transferred from one agent to another because typically you have
to go back into the queue again and that's shown with diagonal lines that go from a conversation
with an agent back up to a queue for a different call type.
So that's a picture of just a slice of time in the middle of the morning at the call centre.
But that's not even the biggest picture I guess, that's still a small picture.
What we can do is actually come over here to the cylindrical display
and we can watch the whole day in the call centre unfold.
So we go back in time a little bit to start at about 7 o'clock in the morning now
in the call centre and we're running a lot faster than realtime
so we're running 128 times real speed and we could see all of the things that we saw
in the static image over there but I'll just talk you through them again where it's live.
So when we set out to draw this picture
of the call centre these are the things we expected to see.
We expected to see calls arriving.
In this case these are bill inquiries arriving in the queue for bill enquiry calls.
We expected to see them progressing through that queue and as you can see they're moving
down the screen there as they progress and we can also see the abandons
where customers are hanging up in the queue up there.
When an agent is available then the call is transferred through to that agent
and we can see calls being answered by agents in the lower part of the screen.
Again these are all billing enquiry calls as they progress through their lifecycle
and again we're moving through the lifecycle of that call progressing down the screen.
So we can see billing inquiries there and we can see a different type of call over here.
We can see fault reports coming in over here and right over on the side over here which some
of you might be able to see we can inquiries about fixed line moves
so we can also see transfers in and out of a particular call type.
So here's sales inquiries are presumably going to other areas that relate to the product rather
than a general sales enquiry and other calls are resulting in a sales enquiry of some sort coming
into this queue, so those are the transfers.
Those are the things we expected to see but then there were a bunch of things we didn't set
out expecting to see but we saw once we looked at the visualisation.
And the first of those was that some types of demand that we would have expected
in the call centre just were not existent.
We would have expected people were calling about iPhones but there is zero demand
in that 200,000 set of calls, the iPhones which is quite an unusual finding and it's not one
that we can necessarily answer with this visualisation
but at least we've found an interesting question to ask,
to understand how we can improve operations.
Then to actually answer that question we need to go somewhere else and we need to dig
into the data and we need to use different tools and a different approach
so this is very powerful in identifying what that should be,
the question that needs further investigation.
Another question that might occur that probably doesn't need
so much investigation is why no one's calling about Blackberries.
That one is probably a bit easier to explain.
We also found again over here on this side of the screen that there were calls coming
in for business inquiries but there were no agents there to handle them.
So that's not really a good outcome from a customer experience
or even from a business point of view that all of those customers are just hanging
up in frustration before they get to speak to anyone.
So another thing that we saw was kind of subtle
but once you've seen it then you can't help noticing it and that is
that these queues are supposed to be fairly orderly.
The call that has been in the queue the longest is supposed to be the one that's answered first
but what we can see here is that there are calls that are spending a long time in the queue
but despite that there are new calls being answered by agents before the calls are taken
out of the queue so you can see calls rapidly falling down in green past the orange calls
that are moving very slowly in the queue and that was an interesting finding
because that's not actually how it works in reality.
It turned out that there was an issue with the way the data was being processed,
so there was a large amount of processing going on between the source systems
and actually a single view of this data of the call centre
and in that processing there were errors that were making it look
like calls were jumping the queue when in fact they were not.
So but the rub here is that this same data was being used to make operational decisions
in a magnitude of millions of dollars a year so it's very important to ensure
that it's quality data and a visualisation like this helps us uncover where there are issues
that we might not otherwise find.
And the final one was an interesting finding around transfers so we might expect that
and here customers are calling with a help request.
We might expect that calls are transferred if customers call with two different types
of inquiries in which case it's a legitimate transfer once the first enquiry is completed.
It might be that the automated system that classifies calls gets it wrong sometimes
which it does and in that case an agent should quickly realise and there should be a transfer
from high up the talking area into a different queue type which is the actual call enquiry
but what we didn't expect to see was calls being transferred within the same type.
We didn't expect to see calls being sent to agents who later said actually I can't deal
with this but it has been classified correctly and so this is great opportunity
to improve the performance of the call centre to eliminate unnecessary transfers
and to eliminate unnecessary cost by ensuring the calls only get to agents
that can deal with them properly.
And I could talk about this all day and we will leave it up once we're finished
but I'll hand it back over to Ray for now to bring us on home.
And that means another baton change.
>> Okay, so we'll just wrap up with a few take aways I guess.
There's learnings that we've had in the course of the work we've been able to do
with our clients and a few of those pieces we'd like to just leave you with now.
So we've covered the benefits.
That's probably worth just quickly talking about them again,
so there is data visualisation can be really useful in presenting complex information simply,
sort of efficiency of understanding is another way of thinking about that I guess.
It can be used to make interacting with data something that's almost exciting,
a bit more visceral, so whether it's like the fireballs instance where the capturing
of the data or the actual reading of data, the consuming can be a much more engaging experience
and then for the call centre example using these high level sort of holistic,
fuzzy visualisations of really complex sets of data can help sort of lead you
to ask more questions where you can do more pointed investigations
and take more pointed actions.
So what are the kinds of triggers you might find in your day-to-day work I guess
within whatever organisation that you're a part of that might be a twig for you
to think okay may be this is an instance where data visualisation could be used?
We kind of covered this a little bit but when there is a complex story
to tell there's complex data behind it but you have the data.
The data is there.
If there is not a shared picture certainly
within an operational sense but you have the data.
The data is there again.
You have the data but it's boring to you.
It's boring to look at or it's boring to create.
Maybe there's an opportunity there for using these kind of techniques.
And finally when it comes to execution, so when it comes to building these kinds
of things the approaches that we generally recommend are starting small and both that's
within the scale of which you're trying to achieve and within the teams
that you have and stay really lightweight.
Use real data throughout so there will be a lot of truisms that people will hold and even people
with a lot of experience can be proved wrong when you actually get into the information,
so definitely use real data as early as possible and use it throughout.
And refine and adapt, so none of these visualisations
that we've seen the endpoint wasn't clear from day one.
It was really an evolution and an exploration.
I mean all of the solutions we've seen they're all custom software, custom created
and they all took on the order of weeks to complete.
So with that I guess I'll stand back up and say thank you
and if you have any questions we'd be happy to here them.
[Applause]
[ Silence ]
Không có nhận xét nào:
Đăng nhận xét