The science of data visualization

00:54:55
https://www.youtube.com/watch?v=l7cAdp0f4X0

Résumé

TLDRIn his presentation, Larry Silverstein explores the science of data visualization, sharing insights from his extensive experience in the field. He emphasizes the importance of creating effective visualizations that facilitate understanding and decision-making. Through various examples, he highlights common pitfalls, such as the misuse of pie charts and 3D graphics, advocating for the use of bar charts and clear, concise designs. Silverstein introduces key concepts like pre-attentive attributes, memory limits, and the significance of color in visualizations. The session aims to equip attendees with practical tips to enhance their data visualizations, ensuring they effectively communicate their intended messages to the audience.

A retenir

  • 📊 Bar charts are more effective than pie charts for comparisons.
  • 🎨 Color should be used thoughtfully to avoid confusion.
  • 🧠 The human brain can only hold about six numbers at a time.
  • ⏱️ The five-second test helps ensure clarity in visualizations.
  • 🔍 Pre-attentive attributes can enhance quick understanding.
  • 📈 Bullet charts provide context for performance metrics.
  • 🖥️ Dashboards should balance exploratory and explanatory elements.
  • 📚 Continuous learning and feedback improve visualization skills.
  • 💡 Beautiful design enhances user engagement and understanding.
  • 🔗 Resources and training are available for improving data visualization skills.

Chronologie

  • 00:00:00 - 00:05:00

    Larry Silverstein introduces the session on the science of data visualization, sharing his background and a story about a poorly designed dashboard for a car company that ultimately failed to meet user needs. He emphasizes the importance of effective visualization in conveying information clearly and engagingly.

  • 00:05:00 - 00:10:00

    The presentation begins with examples of ineffective visualizations, such as a 3D pie chart that misrepresents data. Silverstein advocates for bar charts as a more effective means of comparison and highlights the limitations of crosstab reports for executives needing quick insights.

  • 00:10:00 - 00:15:00

    Silverstein discusses the importance of visual cues in data visualization, demonstrating how color can enhance understanding. He presents a game to illustrate how visual cues can help people quickly identify information, emphasizing the need for clarity in visual data representation.

  • 00:15:00 - 00:20:00

    The speaker introduces the concept of memory limits in data visualization, explaining how visual representations can help users process information more effectively than raw data tables. He provides examples of improved visualizations that make it easier to identify trends and insights.

  • 00:20:00 - 00:25:00

    Silverstein critiques a complex pie chart and suggests that simpler bar charts are more effective for comparison. He discusses the importance of intuitive encoding in visualizations and how to avoid common pitfalls that can confuse viewers.

  • 00:25:00 - 00:30:00

    The presentation covers the impact of interruptions on data comprehension, using real-world examples to illustrate how distractions can hinder understanding. Silverstein emphasizes the need for clear navigation in dashboards to maintain user context.

  • 00:30:00 - 00:35:00

    The speaker discusses color usage in visualizations, warning against the use of emphasis colors that can mislead viewers. He highlights the importance of considering colorblindness and cultural differences in color interpretation when designing visualizations.

  • 00:35:00 - 00:40:00

    Silverstein explains the hierarchy of data types and how to effectively use pre-attentive attributes in visualizations. He emphasizes the importance of positioning, color, and size in creating effective visual representations of data.

  • 00:40:00 - 00:45:00

    The presentation shifts to discussing different chart types and their appropriate use cases, including the effectiveness of bar charts over pie charts. Silverstein provides tips for creating impactful visualizations that convey information clearly and accurately.

  • 00:45:00 - 00:54:55

    In the final segment, Silverstein discusses the importance of aesthetics in data visualization, arguing that beautiful designs can enhance user experience and engagement. He encourages attendees to seek feedback on their visualizations and to invest time in creating visually appealing and effective data representations.

Afficher plus

Carte mentale

Vidéo Q&R

  • What is the main focus of Larry Silverstein's presentation?

    The main focus is on the science of data visualization and practical tips for creating effective visualizations.

  • What are pre-attentive attributes?

    Pre-attentive attributes are visual elements that stand out before we consciously pay attention, such as size, color, and orientation.

  • Why are bar charts preferred over pie charts?

    Bar charts are preferred because they allow for easier comparison of values, while pie charts can be misleading and hard to interpret.

  • What is the five-second test in data visualization?

    The five-second test assesses whether viewers can quickly understand the main message of a visualization within five seconds.

  • How can color affect data visualization?

    Color can influence perception and understanding; it's important to use colors thoughtfully to avoid confusion and misinterpretation.

  • What is a bullet chart?

    A bullet chart is a variation of a bar chart that provides context by showing performance against a target and qualitative ranges.

  • What should be considered when designing dashboards?

    Dashboards should balance exploratory and explanatory elements, ensuring they guide users to insights while being visually appealing.

  • What resources does Larry recommend for improving data visualization skills?

    He recommends books on data visualization and training courses offered by Tableau.

  • What is the significance of beautiful design in data visualization?

    Beautiful design can enhance user experience and make visualizations more engaging, leading to better understanding and usability.

  • How can one ensure their visualizations are effective?

    By seeking feedback, applying best practices, and focusing on clarity and simplicity in design.

Voir plus de résumés vidéo

Accédez instantanément à des résumés vidéo gratuits sur YouTube grâce à l'IA !
Sous-titres
en
Défilement automatique:
  • 00:00:00
    hello everybody and welcome so you made
  • 00:00:09
    it through TC almost almost thanks for
  • 00:00:14
    spending an hour with me this is the
  • 00:00:18
    science of data visualization
  • 00:00:21
    my name is Larry Silverstein I'm a
  • 00:00:23
    strategic sales consultant at tableau
  • 00:00:26
    I've been with tableau for five years
  • 00:00:29
    and twelve years before that I was with
  • 00:00:33
    another business intelligence company
  • 00:00:34
    and honestly I did some regrettable
  • 00:00:38
    things back then
  • 00:00:39
    don't get nervous from a visualization
  • 00:00:42
    perspective I'll give you a story so we
  • 00:00:47
    were working with a car company and we
  • 00:00:51
    thought it would be really cool to make
  • 00:00:53
    a dashboard that looks like a dashboard
  • 00:00:57
    with gauges and dials and meters and all
  • 00:01:03
    other kinds of embellishments and we
  • 00:01:07
    showed it to the executives and guess
  • 00:01:09
    what they loved it for about a day
  • 00:01:15
    because even though it had that coolness
  • 00:01:18
    effect the first time they looked at it
  • 00:01:21
    over time became cumbersome to look at
  • 00:01:25
    there wasn't a lot of information on it
  • 00:01:27
    because of the clunkiness of those large
  • 00:01:30
    gauges and dials so adoption suffered so
  • 00:01:35
    over time I came to tableau and I really
  • 00:01:39
    started to embrace the science of data
  • 00:01:41
    visualization and I'm happy to share
  • 00:01:44
    some of my journey with you today and
  • 00:01:46
    what I've learned so today's session is
  • 00:01:51
    kind of a twofer the first part is the
  • 00:01:54
    science of data visualization but I
  • 00:01:56
    don't want this to just be a TED talk
  • 00:01:59
    where it's you walk out with a feeling
  • 00:02:02
    of well that was really cool I also want
  • 00:02:05
    you to walk away with some practical
  • 00:02:07
    ideas that you could apply and in fact
  • 00:02:09
    as you're watching this presentation
  • 00:02:13
    I challenge you to think about a
  • 00:02:14
    visualization that you might have done
  • 00:02:16
    where when you showed it to people maybe
  • 00:02:18
    you didn't get the result you wanted
  • 00:02:20
    meaning you know maybe people didn't
  • 00:02:22
    understand what you were trying to say
  • 00:02:24
    because my goal is to make you make your
  • 00:02:27
    viewers go from that face on the left
  • 00:02:29
    where people are confused to the right
  • 00:02:32
    where they I have that aha moment and
  • 00:02:34
    you might see some slides you'd like and
  • 00:02:37
    I'd like to remind you you probably know
  • 00:02:38
    by now these slides will be available to
  • 00:02:41
    you on the TC live website after
  • 00:02:43
    conference so let's get into the science
  • 00:02:48
    of data visualization and we're gonna
  • 00:02:50
    start out with some warm-up exercises
  • 00:02:52
    before I get all scientific eye on you
  • 00:02:55
    so here's an exam a real example plucked
  • 00:02:58
    from the web oh really i popping graphic
  • 00:03:02
    wouldn't you say a 3d pie chart but look
  • 00:03:07
    at it for a moment you'll see that there
  • 00:03:09
    are some real problems with this
  • 00:03:11
    visualization now there are times when
  • 00:03:15
    you might want to do something as
  • 00:03:16
    eye-popping don't get me wrong
  • 00:03:17
    maybe you're a blogger and you want
  • 00:03:20
    people to be drawn into your website but
  • 00:03:24
    when you do stuff like this you might
  • 00:03:26
    lose credibility because if you look at
  • 00:03:28
    this you see for example America's is up
  • 00:03:32
    higher than Africa but Americas is 11%
  • 00:03:36
    in Africa's 18 that doesn't seem right
  • 00:03:38
    or compare Americas to China 11% to 13%
  • 00:03:43
    I don't know America's looks bigger to
  • 00:03:45
    me so I don't want to get too far ahead
  • 00:03:50
    of myself but one of the things you'll
  • 00:03:52
    hear throughout this presentation is
  • 00:03:54
    really the the bar chart is your friend
  • 00:03:57
    in many cases when you want to make
  • 00:03:59
    comparisons so here you can quickly see
  • 00:04:02
    for example that you know India is about
  • 00:04:05
    twice as much as China in terms of
  • 00:04:07
    growth and it's sorted so you can easily
  • 00:04:10
    make those comparisons and here's an
  • 00:04:15
    example that you're probably all more
  • 00:04:17
    familiar with the crosstab report it's
  • 00:04:20
    great when you need to look up specific
  • 00:04:22
    information but let's say you gave it to
  • 00:04:25
    an executive and
  • 00:04:26
    job was to figure out which is their
  • 00:04:29
    least profitable subcategory they could
  • 00:04:34
    do it but it's gonna take a long time so
  • 00:04:37
    they may not bother this is not going to
  • 00:04:39
    be very effective now I'm gonna give you
  • 00:04:42
    a visual cue to make it a little bit
  • 00:04:44
    better so now we see the negative values
  • 00:04:50
    in red so it's a cue and we're gonna
  • 00:04:52
    talk about which cues work better than
  • 00:04:55
    others
  • 00:04:55
    but this isn't perfect yet not that any
  • 00:04:59
    of us can ever be perfect but the
  • 00:05:00
    problem is is that you still have to do
  • 00:05:03
    some mental arithmetic you got to hold
  • 00:05:05
    those values in your memory and if there
  • 00:05:07
    were more rows and columns it would
  • 00:05:09
    become an even more complicated task so
  • 00:05:13
    here's a more delightful version now
  • 00:05:17
    we're where we've made it more visual
  • 00:05:20
    now you don't even need the numbers but
  • 00:05:22
    you can tell by the color of the bars
  • 00:05:24
    and the orientation of the bars that
  • 00:05:26
    it's those darn tables so anybody who
  • 00:05:29
    used a superstore knows there's always a
  • 00:05:31
    problem with tables so there you go you
  • 00:05:35
    don't really need the numbers so just to
  • 00:05:39
    get you warmed up here let's play a game
  • 00:05:42
    called count the nines shout out how
  • 00:05:44
    many nines are in here somebody already
  • 00:05:47
    saw this visible you happen to be right
  • 00:05:49
    but I'll pretend I didn't hear the
  • 00:05:50
    answer but if I then give you a visual
  • 00:05:55
    cue the answer that we heard a moment
  • 00:05:57
    ago shout it out if you've counted it up
  • 00:05:59
    how many nines are there I hear 10
  • 00:06:03
    that's correct
  • 00:06:04
    it's pretty obvious well you made it red
  • 00:06:07
    and made it easier but we're gonna talk
  • 00:06:08
    a little bit further about why that is
  • 00:06:12
    another game we're gonna play where it
  • 00:06:15
    is a little more complicated let's say
  • 00:06:17
    you're a Sales Director for the entire
  • 00:06:20
    country and somebody gives you this
  • 00:06:22
    report this looks like a typical report
  • 00:06:25
    lots of numbers and let's say that it's
  • 00:06:28
    X is your store sales and millions Y is
  • 00:06:32
    your store profitability and millions
  • 00:06:34
    and one two three four across the top
  • 00:06:36
    are your regions north south
  • 00:06:39
    east-west whatever all right
  • 00:06:42
    sales directors for the country you've
  • 00:06:44
    got this great data what's your next
  • 00:06:47
    move I know I know you saying yourself
  • 00:06:52
    come on be fair Larry give me some stats
  • 00:06:55
    give me means variances correlation
  • 00:06:59
    coefficients all right I'll give it to
  • 00:07:01
    you almost exactly the same in every
  • 00:07:07
    case sometimes to within several decimal
  • 00:07:11
    places
  • 00:07:11
    now watch your move still hard to tell
  • 00:07:15
    anybody know what this special data set
  • 00:07:17
    is called my last audience was smarter
  • 00:07:23
    than you guys I'm just kidding it's
  • 00:07:25
    called ants comes quartet so Frances
  • 00:07:27
    ants go home was a famous status
  • 00:07:29
    statistician from the 1970s and he
  • 00:07:33
    constructed this data set to prove a
  • 00:07:35
    couple of things first is that to truly
  • 00:07:40
    understand your data it's really
  • 00:07:43
    impactful to visualize it and that's
  • 00:07:45
    what we're all about
  • 00:07:46
    data visualization he also wanted to
  • 00:07:49
    show the impact of outliers on an
  • 00:07:51
    overall data set and that's why we got
  • 00:07:53
    some of those funky results before where
  • 00:07:56
    they all seemed exactly the same but
  • 00:07:58
    let's say now you're that Sales Director
  • 00:07:59
    and you're given this visual report you
  • 00:08:03
    might look at the lower left-hand corner
  • 00:08:05
    and say whoa there's that one store
  • 00:08:08
    that's doing really well let's find out
  • 00:08:11
    what they're doing and drive up the
  • 00:08:13
    other is you could do something
  • 00:08:14
    actionable same thing with the lower
  • 00:08:17
    right hand corner forget that that hat
  • 00:08:19
    wire this way way out there but the ones
  • 00:08:21
    above the line you might want to talk to
  • 00:08:22
    those stores and figure out what's
  • 00:08:24
    better and what if you can encode more
  • 00:08:27
    data into this visualization such as the
  • 00:08:30
    size of the circle might be a different
  • 00:08:33
    measure like you're discounting and you
  • 00:08:36
    might want to use color to represent
  • 00:08:37
    your product categories and show more
  • 00:08:39
    information well it's kind of what
  • 00:08:41
    tableau is all about right so let's get
  • 00:08:46
    into the science a little bit what we
  • 00:08:48
    want to do
  • 00:08:50
    in an effective visualization is to get
  • 00:08:53
    people to use what's called the visual
  • 00:08:55
    cortex that's the part of the brain that
  • 00:08:58
    allows you to quickly see things now
  • 00:09:00
    there is time to use the cerebral cortex
  • 00:09:03
    but remember we want people to look at
  • 00:09:06
    our visualizations and within five
  • 00:09:08
    seconds understand what it's about it's
  • 00:09:11
    not about deep thinking so let's figure
  • 00:09:13
    out how we can exploit that how many of
  • 00:09:19
    you have seen this slide or something
  • 00:09:21
    similar to it there are a number not too
  • 00:09:25
    big so these are called pre-attentive
  • 00:09:28
    attributes these are the things that
  • 00:09:32
    just like the name says before we really
  • 00:09:36
    pay attention to something it stands out
  • 00:09:39
    whether it's size orientation length
  • 00:09:42
    color and so on now we're gonna find out
  • 00:09:47
    a little bit later in the presentation
  • 00:09:48
    that some of them will bore powerful
  • 00:09:51
    than others but it kind of depends on
  • 00:09:52
    the situation so hold on to that but
  • 00:09:55
    these are the things we want to exploit
  • 00:09:59
    so let's touch on some of the other
  • 00:10:02
    facets of date of the science of data
  • 00:10:05
    visualization like we have memory limits
  • 00:10:08
    here's an example suppose I gave you
  • 00:10:11
    this table of data and asked you a
  • 00:10:13
    couple of questions have we gained or
  • 00:10:16
    lost customers over the last four years
  • 00:10:18
    well that first question is really easy
  • 00:10:21
    because I gave you a total line and you
  • 00:10:24
    can see 15 50 compared to 1779 it's gone
  • 00:10:27
    up we've gained great but if I ask which
  • 00:10:31
    city is growing the fastest that's a
  • 00:10:35
    little bit harder but what if I were to
  • 00:10:41
    give you a chart that's the same data
  • 00:10:44
    now it stands out Austin really has
  • 00:10:48
    improved the most right but why is it
  • 00:10:52
    that this is so much more effective it
  • 00:10:55
    probably seems obvious to you the idea
  • 00:10:59
    here is is that well the human brain can
  • 00:11:01
    really only hold
  • 00:11:03
    about six numbers you know you know
  • 00:11:05
    registers right about six but when I
  • 00:11:08
    gave you that table excluding the total
  • 00:11:10
    line that's sixteen values that's a lot
  • 00:11:13
    more than six but by creating that chart
  • 00:11:16
    I've chunked each of those rows into one
  • 00:11:19
    line and I encoded it with a color now
  • 00:11:22
    we can easily differentiate between
  • 00:11:25
    those patterns so let's talk about some
  • 00:11:31
    ways that we can overcome our memory
  • 00:11:35
    limitations so this was actually ripped
  • 00:11:42
    off the web it's a little cut off on the
  • 00:11:44
    screen here there's a website called vis
  • 00:11:46
    WTF I kid you not and it also said
  • 00:11:51
    underneath classic case of would be
  • 00:11:54
    better as a bar chart but the point I
  • 00:11:57
    want to make here is is that you know
  • 00:11:59
    the user meant well they're trying to
  • 00:12:01
    give you a lot of information but by
  • 00:12:03
    making a big pie chart with two circles
  • 00:12:07
    on the inside that it's no longer
  • 00:12:09
    something that we're familiar with and
  • 00:12:12
    the encoding is not really intuitive at
  • 00:12:15
    all for example we got those two things
  • 00:12:17
    in the middle those two circles total
  • 00:12:20
    Internet users and high-speed Internet
  • 00:12:22
    users circle with a circle as people we
  • 00:12:25
    have a hard time understanding sizes and
  • 00:12:29
    you know areas of circles
  • 00:12:31
    so is that outer circle really about
  • 00:12:33
    three times as big as the inner one hard
  • 00:12:36
    to tell and there are some other issues
  • 00:12:37
    here like all those labels pointing to
  • 00:12:40
    things and we have India we see the
  • 00:12:43
    number and then there's something it
  • 00:12:44
    says countries outside of the top 20
  • 00:12:46
    which one is bigger it's really hard to
  • 00:12:49
    tell so this is like a pseudo donut
  • 00:12:52
    chart but not exactly now you can build
  • 00:12:57
    this in tableau but it wouldn't be
  • 00:13:00
    following visual best practices so since
  • 00:13:06
    somebody had commented would be better
  • 00:13:07
    as a bar chart I wanted to prove that it
  • 00:13:10
    is and I wanted to show pretty much
  • 00:13:12
    exactly what this author was trying to
  • 00:13:15
    show so I have
  • 00:13:17
    a bar chart sorted but at the top I
  • 00:13:19
    represent those two big circles and then
  • 00:13:24
    at the bottom I have the countries
  • 00:13:26
    outside of the top 20 is a special case
  • 00:13:28
    and I use a different color but now I
  • 00:13:31
    can easily compare that to India which I
  • 00:13:34
    couldn't tell before I can see it's just
  • 00:13:36
    a smidge lower in terms of users
  • 00:13:43
    interruptions also slow us down
  • 00:13:45
    first I'm going to give you an example
  • 00:13:47
    from the real world and then one that is
  • 00:13:50
    from the viz world so I'd like you to
  • 00:13:53
    study this picture for a moment I'm
  • 00:13:55
    gonna block it out and I'm gonna make a
  • 00:13:57
    change and I want you to tell me what
  • 00:14:00
    has changed study here we go
  • 00:14:02
    interruption and we're back
  • 00:14:06
    anybody can anybody tell me what has
  • 00:14:08
    changed shadow he said nope the leaves
  • 00:14:16
    you happen to be right you're good more
  • 00:14:18
    either you've seen it before you're more
  • 00:14:20
    perceptive but very well done I mean
  • 00:14:22
    most people would know that subtle thing
  • 00:14:23
    right you see that all right really
  • 00:14:25
    subtle all right all right that was fun
  • 00:14:28
    but what's the what am I trying to get
  • 00:14:30
    to so very often I see these paradigms
  • 00:14:34
    where people make a dashboard that
  • 00:14:37
    drools from one to the from dashboard
  • 00:14:39
    one the dashboard - and by the way with
  • 00:14:41
    those new buttons that you saw a couple
  • 00:14:43
    days ago that's gonna make that really
  • 00:14:44
    easy but before we get there I'm gonna
  • 00:14:48
    start with the end in mind where we see
  • 00:14:50
    some product information and some order
  • 00:14:52
    details and the thing is is that when I
  • 00:14:55
    when I get here unless I really know my
  • 00:14:58
    product hierarchy well I might not know
  • 00:15:01
    what category is subcategory I started
  • 00:15:03
    at or what segment we started with so
  • 00:15:07
    now I'll show you the beginning just to
  • 00:15:08
    prove that point so here we are in a
  • 00:15:10
    typical drillable dashboard I'm gonna
  • 00:15:13
    drill into one thing into another you
  • 00:15:16
    know this is a common parrot it's a good
  • 00:15:18
    paradigm and then you have something a
  • 00:15:21
    link to hyperlink that says you'll go to
  • 00:15:22
    see that detail and then we get there
  • 00:15:25
    and that's the point I'm trying to make
  • 00:15:26
    we kind of lost something
  • 00:15:28
    we lost our content
  • 00:15:30
    as we jump from dashboard the dashboard
  • 00:15:32
    so you know where possible try not to
  • 00:15:36
    you know block out the user like that
  • 00:15:38
    try to give them the detail on the same
  • 00:15:41
    page if there's enough room to do so
  • 00:15:45
    alright let's get into some of the finer
  • 00:15:47
    detail around topics like color data
  • 00:15:51
    type chart types and layouts everybody
  • 00:15:53
    wants to know what's the right chart
  • 00:15:55
    type to use we're not going to be
  • 00:15:56
    dogmatic here but I'm gonna give you
  • 00:15:58
    some generalizations and we'll go from
  • 00:16:00
    there
  • 00:16:04
    so tableau comes with many many color
  • 00:16:09
    templates and you could create your own
  • 00:16:11
    but let's just break it down and we'll
  • 00:16:14
    just generalize here two types of
  • 00:16:17
    templates there are the more muted or
  • 00:16:20
    standard colors and emphasis colors and
  • 00:16:23
    I caution you against using emphasis
  • 00:16:25
    colors and and we see them all the time
  • 00:16:28
    the problem is well for one thing colors
  • 00:16:32
    mean different things to different
  • 00:16:34
    people especially when you think about
  • 00:16:36
    different countries so for example in
  • 00:16:39
    the United States red bad green good but
  • 00:16:43
    in China
  • 00:16:44
    red means good fortune so they're gonna
  • 00:16:47
    take something else away from that but
  • 00:16:52
    also remember that roughly 8% of men and
  • 00:16:56
    4/10 percent of women are what we
  • 00:16:59
    commonly call colorblind but really is
  • 00:17:03
    color vision vision deficiency or CVD so
  • 00:17:08
    here we can see the normal vision on the
  • 00:17:10
    left and then different OB is due to
  • 00:17:14
    Ranocchia
  • 00:17:14
    and so on depending on whether you have
  • 00:17:17
    trouble seeing green red or blue and
  • 00:17:20
    that happens to correspond to problems
  • 00:17:23
    with your medium long and short range
  • 00:17:26
    cones in your eyeballs and there are
  • 00:17:30
    websites out there that you could take
  • 00:17:33
    your viz and you submit it to that
  • 00:17:34
    website and it will actually tell you
  • 00:17:37
    this is what a person with let's say
  • 00:17:40
    protanopia this is how they
  • 00:17:42
    see it funny quick story I delivered the
  • 00:17:45
    same presentation a couple of days ago
  • 00:17:47
    and after I was done somebody came
  • 00:17:49
    rushing down to the podium and they was
  • 00:17:52
    really concerned they said go back to
  • 00:17:54
    your slides again on CVD I want to see
  • 00:17:56
    that again he thought he was going
  • 00:17:57
    colorblind it's all it's probably just
  • 00:18:01
    the monitor or something so don't get
  • 00:18:02
    nervous but please don't make any health
  • 00:18:05
    decisions about yourself based on this
  • 00:18:07
    slide so back to those emphasis colors
  • 00:18:13
    so here's the problem with using them so
  • 00:18:17
    on the chart on the left now I don't
  • 00:18:21
    know if 400,000 is a good thing like
  • 00:18:24
    sales or a bad thing like unemployment
  • 00:18:26
    but yet our brains are fooled even if
  • 00:18:29
    momentarily because we see Slovakia is
  • 00:18:32
    red at least in the United States that
  • 00:18:35
    would be our take away and that the
  • 00:18:37
    Czech Republic is green but there's
  • 00:18:39
    really no meaning attached to the color
  • 00:18:42
    so it slows people down and makes it
  • 00:18:45
    harder to understand on the right hand
  • 00:18:48
    side I'm using a neutral color I'm
  • 00:18:51
    telling the story effective they're
  • 00:18:53
    sorted the bars that's all you need to
  • 00:18:55
    know we call this problem double
  • 00:18:57
    encoding so try to stay away from it may
  • 00:19:01
    look pretty but like I said you're doing
  • 00:19:03
    yourself and your viewers a disservice
  • 00:19:07
    it's sometimes okay just be thoughtful
  • 00:19:09
    when applying color to bars so if you
  • 00:19:12
    wanted a group by you know a type or
  • 00:19:15
    something by a category that's fine
  • 00:19:17
    that'll work that could be useful and
  • 00:19:23
    take caution if you must use a
  • 00:19:26
    background to your vis so if I were to
  • 00:19:30
    ask you which one of those inner boxes
  • 00:19:34
    inner squares is darker how many of you
  • 00:19:38
    think the one on the left you know what
  • 00:19:39
    the answer is gonna be right haven't you
  • 00:19:41
    think it's the one on the left is the
  • 00:19:42
    darkest be honest alright some people
  • 00:19:46
    are honest other people are looking at
  • 00:19:49
    their how their plane is doing okay but
  • 00:19:51
    if you draw a square
  • 00:19:53
    they're all the same and we're gonna see
  • 00:19:55
    this problem pop up a little later when
  • 00:19:57
    we get to mapping and too much color it
  • 00:20:04
    can be another thing that slows people
  • 00:20:06
    down just like we can only remember
  • 00:20:09
    about six different numbers we can only
  • 00:20:12
    distinguish around eight colors so if
  • 00:20:14
    I've got a scatter plot my put States on
  • 00:20:16
    to color you're not gonna get a useful
  • 00:20:19
    pattern but if we keep it to eight or
  • 00:20:23
    less like we have here now patterns
  • 00:20:27
    emerge you see clusters this is useful
  • 00:20:31
    all right let's talk about different
  • 00:20:33
    types of data because we're gonna get to
  • 00:20:34
    how people like to see it so two types
  • 00:20:37
    of dimensions you know qualitative such
  • 00:20:40
    as names of states people beers and
  • 00:20:45
    ordinal qualitative data for example
  • 00:20:48
    metals and Olympics gold silver bronze
  • 00:20:51
    survey type data love it like it hate it
  • 00:20:54
    so on and then quantitative your numbers
  • 00:20:57
    your measures whether it's in dollars
  • 00:20:58
    pounds percentages or raw numbers and
  • 00:21:05
    again this is just a generality there
  • 00:21:08
    are times when you're gonna break the
  • 00:21:10
    rules but in general this is the
  • 00:21:12
    hierarchy people like to see things
  • 00:21:15
    first
  • 00:21:16
    remember those pretense of attributes
  • 00:21:17
    position then color then size and shape
  • 00:21:21
    I mentioned some are more powerful than
  • 00:21:22
    others here's your first clue to that
  • 00:21:27
    and this first part isn't so much about
  • 00:21:31
    tableau it's about the science but I did
  • 00:21:33
    want to say that that that biology and
  • 00:21:35
    psychology of our researchers went into
  • 00:21:39
    making tableau so it's no coincidence
  • 00:21:42
    that your columns and rows shelves as we
  • 00:21:46
    look top to bottom left to right that's
  • 00:21:49
    your most powerful thing that's number
  • 00:21:51
    one position the next thing is color
  • 00:21:54
    then size and then if you bring on more
  • 00:21:59
    things there is your shape it's built in
  • 00:22:04
    you don't have to think about it
  • 00:22:06
    and you also may be wondering how do our
  • 00:22:11
    eyeballs track to a screen let's say we
  • 00:22:15
    have a dashboard and we have four
  • 00:22:18
    sections where do our eyes go in general
  • 00:22:24
    your prime real estate is the top left
  • 00:22:27
    so it's got important information put it
  • 00:22:29
    up there if you've only only using a bit
  • 00:22:31
    of the screen you could put it right in
  • 00:22:32
    the middle that's also prime real estate
  • 00:22:35
    but like I said that's that's the
  • 00:22:37
    generalization so I don't know if you
  • 00:22:41
    saw in the data village there was some
  • 00:22:43
    eye tracking studies that you can take
  • 00:22:45
    advantage of here's something that came
  • 00:22:47
    out of the eye tracking study where they
  • 00:22:49
    try to show some of the things that go
  • 00:22:52
    against the generalizations one thing
  • 00:22:55
    they found is that in a dashboard and
  • 00:22:57
    we're playing in a moment our eyes were
  • 00:23:01
    drawn to big numbers and I mean big not
  • 00:23:04
    like this and billions but the fact that
  • 00:23:05
    the font itself is big and that it
  • 00:23:08
    happens earliest in the viewing sequence
  • 00:23:11
    especially the first time you look at
  • 00:23:13
    the VIS so I'm gonna play that so you
  • 00:23:19
    see it coming into focus those big
  • 00:23:22
    numbers kind of pop out earlier they
  • 00:23:28
    came to several other conclusions and
  • 00:23:30
    when you see the slides you'll see the
  • 00:23:32
    link if you want to look at some of the
  • 00:23:34
    other research they did okay
  • 00:23:44
    congratulations you passed the science
  • 00:23:46
    part of this session now for the tips
  • 00:23:52
    and tricks to help you apply what you
  • 00:23:55
    just saw a little while ago lots of
  • 00:24:00
    different chart types that tableau is
  • 00:24:02
    capable of making one is a table and it
  • 00:24:06
    could be useful sometimes especially
  • 00:24:07
    when you need to see specific
  • 00:24:09
    information like if this were a tax
  • 00:24:12
    table you had to look up a value or bus
  • 00:24:14
    schedule really useful but the magic of
  • 00:24:17
    tableau really is is that we find that
  • 00:24:20
    graphs are more powerful for spotting
  • 00:24:23
    trends so you can see playing in the
  • 00:24:25
    background as I look at again we're not
  • 00:24:28
    throwing all the numbers maybe this is
  • 00:24:30
    just an average but if I want the detail
  • 00:24:32
    I can use things like tooltips and last
  • 00:24:36
    year we or during this year we
  • 00:24:38
    introduced vism tooltip to show
  • 00:24:42
    additional detail but we can see the
  • 00:24:45
    trends first that's the thing that
  • 00:24:47
    stands out and here are some
  • 00:24:50
    generalizations if you've got something
  • 00:24:53
    that's based on time you should go on an
  • 00:24:56
    x-axis location on the map comparing
  • 00:24:59
    values you know I love the bar chart
  • 00:25:02
    it's more useful than most people give
  • 00:25:04
    it credit for and so on and maybe you
  • 00:25:09
    didn't realize this that show me
  • 00:25:12
    automatically enforces visual best
  • 00:25:14
    practices if I just click on one thing
  • 00:25:17
    and I go to show me the graph types that
  • 00:25:20
    you can use are actually enabled the one
  • 00:25:24
    that it recommends has an orange bar
  • 00:25:26
    around it if I click more items you'll
  • 00:25:29
    see that they start pot they light up
  • 00:25:31
    because now you can use them and with no
  • 00:25:34
    cost and almost no time you can find the
  • 00:25:38
    visualization that you think tells your
  • 00:25:41
    story the best you don't always have to
  • 00:25:43
    believe tableau with the orange bar
  • 00:25:45
    around it you choose
  • 00:25:51
    so you've probably heard overtime
  • 00:25:54
    debates over using PI's versus bars and
  • 00:25:59
    in fact the pie chart was not even
  • 00:26:03
    included in some of the earliest
  • 00:26:04
    versions of tableau it's not that our
  • 00:26:07
    engineer is warren smart enough to build
  • 00:26:09
    it we just felt that we didn't want you
  • 00:26:13
    to make a bad visualization because we
  • 00:26:15
    just don't think that it's a very
  • 00:26:17
    effective way to show data and I'll tell
  • 00:26:20
    you why if you were to look at the the
  • 00:26:22
    pie chart on the left and let's say we
  • 00:26:25
    want to compare how we're doing us
  • 00:26:27
    versus let's say competitor B even
  • 00:26:30
    though they're right next to each other
  • 00:26:31
    if I didn't give you the number it would
  • 00:26:34
    be very hard to tell just like I said
  • 00:26:36
    circle within the circle that we saw on
  • 00:26:38
    that really munch that monstrosity that
  • 00:26:42
    we saw earlier hard for us to really
  • 00:26:44
    precisely gauge sizes but if I take
  • 00:26:48
    those values in a bar chart and sort
  • 00:26:51
    them and now I don't even need to use
  • 00:26:52
    all those colors I'm just going to use
  • 00:26:54
    those neutral tones and just use a
  • 00:26:55
    darker gray so we stand out you can tell
  • 00:27:00
    even without giving the numbers that
  • 00:27:03
    competitor B is just a little bit ahead
  • 00:27:05
    of us but I've actually been doing some
  • 00:27:07
    reading lately and there are times in
  • 00:27:10
    cases where pipe charts are okay I can't
  • 00:27:13
    believe I'm saying this I gotta gather
  • 00:27:17
    myself okay some people like them
  • 00:27:21
    because they're my kind of soft looking
  • 00:27:23
    there's no axis right it's kind of
  • 00:27:26
    simple we see them every day the kind of
  • 00:27:28
    elegant they sometimes do well on a on a
  • 00:27:31
    map you could do pies on maps and
  • 00:27:33
    tableau and they work well maybe when
  • 00:27:36
    there's just a few slices certainly not
  • 00:27:38
    a lot and the donut chart is a variation
  • 00:27:42
    where you kind of take out the middle
  • 00:27:44
    and you might even think that's better
  • 00:27:46
    because now you're getting rid of those
  • 00:27:47
    sharp angles in the middle but use a bar
  • 00:27:51
    chart please but anyway I let you choose
  • 00:27:57
    stack bars are also a very effective way
  • 00:28:00
    to to show how things are gonna get
  • 00:28:02
    sliced for example let's say we're
  • 00:28:05
    looking at goal attainment here's a tip
  • 00:28:07
    I would recommend so let's say the three
  • 00:28:10
    different the three slices are whether
  • 00:28:14
    we've exceeded whether we've met or
  • 00:28:16
    missed are our goals the one that you
  • 00:28:20
    want to you want people you want to
  • 00:28:23
    stand out should be at the bottom
  • 00:28:25
    because the bottom is really the only
  • 00:28:26
    place where you can make meaningful
  • 00:28:28
    comparisons I mean take a look at it the
  • 00:28:31
    light gray starts and ends at different
  • 00:28:32
    places it's really hard to to measure
  • 00:28:35
    exactly same thing of course or even
  • 00:28:38
    more so with the dark gray items on the
  • 00:28:41
    top and another tip it could be that
  • 00:28:44
    well use an emphasis color I used red in
  • 00:28:47
    this case or when we missed our KPI and
  • 00:28:51
    then I used the neutral colors for the
  • 00:28:54
    others so they kind of fade in the
  • 00:28:56
    background and they also made a choice
  • 00:28:58
    that for any value above let's say 10% I
  • 00:29:03
    actually put the mark on the bar to
  • 00:29:05
    really call it out it's a little bit of
  • 00:29:08
    a story telling method so I started out
  • 00:29:15
    showing that car dashboard and I made
  • 00:29:20
    fun of the gauge you can see gauges
  • 00:29:23
    crossed out in the lower right hand
  • 00:29:24
    corner and yes if you go to tableau
  • 00:29:26
    public you could find people that have
  • 00:29:28
    made gauges if you really want one there
  • 00:29:31
    there be prepared for a lot of math
  • 00:29:34
    there was a lot of math because you're
  • 00:29:36
    actually doing the calculations to make
  • 00:29:38
    the circles and the radio whatever right
  • 00:29:41
    but Stephen few created and he's a big
  • 00:29:45
    name in the data visualization space I
  • 00:29:47
    came up with the idea of what's called a
  • 00:29:49
    bullet chart and it's much more
  • 00:29:50
    effective it gives you real context
  • 00:29:53
    there's a lot of information here so in
  • 00:29:56
    this bullet chart the thick black bar in
  • 00:29:59
    the middle I mean that's where you
  • 00:30:00
    currently are but then you could have a
  • 00:30:03
    comparative measure that little hash
  • 00:30:05
    line reference line if you want to call
  • 00:30:07
    it that
  • 00:30:08
    that could be how you did last year
  • 00:30:11
    where your goal is and then you can have
  • 00:30:14
    qualitative ranges like bad satisfactory
  • 00:30:17
    and good with those neutral tones and
  • 00:30:20
    then we give it really a very definitive
  • 00:30:23
    labeling so you know exactly what it is
  • 00:30:25
    you're looking at and these are really
  • 00:30:28
    effective for executive dashboards
  • 00:30:30
    because you can actually line up a bunch
  • 00:30:32
    of them and then you can see where
  • 00:30:34
    you're above and below your goals very
  • 00:30:36
    effective and as much as I love the bar
  • 00:30:42
    chart there are times when it's the
  • 00:30:44
    wrong chart type to use so if we're
  • 00:30:47
    looking at revenue over time broken down
  • 00:30:49
    by bookings and Billings the thickness
  • 00:30:53
    of all of the bars the height of the
  • 00:30:55
    bars obscures the pattern in the data
  • 00:30:59
    but when I use a trendline you can
  • 00:31:03
    really see how they track against each
  • 00:31:05
    other it's much clearer but with that
  • 00:31:08
    said that only works well with time not
  • 00:31:11
    other dimensions so I've talked about
  • 00:31:16
    pie charts donor charts getting hungry
  • 00:31:19
    yet here's another chart type the
  • 00:31:21
    spaghetti chart so I want you to try to
  • 00:31:25
    avoid the food graphs if possible and
  • 00:31:29
    here I took some data and I decided to
  • 00:31:32
    use brands of spaghetti for fun since
  • 00:31:34
    it's a spaghetti chart and don't make
  • 00:31:36
    any stock buying decisions based on this
  • 00:31:38
    this is completely made-up data but you
  • 00:31:43
    can see why it's called the spaghetti
  • 00:31:44
    chart it's the deed is all there but
  • 00:31:47
    it's really hard to see any trends or
  • 00:31:49
    patterns so what can you do about this
  • 00:31:52
    I'd like to offer you three
  • 00:31:55
    possibilities simplest one get rid of
  • 00:32:00
    the color and then use tableaus
  • 00:32:03
    highlighter who knows about the
  • 00:32:06
    highlighter few people yeah it's been
  • 00:32:10
    around for a couple of releases and as
  • 00:32:13
    long as it's a static visualization I'm
  • 00:32:16
    it's not static people can like choose
  • 00:32:19
    Barilla
  • 00:32:21
    and there you go it goes to the
  • 00:32:23
    forefront the others are grayed out
  • 00:32:25
    that's solution number one here's
  • 00:32:30
    another one we could create a
  • 00:32:36
    calculation let's say we wanted to focus
  • 00:32:37
    on Barilla so I'm again take off the
  • 00:32:41
    brand from color and I'm gonna take that
  • 00:32:43
    calculation and put that one on color
  • 00:32:46
    instead and now I mean you don't have to
  • 00:32:51
    but I'm gonna use some emphasis colors
  • 00:32:53
    I'm gonna use emphasis color for the one
  • 00:32:55
    I want to point out and I'm going to
  • 00:32:58
    neutralize all the others so this will
  • 00:33:01
    work more for a static kind of is as
  • 00:33:04
    opposed to highlighter but there's a
  • 00:33:06
    third solution which is really a lot
  • 00:33:07
    different which I'd like to introduce to
  • 00:33:09
    you and offer you where instead of
  • 00:33:14
    having everything in one you know all
  • 00:33:16
    the brands in one chart I'm gonna create
  • 00:33:19
    what's known as spark lines so each
  • 00:33:22
    brand is going to get their own pattern
  • 00:33:24
    and that really just comes down to some
  • 00:33:26
    formatting maybe I just want to turn on
  • 00:33:28
    the beginning and end values I'm gonna
  • 00:33:32
    get rid of some of the headers and been
  • 00:33:38
    off to watch all of it you end up with
  • 00:33:40
    something like this and if I wanted to I
  • 00:33:42
    could change the color for gorilla if
  • 00:33:44
    that was the important one but the idea
  • 00:33:46
    being now the patterns are no longer
  • 00:33:50
    hidden you get to see all of them at the
  • 00:33:52
    same time but point I'd like to make
  • 00:33:56
    about lines is is that even if the data
  • 00:34:00
    is time-based it's not always fair to
  • 00:34:05
    use it here's an example if you're
  • 00:34:08
    collecting toxin levels but it's at a
  • 00:34:12
    regular intervals and I plotted it like
  • 00:34:15
    this and connected the dots is this
  • 00:34:19
    really telling the story of the data now
  • 00:34:24
    I'm going to show you what that data
  • 00:34:26
    looks like
  • 00:34:28
    not really
  • 00:34:31
    resemble it doesn't really resemble the
  • 00:34:33
    real data does it so what are you doing
  • 00:34:36
    a case like this I'm not showing you the
  • 00:34:39
    only answer but what you might want to
  • 00:34:42
    do is use a dot plot instead the idea
  • 00:34:45
    being again on our brains won't
  • 00:34:48
    necessarily connect all these points so
  • 00:34:51
    it becomes more truthful using dual axis
  • 00:34:59
    is very popular and it's probably meant
  • 00:35:01
    more for an audience that is
  • 00:35:03
    sophisticated here's something from a
  • 00:35:07
    website of spurious correlations that
  • 00:35:10
    there's a lot of funny examples the
  • 00:35:12
    number of people who drowned by falling
  • 00:35:14
    into a pool correlates with films that
  • 00:35:17
    Nic Cage has appeared in all right maybe
  • 00:35:21
    he's a great actor maybe he's not but I
  • 00:35:23
    don't think he causes people to jump and
  • 00:35:25
    drown into pool drivin pools but the
  • 00:35:28
    problem with the dual axis is that for
  • 00:35:31
    one thing it makes your eyes dart back
  • 00:35:33
    and forth and that kind of slows us down
  • 00:35:35
    and then the worst part is especially if
  • 00:35:39
    you're not a sophisticated analyst you
  • 00:35:42
    might actually see this as having a
  • 00:35:44
    correlation but the problem here is that
  • 00:35:46
    the order of magnitude of drownings is
  • 00:35:48
    far greater than the number of movies
  • 00:35:51
    that Nic Cage has appeared in so you
  • 00:35:54
    might want to simply have them in
  • 00:35:55
    separate charts one on top of another
  • 00:35:58
    that might be an easy solution so if
  • 00:36:04
    you're a new tableau user you might be
  • 00:36:06
    wondering I've gotta find where in
  • 00:36:08
    tableau I can get a 3d visualization you
  • 00:36:13
    could look all you want you're not gonna
  • 00:36:15
    find it and there's a reason for it
  • 00:36:18
    because of all the science that goes
  • 00:36:21
    into making tableau we know that it's
  • 00:36:24
    suffers from the problem that we call
  • 00:36:27
    data occlusion meaning that data is
  • 00:36:31
    hidden so I'm going to try to stand
  • 00:36:34
    close to the edge here and look down to
  • 00:36:36
    try to find data for December of 1900
  • 00:36:41
    I can't look down and see well the
  • 00:36:44
    problem is it's a 3d representation on a
  • 00:36:47
    2d plane I just can't do it so somebody
  • 00:36:50
    might try to outsmart me hey Larry
  • 00:36:52
    suppose my software could spin the cube
  • 00:36:55
    that would do it and I would say nice
  • 00:36:58
    try but not exactly because the moment
  • 00:37:01
    you spend the cube guess what
  • 00:37:02
    you've now hidden the data on the other
  • 00:37:04
    side of the cube with tableau we say use
  • 00:37:09
    something called small multiples you
  • 00:37:11
    could see multiple dimensions at the
  • 00:37:14
    same time and see all the detail like we
  • 00:37:17
    show here using a tooltip in fact we're
  • 00:37:20
    showing more data the 3d chart is only
  • 00:37:23
    showing every 10 years our viz is
  • 00:37:27
    showing every year you can see patterns
  • 00:37:31
    in a v' is made with small multiples
  • 00:37:38
    alright let's talk about mapping now I
  • 00:37:41
    don't want you to come out of there and
  • 00:37:43
    said Larry told us not to use maps I'm
  • 00:37:46
    not saying that I'm just gonna say just
  • 00:37:48
    be careful maps are great when you have
  • 00:37:50
    location state city zip codes but
  • 00:37:52
    remember back to our hierarchy from a
  • 00:37:54
    little while ago the number one pretense
  • 00:37:58
    of attribute the most powerful is
  • 00:37:59
    position that's the row and column shelf
  • 00:38:03
    but guess what goes there when you have
  • 00:38:05
    a map latitude and longitude that's
  • 00:38:07
    spent so you can no longer have bar
  • 00:38:09
    charts and measure the length you're now
  • 00:38:11
    left with some weaker pretense of
  • 00:38:14
    attributes like size which we can get a
  • 00:38:19
    general idea of which is bigger but it's
  • 00:38:21
    not precise and color and you know it's
  • 00:38:25
    sometimes hard to exactly distinguish
  • 00:38:27
    colors but it is a very effective
  • 00:38:31
    visualization to use what I'm saying is
  • 00:38:35
    don't always assume just because you
  • 00:38:37
    have a geography you have to use a map
  • 00:38:39
    sometimes a bar or other chart type will
  • 00:38:42
    actually be better but please use maps
  • 00:38:46
    because I don't want to get in trouble
  • 00:38:48
    from management
  • 00:38:51
    we knew of one measure you could create
  • 00:38:55
    what's called a choropleth in tableau
  • 00:38:58
    that we use a more a simpler name a
  • 00:39:00
    filled map but this also suffers from a
  • 00:39:05
    problem we talked about before remember
  • 00:39:08
    this problem because like in this
  • 00:39:12
    example we see Texas is dark our eyes
  • 00:39:15
    might play some tricks on us
  • 00:39:17
    unintentionally about the things that
  • 00:39:19
    are around it you can use a filled map
  • 00:39:21
    it's sometimes okay for generalizations
  • 00:39:24
    now for two measures you're gonna use a
  • 00:39:26
    symbol map the size of the circle is
  • 00:39:28
    going to represent one measure the color
  • 00:39:31
    the second dimension it's a second
  • 00:39:33
    measure but even if you have one measure
  • 00:39:37
    you might want to still use the symbol
  • 00:39:40
    Mac it's a symbol map instead of the
  • 00:39:44
    choropleth map and gonna think of
  • 00:39:47
    Greenland they're our projection issues
  • 00:39:50
    with maps to green looks really big on a
  • 00:39:53
    map and when you put it in a color you
  • 00:39:56
    might start to think that it's more
  • 00:39:58
    important than it really is in your
  • 00:40:00
    overall map but if you put a symbol on
  • 00:40:04
    it and that symbol is teensy weensy
  • 00:40:06
    you'd come to realize all right you know
  • 00:40:09
    let's it sails not a lot of sales in
  • 00:40:10
    Greenland but the cool thing with maps
  • 00:40:15
    is you can actually use most any picture
  • 00:40:19
    that you could turn into an XY
  • 00:40:23
    coordinate so here's a visit of a
  • 00:40:25
    baseball perfect game showing where all
  • 00:40:30
    the pitches went and whether they were
  • 00:40:32
    strikes or put into play and so on so
  • 00:40:35
    you could do that with tableau take a
  • 00:40:37
    picture and map your dated XY
  • 00:40:39
    coordinates and your picture is
  • 00:40:41
    essentially a kind of map now this
  • 00:40:47
    session isn't intended to be about
  • 00:40:50
    storytelling but I do want to impart
  • 00:40:53
    upon you that as a visitor you have the
  • 00:40:56
    obligation to tell the truth with data
  • 00:40:59
    kind of like that toxin level vis I
  • 00:41:02
    showed before where
  • 00:41:03
    connected bats that's not really
  • 00:41:05
    truthful here's another viz taken from
  • 00:41:07
    the real world it's about regulation of
  • 00:41:11
    the cable industry and I'm not gonna
  • 00:41:13
    make any arguments about whether it was
  • 00:41:15
    good or bad I just want you to look at
  • 00:41:17
    this for a second and see if you spot
  • 00:41:19
    anything that it may be untruthful just
  • 00:41:22
    by looking at this the dates well that's
  • 00:41:28
    a great point for one thing the first
  • 00:41:30
    bar has four years the second bar is
  • 00:41:33
    five not apples to apples exactly and
  • 00:41:35
    it's not even telling us if it's taking
  • 00:41:39
    inflation into consideration now I'm
  • 00:41:42
    going to show you what really happened
  • 00:41:44
    so actually after regulation in 1992
  • 00:41:52
    there was great investment but then it
  • 00:41:55
    went down a little bit maybe because of
  • 00:41:58
    the financial collapse that time and
  • 00:42:01
    then it went up a lot due to the doct
  • 00:42:04
    come come bubble and then it actually
  • 00:42:07
    sank again so again I'm not trying to
  • 00:42:09
    argue the merits or demerits of you know
  • 00:42:12
    whether regulation is a good or bad
  • 00:42:15
    thing I'm saying in this case I don't
  • 00:42:17
    think the visitor really told the truth
  • 00:42:21
    with the data and that's an obligation
  • 00:42:22
    that you have next one I want to make is
  • 00:42:27
    it was great as tableaus this ql engine
  • 00:42:31
    is that's our secret sauce it's what
  • 00:42:33
    tells you what biz's are possible or not
  • 00:42:36
    possible with your data when you're
  • 00:42:38
    going to show me but you know the
  • 00:42:41
    default isn't always the best so if I'm
  • 00:42:43
    looking at drought data over time broken
  • 00:42:47
    down by States you know my default might
  • 00:42:49
    look something like this and this is a
  • 00:42:51
    variation of the spaghetti chart I
  • 00:42:54
    showed you a little while ago I'm trying
  • 00:42:58
    to say is it's so easy and fast in
  • 00:43:02
    tableau to just try different things it
  • 00:43:05
    may take a little training to do some of
  • 00:43:08
    the more advanced things but not a whole
  • 00:43:09
    lot you're not gonna break your data by
  • 00:43:13
    trying things so here's a much more
  • 00:43:15
    effective visualization
  • 00:43:17
    where it's really an array of maps where
  • 00:43:20
    we can see visually trends over time and
  • 00:43:24
    sometimes they cross over years you can
  • 00:43:26
    see where our drought might have started
  • 00:43:29
    and ended or maybe it was very regional
  • 00:43:32
    that's hard to pick up in a line chart
  • 00:43:35
    works well here
  • 00:43:41
    so far I've really been focusing on the
  • 00:43:44
    single vis and best practices but let's
  • 00:43:48
    touch on dashboards a little bit
  • 00:43:50
    specific best practices there here's a
  • 00:43:57
    rhetorical question
  • 00:43:58
    our old dashboards the same and what I
  • 00:44:01
    mean by that is you know are there
  • 00:44:02
    different types of dashboards of course
  • 00:44:05
    there's an infinite number of dashboards
  • 00:44:06
    you can make but let's say there are two
  • 00:44:09
    types and this this is the belief of
  • 00:44:11
    Andy Kirk a visualization expert he says
  • 00:44:14
    the two types are explanatory and
  • 00:44:17
    exploratory dashboards well start with
  • 00:44:21
    the exploratory dashboard here's an
  • 00:44:23
    example and we see these all the time
  • 00:44:25
    they're beautiful dashboards tableau
  • 00:44:28
    does them very well they help you
  • 00:44:30
    monitor your business we see the facts
  • 00:44:32
    along the top we see lots of trend lines
  • 00:44:35
    reference lines you drill filter it's
  • 00:44:38
    beautiful but it's neutral meaning when
  • 00:44:45
    it's effective it's really just begging
  • 00:44:47
    for you to click in it and find your own
  • 00:44:50
    truth right you're trying to monitor
  • 00:44:53
    your business it doesn't know ahead of
  • 00:44:55
    time where the problems are how many of
  • 00:45:00
    you have seen this visible or I'm just
  • 00:45:02
    curious it's been around a long time
  • 00:45:04
    half a dozen or so it's a brilliant
  • 00:45:08
    visits and very sad biz but it's it's
  • 00:45:11
    special in its characteristics
  • 00:45:13
    it's an explanatory visit which means it
  • 00:45:16
    has an opinion it's its goal is to make
  • 00:45:20
    you feel something or to take some sort
  • 00:45:24
    of action the title could send shivers
  • 00:45:29
    down your spine
  • 00:45:30
    Iraq's bloody toll now this visitor used
  • 00:45:34
    a non-standard type of is actually the
  • 00:45:38
    x-axis for time is along the top and
  • 00:45:41
    normally you might say well that's gonna
  • 00:45:43
    make it hard for people to understand
  • 00:45:44
    but in this case there's no doubt
  • 00:45:47
    especially due to the red dripping blood
  • 00:45:50
    right you know exactly what it's telling
  • 00:45:53
    but here's the interesting thing about
  • 00:45:55
    an explanatory or opinionated viz I can
  • 00:46:00
    take the same vis and tell another story
  • 00:46:01
    with it in fact I'm gonna turn it upside
  • 00:46:07
    down and put the x-axis where you'd
  • 00:46:09
    normally see it on the bottom and now
  • 00:46:14
    I'm gonna change the title for my Iraq's
  • 00:46:16
    bloody toll to Iraq deaths on the
  • 00:46:19
    decline and because I'm telling a a
  • 00:46:23
    better sir it's not a happy story but at
  • 00:46:25
    least it's good news I don't really need
  • 00:46:27
    the red color anymore I'm gonna make it
  • 00:46:29
    neutral you can't do that with an
  • 00:46:31
    exploratory dashboard that's unique to
  • 00:46:33
    something that's explanatory I hate to
  • 00:46:39
    give you the answer if it depends
  • 00:46:41
    for which ones you use but I would say
  • 00:46:43
    this most people focus on the dashboard
  • 00:46:47
    on the right wouldn't you say the ones
  • 00:46:49
    that are exploratory tab what makes it
  • 00:46:51
    really easy to drag in your your
  • 00:46:55
    different views and put it in filters
  • 00:46:57
    and use actions to drill down I think
  • 00:47:01
    the more subtle and nuanced skill and
  • 00:47:03
    the one I want to see you invest more
  • 00:47:06
    time in is the one on the left where you
  • 00:47:09
    can go to your management and say I
  • 00:47:13
    think we need to do this we need to
  • 00:47:16
    change our product mix or change our
  • 00:47:20
    discounting practices based on Veda I'll
  • 00:47:27
    get off my soapbox the idea is that no
  • 00:47:31
    matter which one you use you want to
  • 00:47:33
    make better data-driven decisions and
  • 00:47:36
    really just change the world for the
  • 00:47:38
    better whether your world is in business
  • 00:47:42
    private practice you're a blogger what
  • 00:47:45
    what have you all right now we get to
  • 00:47:49
    the crux of the presentation remember
  • 00:47:51
    the the phases of being confused to
  • 00:47:54
    saying aha
  • 00:47:55
    how do you make a viz to pass the
  • 00:47:58
    five-second test here's a great example
  • 00:48:00
    finding bigfoot' and I didn't know that
  • 00:48:02
    we were gonna see so much a sasquatch
  • 00:48:04
    the other day but here it is again but
  • 00:48:08
    this visit is fantastic the name is very
  • 00:48:10
    good it's very clear we see a map we
  • 00:48:13
    know that we're looking at sightings
  • 00:48:14
    broken down by season over time then we
  • 00:48:17
    get our factoids along the bottom and
  • 00:48:20
    here are some tips for making a viz for
  • 00:48:23
    the five-second test is we talked about
  • 00:48:26
    real estate most important things should
  • 00:48:28
    probably go upper-left didn't talk about
  • 00:48:31
    legend so much but put them near the
  • 00:48:33
    view and they shouldn't be really big
  • 00:48:35
    that's just way still out of space be
  • 00:48:38
    careful about your color schemes you
  • 00:48:40
    don't want to make people's eyes go
  • 00:48:42
    haywire with a lot of different color
  • 00:48:44
    schemes and the number of views four or
  • 00:48:48
    five maximum I mean you can go bet you
  • 00:48:50
    can go more I've seen good visitors with
  • 00:48:52
    10 views I've seen bad ones with two
  • 00:48:55
    views right so it depends and where you
  • 00:48:58
    can provide interactivity so people can
  • 00:49:01
    drill and find more information the next
  • 00:49:06
    one is about using your words and that's
  • 00:49:08
    a double-edged sword what i mean by that
  • 00:49:11
    is edward tufte hands who knows that
  • 00:49:13
    we're tough they were read books by
  • 00:49:14
    tough day very smart audience good to
  • 00:49:17
    see he came up this idea of dida ink
  • 00:49:23
    ratio which means as much as possible of
  • 00:49:28
    your real estate should be focused on
  • 00:49:31
    the data not embellishments so you're in
  • 00:49:35
    this case notice even grid lines are
  • 00:49:38
    barely visible if at all since this is
  • 00:49:44
    an executive dashboard it rather than
  • 00:49:47
    going down to the penny level we round
  • 00:49:49
    it to let's say six point four three
  • 00:49:52
    million total revenue
  • 00:49:55
    that's good enough and we use the letter
  • 00:49:56
    M to hide all the other values we talked
  • 00:50:01
    about using tooltips before that's
  • 00:50:03
    another way to show more information on
  • 00:50:05
    demand without spending a lot of that
  • 00:50:09
    data Inc you see some reference lines in
  • 00:50:13
    the lower right hand corner you might be
  • 00:50:14
    wondering what about that Larry that's a
  • 00:50:16
    good idea
  • 00:50:17
    reference lines are helpful because it
  • 00:50:19
    actually helps give you context for
  • 00:50:21
    their data now I'm going to touch on
  • 00:50:27
    some formatting and coloration this is a
  • 00:50:31
    viz by our own anti cat grieve you saw
  • 00:50:33
    him during re-envisioned maybe attended
  • 00:50:35
    one of his sessions he won an internal
  • 00:50:37
    iron vis contest with this submission
  • 00:50:41
    which is done kind of like in the style
  • 00:50:43
    of a I don't want to say it's a cartoon
  • 00:50:46
    but it's a panel a bunch of panels but
  • 00:50:49
    also notice that he's using mostly line
  • 00:50:53
    charts and line chart variations like
  • 00:50:56
    the upper right hand corner is called
  • 00:50:57
    the slope chart but in the middle and
  • 00:51:00
    he's not just doing it for the sake of
  • 00:51:02
    not using a line chart
  • 00:51:03
    he's showing he's experimenting with the
  • 00:51:07
    VIS and saying a heatmap of time is very
  • 00:51:11
    effective here because it really points
  • 00:51:13
    out winds and most Road fatalities occur
  • 00:51:16
    this is road fatality data January first
  • 00:51:19
    Christmas July 4th they all stand out
  • 00:51:22
    but also because Andy knows that if
  • 00:51:25
    you've got too many of the same vis it
  • 00:51:27
    creates kind of a visual exhaustion so
  • 00:51:30
    I'm not saying use five different
  • 00:51:31
    varieties of chart types just to show
  • 00:51:35
    how many chart types you can use but
  • 00:51:38
    when you have several on one page you
  • 00:51:40
    might want to mix it up a little bit and
  • 00:51:42
    finally notice the effective use of
  • 00:51:45
    color here it's bad news road fatality
  • 00:51:49
    so he's using red but he's also using a
  • 00:51:52
    very soft color to make that emphasis
  • 00:51:56
    color standout and when you're - warning
  • 00:52:02
    for the five-second test I want you to
  • 00:52:05
    get feedback because you're craving the
  • 00:52:08
    vis you feel somewhat married to it I
  • 00:52:11
    get that but if you show it to others
  • 00:52:14
    whether it's if it's in business you
  • 00:52:16
    show it's your colleagues if it's
  • 00:52:18
    something you could show to spouse
  • 00:52:20
    friends whatever even if people don't
  • 00:52:22
    know what business you're in they should
  • 00:52:25
    get the gist of it in five seconds or
  • 00:52:26
    less if not go back to the drawing board
  • 00:52:29
    you don't have to take every piece of
  • 00:52:32
    criticism you get and at some point on
  • 00:52:36
    the flip side publish the vis at some
  • 00:52:40
    point you've got to say it's good enough
  • 00:52:41
    but don't make that your final it is get
  • 00:52:46
    to get more feedback even afterwards
  • 00:52:47
    there's always time for a version - so
  • 00:52:52
    the last point I want to leave you with
  • 00:52:53
    is his beautiful design important that's
  • 00:52:56
    a question now if you're just doing
  • 00:52:58
    something for yourself to get a quick
  • 00:53:00
    answer to something visnu l in tableau
  • 00:53:04
    will give you a nice viz right out of
  • 00:53:06
    the box you don't have to format it
  • 00:53:08
    you're good to go right but I want to
  • 00:53:14
    talk about what we're looking at here is
  • 00:53:16
    this is method soap when it was first
  • 00:53:21
    released it actually had a leakage
  • 00:53:25
    problem but because of its
  • 00:53:28
    anthropomorphic design people decided it
  • 00:53:32
    was something they wanted to put out in
  • 00:53:35
    a place where people can see rather than
  • 00:53:37
    where so goes under the cabinet right so
  • 00:53:41
    people willing to overlook this defect
  • 00:53:44
    because it was so beautifully designed
  • 00:53:46
    so when it comes to business
  • 00:53:49
    scientific studies have shown that a viz
  • 00:53:52
    that is beautiful people will consider
  • 00:53:55
    it easier to use more delightful even if
  • 00:53:59
    it may not be so beautiful design is
  • 00:54:02
    actually important so spend time doing
  • 00:54:05
    it if you can so with that I'd like to
  • 00:54:09
    leave you with these resources here are
  • 00:54:11
    some great books that I've read to help
  • 00:54:13
    prepare for today's presentation in the
  • 00:54:16
    lower right hand corner you see
  • 00:54:18
    something about training so if you're
  • 00:54:19
    new to tableau I really do
  • 00:54:20
    urge you to go for desktop 1 training
  • 00:54:22
    and then this topic that we discussed
  • 00:54:25
    today is in the visual analysis class
  • 00:54:29
    which if you can't get out of the office
  • 00:54:31
    we actually offer virtually in two and a
  • 00:54:34
    half hour sessions over five days so I
  • 00:54:38
    thank you very much I know you all have
  • 00:54:41
    places to go like the airport please
  • 00:54:43
    take a moment if you would and please
  • 00:54:46
    fill out the survey I hope you had a
  • 00:54:48
    great tableau conference safe travels
  • 00:54:51
    see you next year yeah
  • 00:54:54
    [Applause]
Tags
  • data visualization
  • Tableau
  • Larry Silverstein
  • bar charts
  • pre-attentive attributes
  • color theory
  • dashboard design
  • effective communication
  • visual analysis
  • best practices