8 Rules for Better Data Storytelling | Get Your Insights Understood

All right, all right, all right. Hello. Hello everyone. It's great to be here. I am so excited to be hosting the session on eight Rules for better data storytelling. I already see quite a few people from all over the world joining us. Everyone do let us know where you're joining from in the chat. I'm going to be sharing here. Some of the folks sharing, we have Sherry M. From Wisconsin, we have Miracle. Where are you joining from? Let us know. Sidita from Albania, Miracle's joining from Nigeria. LinkedIn user here from Israel. We have Andres Tumedias from Peru. Great to see you. We have Amna Korshid from India. We have Juan Camilo village from Colombia. Awesome. Good to see people from all over the world seeing someone here from Mexico, Jose from Mexico. We have Michelle Evans from Ireland, Iqbal from Pakistan. People from all over the world, Levy from Zambia. We have you can see my cat here in the background we have Edward from Kenya, Chuse from Somalia. Connecticut. Connecticut. Connect. Connecticut, Connecticut. Okay, I'm gonna stop pronouncing it Connecticut, usa. Okay. We have Jamel here, Michael from Poland, Gene Dobre, Michael, Mike from Bolivia. Awesome to see people. We are seeing quite getting a lot of messages. People from Pakistan, Sri Lanka, Ivory Coast, Costa Rica, Canada, Las Vegas, Sin City here, Thesi from Sin City. And then here we have tuning in from Singapore, Brazil. So I think we're covering all the continents. We've seen South America, North America, Europe, the Middle East, Africa, Asia. Anything I missed? Do we have any Australians in the house here? Maybe we have. Hello from Gemma from Wales. Good to see you Jim. Actually Reese is from Wales as well. Awesome. Good to see people from all over the world joining us. I am so thrilled to be hosting the session on eight rules for better data storytelling. And you know, as we always say, the last mile of analytics is always the hardest one because this is where we drive real impact. And this is why today we're going to be talking about that last mile of analytics in in better data storytelling. So I'm going to go through my slides here but before just a bit about myself, a bit about Data Camp. Our mission is to democratize data and AI skills for everyone. So in case this is your first time interacting with Data camp we do this by helping both learners and organizations future proof their skills whether individuals organizations, we give them best in class learnings with access to market leading training with data Camp learn we help them build work ready skills with data Camp Data lab we help them grow their data team and their skills and cement their skills with data Camp. Certifications. And this is why more than 14 million learners and 4,000 organizations work with us to upskill and transform their data and AI skills. A bit about myself. My name is Idal. I'm VP of Media at datacamp. I'm a graduate in Economics. I told everyone. Where are you from? Right. You joining from. I'm joining from Belgium, but I'm originally Lebanese. I graduated from Economics at American University, Beirut. I did my master's in business analytics and data science from ISEC Business School in France. And I've been a data science educator and evangelist at Data Camp for the past five years. I also host the Data Frame podcast. So a lot of the insights that you'll see here come from the many experts that I've spoken to across the board. We have someone here from Brussels as well joining us, Henrietta. I'm really nearby, so I'm in Sanjeet right now in Brussels, and I'm very excited to be here. And we also have Gavin from Australia, so I think we've completed our continents, unless I'm missing a continent here. Good to see everyone here. And then Alexander from Ukraine, we really appreciate you joining. So with that, let's get started. We're going to talk about data storytelling, the last mile of analytics, right? We're going to talk about eight rules for better data storytelling. We're going to focus on four rules for better data visualization and four rules for better data for better data narratives. And then finally, we're going to talk about how to become a better data storyteller. So data storytelling, last mile of analytics. Why is data storytelling important? Right? And I'm going to rely here on the shoulders of giants, right? I mentioned that, um, a lot of the experts that I've spoken to on the Data Frame podcast on these webinars influence a lot of my thinking here when it comes to this topic. You know, Colin, as Baron McNaflik, author of Storytelling with Data, right, mentions that despite having all of the information in the world at our finger at our fingertips, it doesn't make it easier to communicate it. If anything, it makes it harder. The more information you're dealing with, the more difficult it is to filter. Andy Cotgrave, technical evangelist at Tableau, great friend of the show, mentions that data storytelling enables data teams to wield the power to frame arguments and persuade with data responsibly and deliberately. Gert, the Gator machine learning lead at Deloitte, to be impactful with data science, organizations need data scientists with stories. And then Brent Dykes, who's the Author of Effective Data Storytelling mentions that data stories are powerful vehicles for sharing data insights to influence and drive change within an organization. So all in all, data storytelling is an incredible tool at everyone's disposal here to drive change with data. So if we go from a data here, from ideation, if we go from, you know, any data science project, it starts with ideation and ends at acceptance, right? And the thing is that data storytelling is that last step, step before you reach acceptance. Because what you need here, even if you build the most, you know, incredible dashboard, most important, machine learning model, the most effective visualization, if your stakeholders are not able to internalize many of the takeaways that you provide, you're not going to be able to drive impact with data. So data storytelling is a methodology for communicating data insights with compelling narrative and visuals. It enables data scientists, data analysts, data practitioners of any kind of to drive change with their insights. So I love this definition here. Kind of the different components of data that are created by Brent Dykes and his book Effective Data Storytelling. Highly recommend that you guys check that book out. And data storytelling, in a nutshell, what it tries to do is combine narrative, data and visuals to explain, engage and enlighten with data so that we can drive change. Today, we're not going to be talking about data, we're going to be talking about visuals and narrative. So this is why we're going to start off by talking about eight rules for better data storytelling. And we're going to first talk about rules for better data visualizations. Rule number one, choose the best visualization for your story and we're going to talk about it here. There are many ways that you can cut a cake here. When it comes to data visualization. There are different plots that can take different shapes and forms that are applicable for different different use cases. It is always important to choose the right visualization for the right task. And I'm going to give here a few different examples. But the main role to think about is to always work backwards from your data. Don't start with what visualizations you want to show, think about what data you're trying to show and adapt your visualization to fit the data. For example, if you want to show comparisons of different categories, such as different banking products and associated customers, or branch cost data broken down into different verticals. Or if you want to visualize different customer segments, a bar chart is super useful here. For example, here, if I am a bank and if I want to visualize the number of customers by loan products, whether business loans, loans against property Commercial vehicle financing, construction equipment loan, etc. Etc. I can visualize those with a bar chart. If I want to show the changes of a variable over time, I would choose a line chart, for example, a stock price change over time. The number of app users over time, right? The number of customer support tickets over time. Here, for example, let's say these are investments that I've gained over time from an investment account. I wish this is what my investment account looked like, but this is a line chart. On the x axis we have a temporal, you know, we have dates. And on the Y axis we show the wealth accrued. We have our measure that we're trying to show. Now the third type of problem here that we're going to discuss is showing the distribution of a variable over time, right? So let's say for example, the distribution of savings for customer and saving accounts or app opens for all customers in a given time period, right? And here a great visualization would be something like histograms. So let's say I'm trying to visualize saving account size by number of customers, right? So for example, here around, I would say 25% of my customers have $20,000 or less, right? And 25% of my customers have around $100,000 or more. And when you use effectively, you're going to talk about labels here, when we talk about using labels, if you use something like a median, for example, where you say 50% of our customers have 80,000 or more saved up in savings account, this really helps you tell a story. Now the fourth one I want to show here is a relationship between two variables, right? And a great plot for that would be scatter plots, right? So the relationship between historical credit scores and number of loans taken, the relationship between customer lifetime value and number of products purchased, right? These are all great ways to show these types of plots. So common use cases of scatter plots here, for example, are, you know, if you want to look at what drives customer loyalty at an organization, right? So let's say you're a telecom provider and you sell multiple products, right? And you want to see what is customer lifetime value. The, you know, the lifetime value of a customer for that particular product. We can see that the more products they purchase, the higher customer lifetime value they have, which is a natural conclusion given that they have purchased more products. That is rule number one, choosing the best visualization. The second visualization, which is often the biggest pitfall, especially early folks working with data visualization fall into, is keeping visualizations minimal and avoiding clutter. It's really easy to add clutter to your visualizations. So let's talk about how we can minimize clutter and visualization by using one really good example here. Now there is this really great example from dark horse analytics that I'll let you guys check out where we talk about the trade off between cognitive load and effective effectiveness. Right? So as a rule of thumb, what you want to do is remove all of the elements of a visualization that don't preserve serve a purpose. So if you think about the signal to ink ratio, it needs to be incredibly high. Where signal is value, you provide ink is the ink that you provide that you use to provide that signal. Right. What you want is your signal to be way higher than your ink. So let's talk about it in a bit more depth here we have this visualization that shows calories per 100 gram for different foods. And we show different types of foods from french fries, potato chips, bacon, pizza, chili dog. As you can see, this is a really colorful slide and shows quite a lot of information. We use a lot of ink here. So what I'm going to try to do here is I want to preserve the same amount of information. I want to preserve the information which is just showing the number of calories per 100 grams for these foods. But I want to strip out as much ink as possible. So let's get started. First thing I'm going to do is I'm going to remove backgrounds. So we went from like this brown background to a white or transparent background here, right? And this helps us one focus in more on the visualization rather than focus on the color of the background. Second thing I want to do is I want to remove redundant labels. So if you look at for example french fries, fries, potato chips, bacon, pizza, they're there twice. So if I look at for example also type of food, number of calories. Right. A lot of what's covered here, right. Is redundant. So I have calories for 100 gram for different foods I have here, type of food I have here, number of calories I have here, the different labels. Right. So what I can do to remove redundant labels here is one, remove the labels that were here, remove the x axis and Y axis labels and then just keep the informations there. I see someone here talking about the shadows of the bars. That's going to go soon. Don't worry about it. The third thing that we're going to talk about is removing borders. Borders here are really important to think about, right. Because they're adding as well needless information. And you know, you may say here, but maybe I want to see what is relevant here. What is hitting the 600 or 500 mark, right? To be able to use as a guide for visualization, we're going to look at different techniques that will help us do that. So first thing I'm going to remove the borders, right? Especially from the visuals, right? And then thing I want to reduce colors where it doesn't matter, right? So look at color here, right? We have different colors for different types of visualization for different bars, right? Maybe you want to standardize on one color, for example, or maybe we want to highlight only one particular food group. So here, let's say I work for the bacon lobby. I want to be able to show that bacon is healthier than potato chips and french fries. And I can only highlight bacon right where it's relevant. Third, fifth clutter removal technique is removing needless effects. So here this person mentioned shadows. We're going to remove those shadows because these are needless effects here because there's no additional value that we're getting from this visualization by keeping these shadows. And then an additional thing here is the bolding that we see as well. We can also remove the bolding and keep it focused only on the data visualization. Now removal technique number six is remove the access lines when not lead it, right? So for example, sometimes we may need to keep the access lines, sometimes we may not. Right here it's all about making that trade off between signal to ink. So here I'm going to remove these access lines and you may say, okay, but what do I do about, how do I make sure that my readers know that the information is presently exactly. They know the exact amount of calories per 100 groups for these, for this particular visualization. What I'll do here is instead of adding access lines, I'll add labels directly on the plot. So here I'm able to get immediately french fries calories, potato chips calories, bacon calories, pizza calories, chili dog calories per 100 grams. So all in all, what we did is that we used white spaces instead of background colors. We, we removed chart borders, we removed grid lines or axis, we cleaned up axis labels, we labeled data directly as opposed to using a legend. We removed data markers and we used special effects like where we bold, underlined italics, shadows sparingly, right? So if we look at here how we started versus how it ended, massive difference in terms of the amount of ink reused for the same amount of signal provided. So yeah, there's a comparison chart Here. So now that gets us to our third rule, which is using colors creatively. And if you think about it, we use color creatively here to showcase bacon. So a really great way to use color, for example, is to distinguish between groups. And, you know, the best place place to get this is to. To kind of get, get that, you know, you creative use of colors come comes from election charts, data, right? This year, for example, there's going to be a election in the states in November of 2024, and we're definitely going to see these charts pop up more and more, right? But this comes from 538, which is a great site focused only on statistical analysis for, you know, especially political issues and topics, right? And this shows here colors to distinguish between groups, right? So, and this is a great way, right, to illustrate what would happen if only women voted versus what would what would happen versus if only men voted, right? So, for example, this is especially for the 2018 midterms based on patterns. And, you know, in the states, you have two political parties, the Democrats, which tend to have a blue color, and the Republicans, which tend to have a red color, right? So here we're using color very creatively to show how many seats Democrats would win if only women voted and how many seats Republicans would win if only men voted, right? And this really shows the power of using color because it instantly is recognized and in people's, you know, in the kind of subconscious cognition of people that they associate Democrats with blue and Republicans with red. So how can you use this, though, in a normal data set? Not all of us work with political data or polling data or election data. Let's go back to this example of what drives customer loyalty at an organization where we look at, you know, the correlation between customer lifetime value and products purchased. Let's say we have two segments of customers, segment A and segment B, right? And we want to show this relationship in the same chart segmented by customer category. We can easily add this nuance and color instead of having to create multiple charts at the same time. And this really shows here the power of using colors creatively to be able to show a simple nuance in our data set. Another use of color that you can use to, to use in your visualizations in your data storytelling is to color, to use color to highlight intensity. You know, we all remember Covid and how it was like, and, you know, Covid in a lot of ways was the data story of the century at the time. And then we looked at here, for example, another visualization from538 where it shows demand that COVID 19 testing sites depends on the community you live in. So it really shows, for example, that in certain areas, right. There are extremely high demand and a lot of busy testing sites. Right. So in California, for example, we see many testing sites have a potential demand of 200,000 or more patients. Light colors showcase that there's not a lot of demand for testing sites. Highly intense colors shows that there's really high demand. Right. So another great way to highlight intensity safety as well. Another way we've seen this as well is infection hotspots. For example, we see here which states in America are more likely to have infected people at a school of 500, right? Where if you have 10 plus people, let's say more than 10 plus people at a school of 500 that are infected, you will see the county here visualize in really strong red. If low numbers of amount, you'll see them in white or a pale red color. Now, there's also additional creative ways that you can do to use color to distinguish between groups. Right. So I'll use this election example again. You know, Democrats from every states are donating to Georgia. You know, the American election system has an electoral college, an electoral college system, I think not college, but I forgot what the name is. But different states have different, different, different power, a bit different number of votes associated with them. Right. So here, for example, this is an example from the Senate, right. Where there's a Georgia Senate runoff, right. Where all of the Democrats are donating to Georgia because that's where there is a key swing state or there's a key election happening in Georgia, for example. And here it's really interesting because we're using blue across the board, but we're using color intensity to showcase focus on the thing that we're focusing on here, which is the state of Georgia. So how can this be used also in normal data sets? Right. Let's say we're visualizing the number of customers by loan product at a bank. Right. Let's say we really don't want to focus on one type of loan only or two types of loans, business loan and loan against property. We can gray out the other ones that are not relevant to us and focus in and increase the intensity on the loans that we want to focus on while we tell our data story. This will help capture the attention of our audiences. We're almost reaching the fourth rule here of the data visualization rules, which is using texts appropriately. Labels can be extremely effective at highlighting insights. And I'm going to illustrate this example, let's say this is a customer support example where we're looking at number of tickets received versus processed and we're trying to convince leadership of investing more in the customer support team because the amount of tickets that we're able to process are unable to be met on a regular basis. Why is that? First thing I can do is I can add labels on my data. So I want to reach a point where I have a very convincing visualization. So the first step that I'm going to do is I'm going to add labels to show the delta that we're experiencing between numbers of ticket received and number of tickets persist. Second thing I want to do is I want to add the story and the labels here that provide context over this business context over this business problem. So first thing here I add the label that says two employees quit in May. We nearly kept up with the oncoming volume in following two months, but fell behind with the increase in August and haven't been able to catch up since. Right? So that here provides context. The second thing here that we're doing is we're using the title very creatively. Please approve the hire of two full time employees to backfill those who quit in the past year. So here we're trying to anticipate the request using the title. So labels can be extremely effective at highlighting insights. We also go back to this particular COVID19 example from 538. They used labels here, for example, right? To provide context over what's happening in California. However, they also provide context, a different context on what's happening in South Texas. Saying there are so few testing sites in South Texas that potential demand exceeds 600000 patients per site. Right? So here for example, we see in California that people are naturally inclined there, right, to go get tested. Whereas in South Texas the problem necessarily is not. It's not necessarily just on related to. Related to demand of tests, but also availability of testing sites. And here, for example, areas have a smaller potential patient demand where testing sites are more prevalent and evenly distributed. So there is high demand here, but there's just much more testing sites. So rules of thumb when using labels, label access and titles for clarity, label data points when necessary, play around with font size when highlighting specific messages and make sure that you anticipate common audience questions and how they can go into labels. I'll maybe focus in on that last one on common audience questions. Right? A common audience questions here will be why are we not able to receive to. To follow up with process to process received tickets. So here we added a label to answer that question way in advance. Now that we covered our rules for better visualization, let's talk about rules for better narrative. First thing is know your audience and know your format. So first thing is, let's talk about this in a bit more depth there. We have different stakeholders and different formats depending on the, the, the presentation that we have, we may want to present to an executive. And they have relatively low data literacy, right? They only care about outcomes and decisions, right? And then maybe a format that is relevant to them. Here is a presentation. If you're a data scientist and you're presenting your analysis to a data science leader or a partner, right? What they care about maybe is the rigor of your insights, right? So you may want to share here a coding notebook, for example. If you're working with a business partner, they have high data literacy, they care about tactical next steps. What they care about here necessarily is a detailed written report, for example, on what we ought to do based on the story that we're trying to tell. What's really important though is that you know the priority of the audience, right? So the executive, they care much more about the business impact of your model, your dashboard, your analysis than about the way you got there, right? So I'll give an example of a machine learning model, right? The executive cares much more about the business impact of implementing the this machine learning model. Then about how you've been able to improve a model by 1% using this new technique and this data cleaning technique, right? So let's say I'm doing a presentation on a customer churn model, right? I would say we've built a customer churn model. The expected impact is X amount in dollars from being able to do X and Y Z. Next steps for being able to implement it are X, Y and Z. Here's what we need to make that a reality. Rather than we used random forests and built these features and did this and did that and did use the ROC curve to determine accuracy, yada, yada yada. A data science leader, for example, cares much more about how you arrived at your insights and to battle test them for rigor, right? So here you actually need to share the ROC curve, how you arrived there, because probably data science leader wants to vet what you did before going to the executive. A business partner cares much more about how your analysis impacts their workflow and what should be the main takeaway from the data story. So for example, if you're working with a marketing stakeholder on if you're working with a marketing stakeholder on the customer churn model, right? What they care about maybe is how can we improve the targeting based on the customer churn model and provide recommendations, right? So practice empathy with your audience. Does the audience have the necessary prerequisite knowledge to understand a particular metric? For example, how much time does the audience have to consume the state of story? If you're doing a 30 minute presentation, do you really need to go into the details about the machine learning model? What is the medium of the presentation is written oral that the audience prefers, right? Think about the empathy here that you can build with your audience. More importantly, what you need to know is know their level of data literacy or subject matter expertise. And a great examples here is comes from the Economist. So the Economist does incredible visualizations, right? They're really well known for really high quality visualizations. But one thing that they did over the past few years is take their visualizations and put them on social media and for channels like Instagram, for example. They need to be much more accessible and less detailed, right? So, and because you're effectively also marketing these, these visualizations to a younger audience, right? So what they did here is they took the same visualization on the left and simplified it because they are, they know that their younger audience on Instagram has a lower attention span and probably doesn't want to go read into the details on such a small screen, right? Rule number two, begin with the goal in mind. It's very tempting to throw everything you got at a data story, right? Every chart you have, every single feature, every single piece of analysis. It's really easy to be like Oprah here and be like, you get a chart, you get a chart, everybody gets a chart, right? But what you want to avoid here is overburdening the audience with information that is not necessarily relevant for the objective that you have for the data story. So always start with the goal in mind. So for example, here convince management of investing additional resources on the support team. We saw that example. That is the goal, right? So who is the audience? The more specific the audience, the easier it is to build a data story that resonates. And what should the audience know? So know your recommendation before you craft the data story. Then what data can you use to convey your point? Identify the data that is relevant for this data story. However, this doesn't mean that your goals should determine the data story. What the data is telling you always takes precedence. So we shouldn't lie with data, but we should filter the unnecessary data. That doesn't help us achieve Our goal rule number three, which is what we just discussed here, do not mislead with data stories. It's really easy to mislead with data stories, whether intentionally or unintentionally. And this is the fastest way to losing credibility. I'll show you here an example based on the chart that you see on the left hand side, right? So we have global temperature anomalies, right, from the 20th century over the past decades, right? And if you think about this chart, right, it's showing you like global temperature anomalies over the past decade, right? And the in Celsius you would be like, you know what this is? You know, there's no global warming, there's nothing here that's happening, right? But only because what we're using is a Y axis that is extremely small, like really large data. Sorry, not Y axis. If the X axis goes really bad back in time on a short frame, we're going back to 2002, right? However, that's not how climate data works. Climate data works over the long haul, works over decades. So if you go back to the 1800s, for example, from the time that we had measurements in the 1800s, right, you'll see that there has been a market increase, right? So always make sure that your time horizons make sense given the data you're treating, right? So a good example would be, let's say you're a social media manager and then you show that in the past year you, the number of visits to the website from social media has gone down quite a lot, right? If you show only the past two weeks in data that sees like it's going up and you're like, everything is going on the uppity up, right? But if you look at it over the period of 12 months, maybe it doesn't tell the same story, right? So that's a really good example here of how you can lose credibility. The second one here is a pitfall number two to avoid is ensuring that axes all start with zero, right? So a good example here comes from, you know, interest rates, right? So here we're looking at same data, but different Y axises. If you use really incredibly small, you know, unique increments on the Y axis, right, where you use one 0.01% unit, 0.001% unit increments, right? You would assume that from 2008 to 2012 there's been an explosion in interest rates, but actually it's been relatively stable. If you look at a unit increment of 0.5%, which here makes much More sense to use. More importantly as well, ensure that your axis start with zero, right? So here we see that the axis starts at 3.14% and then jumps up to the top, right? Whereas here we start with zero percent as well. Now pitfall to avoid number three is that ensure that access of scale are appropriate, right? So on the original side here on the left hand side, this is also an economist. Economist one, right? They were showing the average number of likes per Facebook post, right, across different politicians in the uk, right? If you look at Jeremy Corbyn here and then look at the Labor Party, Jerry b. Corbin has 5210 average likes per post. Labor party has 800. But from the visualization you'd assume that Labor Party has almost half of what Jerry B. Corbin does, maybe 60% of what Jerry B. Corbin does. Not necessarily 1/5, right? A better visualization here is ensuring that you're using the access scales appropriately with the visualization. So then they changed it to be more accurate. Other best practices to ensure that data stories are not misleading. If you're sampling data, make sure that they're representative of the population. So make sure you use a random sample. If you're using use centrality measures, e.g. median mode mean to ensure context around the population is taken into account. For example, example, right? Rule number four, develop a narrative around your data. Different narrative structures to choose from, right? I'm not going to pretend to be a storytelling expert, right? But you know, you have Aristotle, strategy structure, Freight tax pyramid, Campbell's Hero's Journey. Campbell's Hero's Journey is actually, you know, Star Wars, Lord of the Rings are based on Campbell's Heroes journey, for example. And these are all mentioned in the book I mentioned by Brent Dykes on effective theater storytelling, which I highly recommend that you gu guys check out. This is also from his book Brent Dykes. He talks about the experiments like this, this narrative structure that are really that I really like. You set up, you set up a background in the hook, right? You then discuss rising insights. So here, for example, supporting details that reveal deeper insights into the problem or opportunity, you find the aha moment, right? Major findings or central insight. And then you talk about solutions and next steps. I'll give you guys an example from my work at datacamp. One of the main traffic, main metrics that I think about at Datacamp are, for example, podcast listens, right? Let's say podcast listens on Data Frame suddenly dropped down quite a lot. So here the background is Data Frame. Podcast lessons have decreased in the past six months. We have yet to know why. Well, rising insights essentially focus on revealing deeper aspects of the data. So here we look at, we break down podcast listens by channel. For example, something like Spotify, Apple Podcasts, the Google Podcasts, right? And then aha moment here for example would be, oh, we find that one channel is specifically responsible for the decline. Right? And then solution and next steps here would be investing more in that channel or finding alternative channels. Right? And here for example, this is a, an arc of a story that we can see. So another one yet? I'll give you another one here. Our organic traffic has gone down. We've noticed it's specifically coming from our blog. These articles have been outranked by competitors. We need to update these articles to make them more useful for, for our audience. I highly recommend that you check out these, store these resources on narrative because narrative is a really big area just in its own and there's no one size fits all. It really depends on your style. So with that, we're nearing the end of our presentation. I want to make sure that everyone does ask as many questions as possible because I want to be able to answer them and we're going to have a bit of time. I want to talk about how everyone will become a data storyteller and I want to talk about resources to improve your data visualization skills. So one, data storytelling is stable stakes, right? Every year at the end of the year, we get our Spotify wrapped and then we see, you know, we see a story about ourselves. And it's a data story. It's a story that shows how much how. What was our favorite song in April? It was. It showed us what is our favorite author, favorite singer, what is our favorite genre, right? This is a data story about ourselves, right? And that's why it resonates a lot. Data storytelling skills are table stakes in a world where we are always working with data. More importantly, another way that data has been table stakes is through COVID 19, right? Everyone was glued to their screen looking at infection charts in the past, you know, in 2020 and 2021, right? There's one screenshot I took two years ago in Belgium, right? And then in 2024, we're going to have an election where we're going to be really focused on election data. These are all data stories. So here comes the, you know, shameless plug section of the presentation. Level up your data visualization skills with Data Camp, right? So we have quite a ton of data visualization and data storytelling skills at Data Camp from Data visualization with Python to data visualization with R, whether data visualization with Tableau or Power Bi. And with that comes the point where we can start answering questions. But before that, I'm going to drink a bit of water just to make sure that I'm ready to answer all these questions. Okay, Can I just come seeing here quite a few comments as well before I start getting questions? Right. This is a great comment here by Daniel. I'm a huge advocate of internal style guides with easy to use templates, light and dark color palettes, proper contrasting points, contrasting typefaces. Defaults are important and most people don't change them. I completely agree. In fact, I'll share the these examples of. For example, I'll show this Airbnb data visualization theme. So Airbnb, for example, right, have this internal package, right, that they use. I forgot which is the one that they use here? No, these are. This is their own design language. But yeah, they have a ggplot2 theme that they've developed internally, for example, to make all of their analysis. Airbnb branded. Right. Another comment here that I've seen as well. This is also super important. It's a bit of a long one. I don't think I can show it all, but you need to be careful with colors. Some color combinations can be hard on the eyes or create comfort or discomfort on eye strain. Or for example, you may also have people that are colorblind. Right. These are all super important considerations as well when it comes to visualizations. Question here that I'm getting. Are there any tools or software that you recommend for effective data storytelling? Well, there's tools and softwares for great visualizations, but they're not necessarily for great data storytelling. Data storytelling is really the human side of data visualization and this is where you need to work on those soft skills here. Another question here from Ilya. How can you measure the success and impact of a data story within an organization? Well, did your stakeholders accept the insights? Did they take away something actionable? What are some common mistakes people make when crafting a data story and how can they be avoided? I think the biggest common mistake is throwing. I'll go back to that Oprah slide. Throwing everything you have at a data story. Right? It's very tempting to throw everything you got at a data story, and this is where you need to avoid that pitfall. Karen, Matt, do you have a slide with all of the rules listed? So we're going to share the slides, but also there is a white paper version of this that you can access for easy convenience. Another question here from Christy. What advice would you give for making data stories more engaging for non technical stakeholders? You mentioned it in your answer. Non technical. Make it non technical. Focus on the business value and what they care about. Have empathy for the, for the audience. We have another one here from Laura. How does one ensure the data story aligns with the overall business objectives? Right. So ensuring that their story aligns with it really depends on the context and the audience, right? I'll go back to this slide. For example, really important that you focus on what the audience cares about. Another question here from Danielle. Visualizations have always, always lead to a solutions or just lead to know that there is a problem and where is the problem. So visualizations don't need to always lead to the solution, right? A data culture essentially is where there is common understanding of where the problems come from and data visualizations can be a great tool for that. I hope that answers your question. Here. I'm seeing this question. I really hope I'm understanding it. Well, what is, how would you write, how would you advise someone to write? Data storytelling, I assume get started with data storytelling. You mean highly recommend that you check out our courses, right? But data storytelling is, you know, focus on crafting the right narrative for your audience and think about who the audience is. Oh, this is a good question. How do you find a good balance with data storytelling narratives and use of dashboards where you have multiple pixels, that is really great. So when it comes to, you know, embedding a data story in a dashboard, think about organizing the dashboard in a way that logically flows. So think about how a data story starts with certain questions that are high level and then you want to drill down, right? So if you replicate this, for example, in a dashboard, right. Let's say here I'm going to look at traffic, right? I want to first look at overall traffic, right? And see like changes over time. Then I want to look at sources of traffic here, for example, would be the blog, right? Or the courses or whatever. Then I want to see how specific articles are performing, right? Then I want to make sure that I get also a visual that shows me what blogs to update, right? Really depends on the business problem. But you want to make sure that you optimize your dashboard for a flow that makes sense. Another one here. What advice will you give an upcoming learner in terms of tool to learn first for data visualization? So if you are coding, if you want to code, I would recommend either Python or R if you want to use no Code tools. I would recommend Power Bi or Tableau. Also, Excel is an option here. Oh, so here, how about a highly technical chief data officer or manager? Right? That's a good question. What I would do here is I would lead with the business impact and then put the technical details in the appendix. I don't see any more questions. A question here from Rachel. What are some ways that you can maintain ethical considerations that you avoid lying with data stories? So. Here we go into this, right? Quite a few different pitfalls that you want to avoid, but make sure that you're being intellectually honest about what the problem is. And this actually connects to a great question here. How do you push against a leader who wants to manipulate a visualization so it shows a better story? It's a tough one, right? Because if you, if the leader is doing this. With, with bad faith, right. Then there is a bit of an ethics problem here that you want to push back against. From. So maybe you want to talk to their leader. Right. But if you. But the best way, in my opinion, to create shared consensus. Right. Is to, you know, I don't know what this particular example that you're mentioning here, right. But as an analyst, right. There are different stakeholders that you have to manage. There's often politics that you may not be aware of. And what you want to do here is adopt a mindset of truth seeking. Right. And be objective. So, for example, if the leader is like, oh, no, can we update the visualization to look like that? Right. I would actually create a public forum where you say, you know, this is what I found in the data. Right. I'm not sure if I'm interpreting it correctly. What do you guys think? And then this creates a debate internally. Right. And if the leader is still approaching this in bad faith, this is where you may want to escalate this to another leader. And maybe thi c here you mentioned that this happened in your previous roles. How did you navigate it at the end? How would you adapt data storytelling to an academic audience? So if they're technically minded, you probably want to be a bit more technical in your approach and be really focused on the, the rigor here. I don't see any additional questions. So, yeah, thesi, I presented the truth and backed it up with science. That's, that's the approach, truth seeking here. All right, I think this wraps about. Oh, and then additional context here from thesi. Why we present things in certain ways. But I asked my boss for help. That's a. Yeah, great, great way to, to, to manipulate this I think you did a great job. That said, I think we are almost wrapped up. I just want to make sure that everyone is aware to go to our upcoming webinars page for future events. Right? Right. We next Tuesday we have getting started with machine learning using ChatGPT. Do check it out. And then we also have another session on May 8th on how you and your team can learn data skills more effectively. And also at the end of the quarter we have Radar AI. Now this is the previous Radar AI. Oh wow. If you go to Radar AI 2024, we have on June 26th, June 27th, an incredible two day event. If you've ever been to our Radar sessions or Radar conferences, you would know what I'm talking about. And I think this Radar is probably the most ambitious radar to date. We have two days of events with incredible speakers and thought leaders in the AI space. From Hugging Face to Monte Carlo to Pinecone to Atlant. Thomas Dungush is joining the Data Lab CEO is joining. George Frazier, CEO of five friends, joining Chandra Dunelson, she's the Space Data analytics officer at the U.S. space Force. Cindy Houston is joining. Lots of amazing folks are joining us for Radar at the end of the quarter. Everyone do check that out. Do register. We want to see you there. It's all free, it's all virtual. Choose whatever session you want to join. We'd love to see you there. And with that, I will let you be. I'll give you back eight minutes back of your day. And again, please. People joined us from all over the world. I can't stress this enough. I'm so appreciative of everyone who joins these sessions from different time zones. People joining us from Singapore, it's like what, what time is it Singapore right now? I think it's like 10pm it's almost midnight in Singapore right now. So I really, really appreciate everyone joining in and it is truly an honor. I cannot stress how, how great these presentations are and how much joy I get from being able to.