2026-04-23 22:00:59
Recently, I had the chance to be part of a panel at Canva Create on Data Storytelling in the Age of AI, alongside Duncan Clark and Tey Bannerman. We discussed how quickly AI is changing what’s possible and what that means for people who work with data. The conversation kept coming back to a provocative question, If AI can generate the charts, what’s left for the human?
Here’s where I’ve landed: the hard part hasn’t changed.
AI can make graphs in seconds. Entire dashboards. Draft presentations. Even suggested narratives. And while the design quality is currently inconsistent, that gap will likely narrow quickly. This is redefining what it means to be “good with data.” Historically, it meant skill with tools: knowing how to write the code, wrangle the data, structure the dashboard, or build the graph.
AI is shifting that. But maybe not in the way people think.
Tools have been making graphs for decades. Excel. Tableau. Flourish. The mechanics were never the most difficult piece. The hard part has always been deciding what matters: understanding your audience, interpreting what the data means, and figuring out what someone should do differently because of it.
Let’s look at an example.
The following side-by-side shows manager performance scores before and after a pilot manager training program. On the left is the AI-generated chart. On the right is the one that I designed.
The AI version isn’t wrong. But it also isn’t right for my specific scenario. That’s where I, as the data storyteller, come in. Knowing who my audience is, I can make smart decisions about how to visualize the data, what clutter to eliminate, where to focus attention, and what action to drive. That is data storytelling.
While AI might make it easier to generate output, it doesn’t make these important decisions for you—and it shouldn’t. If anything, it makes the humans making them more important.
A practical strategy I shared on the panel (one I was teaching well before AI) is to start low tech.
The temptation with any tool—AI or otherwise—is to jump straight into generating graphs and slides. The risk is that you skip the most important step: determining what you actually need to say.
Start analog. Gather your team around a whiteboard or reach for a pen, paper, or my personal favorite low-tech tool: sticky notes. Figure out the plan and the story. Who is it for? What decision are you trying to support? What action do you need to drive?
That gets at something I’ve been thinking more about lately: the role of friction.
Friction, in the classic sense, slows us down. In some instances that’s undesirable. This unhelpful variety comes in the form of confusing graphs, cluttered slides, and unclear structure. This is the friction I’ve spent my career helping others remove. If AI can help eliminate this type of friction, I think that’s great.
But there’s also productive friction. This is the type that forces us to think more clearly before we communicate. Go back to that rule I shared about starting low tech: there is literal friction as you move a pen across paper. That slows us down in really useful ways—to ask what matters, question the data, decide what action makes sense.
This reminds me of a conversation I had with Ken Field for the SWD podcast. Ken is a cartographer, and he talked about how much mapping has changed in the past couple decades. Because the materials to physically make a map were expensive, there was an incredible amount of planning before ever putting pen to paper. All of this changed with GIS tools—suddenly, basically anyone could make a map. And that planning step was no longer required to create one. But it remains just as important for creating an effective one.
That stuck with me, because it’s not really about maps. It’s about thinking.
AI is doing something similar—at a scale far beyond maps. It can remove a lot of the production friction, the mechanics of building graphs and slides. That is incredibly powerful.
But it also makes it easier to skip the most important step: thinking. That’s a major problem. This means the goal shouldn’t be to eliminate friction entirely. It should be to remove the parts that get in the way, while preserving the parts that make us better. Don’t let AI replace your critical thinking!
That tension—between speed and judgment—is key. AI is incredibly powerful at generating options. It removes the blank page and speeds up exploration. These are all good things. But someone still has to decide what’s worth focusing on.
It’s important to note that this isn’t a new phenomenon. We’ve seen it before. There was a time—not that long ago—when organizations believed hiring data scientists would solve their data problems. Before that, it was dashboards. While technical expertise is critical, it didn’t crack the real issue: you can generate all the insight in the world, but if people don’t understand it or know how to use it, nothing changes. In this world, an effective data storyteller bridges the gap between insight and action. It is the data storyteller who will bridge the gap with AI as well.
AI can generate answers exceptionally fast. But someone still needs to decide what questions are meaningful, which data is trustworthy, and which action should follow. AI might be able to generate the chart, however humans still own the interpretation. They must decide what it means and what to do next.
It’s also worth highlighting that AI can generate polished-looking outputs very quickly. This can make it easy to overlook flaws. When you do use it for analysis or content creation, it’s critical to remember: you are responsible for what you put in front of an audience. A simple rule still applies: never present a graph you can’t explain.
If you can’t clearly articulate what the data shows, why it matters, and what people should do with it, then it’s not ready. AI doesn’t change that—it just makes it easier to skip that step. But the accountability hasn’t shifted. If your name is on the slide or you are the one presenting it, your judgment should be behind it.
I was talking with a lawyer friend recently who shared something that stuck with me. She had been using AI to help draft briefs and emphasized how important it is to provide strong context up front to get useful output (the corollary to a common warning from my past life building statistical models: garbage in, garbage out). But she also said something else: once the draft is done, it still needs a critical review.
That applies here, too.
AI is designed to satisfy the user. Which means if you’re not careful, it will start telling you what you want to hear. That’s where a second set of eyes becomes incredibly valuable—someone outside the back-and-forth who can question what’s there without bias.
I’ve long been an advocate of getting another perspective on your work. When AI is involved, that step becomes even more important.
One of the things that struck me most at Canva Create wasn’t the technology—it was the people. I had the chance to connect with so many thoughtful, curious, creative humans, who care deeply about how their work lands and the impact it has.
There’s a lot of focus right now on what AI might replace, and a legitimate concern that it could stunt thinking or dull creativity. But I left with the opposite view. If anything, this is a moment to double down on what makes us human. Storytelling has never been about charts or slides alone. It’s about connection. It’s about understanding your audience, meeting them where they are, and helping them see something in a new way.
While AI can generate content, suggest structure, and speed up production, it doesn’t care. It can’t read a room or build trust. It can’t decide what matters in a way that reflects human context, nuance, and judgment. That part is ours. So rather than asking what AI will take away, a better question is: what does this make more important?
For me, the answer is simple: human connection. Data storytelling has always been about that. In the age of AI, it matters more than ever.
2026-04-20 21:00:11
Gantt charts are a popular choice for illustrating the start and duration of events, which is common practice in project management. While useful for representing timelines, these charts can quickly become busy and difficult to interpret, especially when dealing with complex workflows.
Let’s consider an example.
Inspired by NASA Schedule Management Handbook
This is a project schedule with risk assessments showing the probability distributions for each activity of interest. The colors of the bar represent the likelihood of completion. For instance, the green portions towards the left show optimistic timing (10–50%), while the red portions indicate later, more likely completion dates (70–90%).
The black triangles represent the planned date, or the scheduled target, of each activity listed on the left, while green triangles show the need date, the required date to meet an on-time delivery. A table at the right lists the probability of meeting each date, adding a lot of information, but also visual complexity.
Simplifying involved views like this allows the audience to spend less effort decoding the structure and more time engaging with the insight it conveys. Imagine we want to use this chart to show where the project team should focus its efforts to meet the desired delivery date. Specifically, the Functional Test, Vibe, and Therm-Vac workstreams are showing no chance of on-time completion and will require corrective action to meet current deadlines. Some thoughtful design changes could help improve the effectiveness of this message.
The original view has many lines contributing to the cluttered look. Removing the gridlines is a quick and easy way to begin to bring the pertinent information to the foreground.
With fewer borders, greater contrast is created between the background and the data, allowing the horizontal bars and milestone dates to stand out more.
Another way to make it easier for our audience to consume this information is to be thoughtful about the order and alignment of the data and labels. Currently, the dates are at the bottom, which means they will likely be the last thing that is seen. Since dates are such an integral part of a project timeline, we should move them to a more prominent place at the top of the chart.
By moving the dates higher up, we can make sure they are seen first. Additionally, by simplifying the date formats on the axis, it becomes possible to rotate the text horizontally, making it quicker and easier to read.
Our next step is to rethink the use of color in this chart. The current green, yellow, red gradient is a challenging combination to discern for people with color vision deficiency.
A color palette that uses shades of gray for bars eliminates the visual noise from multiple bright colors, allowing the viewer to focus on the timeline and task durations. For the milestones, more colorblind-friendly colors—blue and orange—make them stand out. Opting for circle markers is visually smoother and feels less alarming than the pointy triangles. Finally, applying the same color used for the circle icons to the probabilities listed in the table on the right helps to visually tie them together.
Now that the chart is less busy, we can turn our attention to guiding viewers toward the information that matters most. Ideally, we hope to draw our audience’s eyes to key elements or insights in the graph. We can apply some intentional focus techniques to emphasize these key takeaways for our audience.
As things stand, the Functional Test, Vibe, and Therm-Vac tasks have zero probability of being completed on time. A simple way to quickly get this point across is to add a takeaway in words at the top of the chart to tell our audience what we want them to know.
The added text indicates what we want our audience to see in the chart. We can also make additional adjustments to the chart itself to ensure that the words our audience reads are reinforced visually.
A light orange band spanning the three tasks named in the title draws attention to the workstreams at risk. Sparing color highlights just the plan dates (since these areas do not have need dates), while deeper intensity of color spotlights those projected to miss the scheduled target. Finally, placing the plan and need date probability columns on the left makes them quicker to find, and bolding task names and zero-percent probability values adds prominence.
By applying decluttering and focusing techniques, we can bring clarity to a complex Gantt chart, leading to quicker decisions and a clearer understanding of schedule risks.
To see more examples of timeline visuals, check out our April 2026 Challenge.
2026-04-07 03:55:48
I’ve always loved the hiring process.
I remember when I was building my team at Google and spending an inordinate amount of time designing the interview case study. I wanted something that would help us assess candidate’s skills—but also give applicants a realistic preview of the type of work they’d be doing if they joined the team. (That is, until the case study was leaked and I had to redesign it!)
What I loved most wasn’t only designing the process. It was seeing the organization through fresh eyes. Through candidates’ questions and perspectives, I was reminded how special the work—and the impact we were having—really was.
I’ve felt that same sense of perspective over the past couple of months as we’ve focused on hiring for the storytelling with data team.
When we opened applications for the Data Storyteller role earlier this year, we weren’t entirely sure what to expect. What followed was both energizing and humbling: more than 150 people applied!
As a small team, every addition matters enormously. Each new person changes how we show up for our clients, our community, and each other. Because of that, we take an incredibly intentional approach to hiring. The process is a significant investment for our team. From reviewing applications and candidate exercises to conducting interviews and thoughtful debriefs—I’m incredibly grateful for the time and care the team brought to every step.
An aspect that made things difficult wasn’t narrowing down the field because of a lack of qualified candidates—it was quite the opposite. We met many talented people who are doing meaningful work to help others communicate more effectively with data. It was hard to say no to so many great people.
One of our goals throughout the process was to make sure that even candidates we didn’t ultimately select had a positive experience. Many of the people who applied are part of our broader community, and we hope they continue learning with us, participating in challenges, reading the blog, and pushing this field forward alongside us.
From this incredible group, we’re excited to welcome two new teammates.
Today, I’m excited to introduce two people joining our team: Alli Torban and Ryan Biesecker.
Alli brings deep expertise in data visualization and data literacy, along with a passion for helping people truly understand data. She is the host of the Data Viz Today podcast, the author of Chart Spark, and has been a thoughtful and creative voice in the data visualization community for years.
At SWD, Alli will be contributing across our workshops, content, and community initiatives, including taking ownership of the SWD podcast, helping shape its next chapter while continuing to bring thoughtful conversations to our audience.
Ryan joins us with a background in consulting and analytics, where he has helped organizations use data to answer complex business questions and drive better decisions. He brings strong technical skills, client experience, and a natural focus on translating analysis into clear, actionable insights.
At SWD, Ryan will contribute across our workshop delivery and content while also helping us better understand patterns in our client work—bringing a data-driven lens to how we think about relationships and opportunities to expand impact.
You’ll start to see more from Alli and Ryan soon, as they begin contributing across our workshops, blog, podcast, and other SWD channels.
Growing our team allows us to increase our capacity to work with organizations through team trainings. More importantly, it allows us to do this while maintaining the thoughtful, high-quality experience we strive to deliver in every engagement. This means we can continue to serve our clients and community, while expanding the reach of the work we care so deeply about.
We feel very fortunate that Alli and Ryan chose to join us. Each new teammate helps us further our mission: to help people communicate more effectively with data and drive positive change through better understanding. Please join me in welcoming Alli and Ryan to the storytelling with data team!
2026-03-09 21:00:11
When you’re the one working with data, you likely know it better than anyone else. While this is great—it puts you in a fantastic position to help others derive value from it—it can also be problematic. Because when we’re close to a topic, it can be hard to detect when specialized language, acronyms, and abbreviations that all seem super obvious to us, can totally obfuscate our message. (Mike writes about this phenomenon, known as the curse of knowledge, in Chapter 8 of Before & After.)
Consider the following example. Take a moment to scan the slide below and see if you can quickly identify the main takeaway.
This slide summarizes performance for a SaaS (software as a service) company, showing how revenue from newly acquired customers grows over time based on how they were acquired and how many products they adopt.
Before we can start interpreting the data, we have some serious decoding work to do! There are numerous acronyms and abbreviations that are used repeatedly across the graphs and commentary.
By my quick count, there are more than 50 acronyms and abbreviations on this single slide!
Let’s start by defining the main ones:
ARR: annual recurring revenue
MOB: months on book (time since account opened)
AE: account executive, or sales-assisted channel
PLG: product-led grown, or self-serve channel
MP: multi-product customers
SP: single-product customers
LTV: lifetime value
Accts: accounts
For the people who produced this slide, these terms likely feel perfectly normal. For those outside that team—even someone from another part of the business—this slide requires a ton of translation before we have any chance of understanding the message.
Speaking of the message, if we refer back to the subtitle of the slide, we can learn the core insight: High Value Disparity in MP vs. SP Accts.
In plain language, this means: customers who adopt multiple products generate far more value than customers who use only one.
That key insight is buried under layers of shorthand and reporting language.
I reworked the slide with this insight in mind, articulating it succinctly at the top. I also defined acronyms, spelled out full words, and stated things simply. I condensed the data that was previously in four graphs to two, making it easier to compare each metric across all customer segments. I organized titles and commentary together with each graph to make it easier to connect the words to the data that support them. I added a discussion topic to help direct next steps.
The original slide might have worked perfectly well in a regular report. Reports often prioritize completeness and precision.
But when you need people to truly understand something and pay attention, it’s worth taking the time to translate. This means translating:
acronyms into language your audience understands
jargon into meaning
analysis into a clear takeaway
The challenge isn’t the data—it’s the translation.
When we’re deep in our own analysis, it’s easy to assume that others speak the same language we do. But our audience doesn’t live inside the metrics, acronyms, and reporting structures. If we want them to engage—and act—we have to translate first.
This is exactly the kind of “aha” moment we see in our storytelling with data team trainings, which are customized around examples from your team’s own work. Often the biggest breakthroughs come when people see how small changes—like clarifying language, focusing the message, and simplifying the visuals—can completely transform how their data is understood.
2026-03-03 23:00:00
The importance of an effective slide title cannot be overstated. Positioned in prime real estate at the top of the page, it is often where an audience’s eyes will land first. With that in mind, it is worth investing time to craft a title that introduces the content below and establishes a clear purpose. Too often, this valuable space is used for purely descriptive statements. Let’s look at an example.
With the title “CONSUMER ENGAGEMENT ANNUAL REVIEW,” an opportunity is missed to indicate what truly matters with the slide, which shows customer engagement for a department store across several retail areas. The audience is left to determine independently what deserves attention. This can be improved by making a key observation about the data. Doing so primes the viewer to look for evidence of that statement in the chart.
This is better. The most important insight from the visual, a concerning decrease in engagement for the Home & Garden area in November, is now clearly communicated. Yet, committing to a single interpretation of the data can feel uncomfortable. Perhaps this is part of the reason we sometimes see slide titles posed as questions.
At first glance, this feels engaging. A question invites curiosity and perhaps debate. But good intentions can introduce complications. When faced with a question, the audience may:
Mentally attempt to answer it: Now those present have been asked to think. That can be engaging—but it’s also demanding. They’re trying to answer the question, interpret the chart, and listen to you at the same time. Not all of those tasks can happen simultaneously. Most likely, what suffers is attention to the presenter.
Interrupt the presentation: Questions create a knowledge gap that the brain wants to fill. This can be powerful, but only if the answer comes quickly and clearly. More outspoken participants may try to answer immediately, steering the discussion in a direction not planned or practiced.
Come to the wrong conclusion: In trying to answer the question, the audience might need to make assumptions or draw their own conclusions. This risks putting them on a different path from the intended narrative.
Undermine presenter credibility: Posing a question can unintentionally signal uncertainty, implying that a firm conclusion has not yet been reached or the presenter isn’t confident in the insight. A heading like “Are our campaigns effective?” may leave those in attendance wondering whether there’s a clear answer coming at all.
Instead, consider using an active takeaway title that explicitly states the key message—and signals how the information should be interpreted or acted upon. In the case of our example, the promotional content for the critical Black Friday period wasn’t as extensive as it could have been, resulting in consumers exploring competitors' stores for deals. A further revision of our slide demonstrates the effect this clarity of message can have.
With a clear purpose established for the slide, the added benefit is that the graph can now be revised to tie to the statement visually. Let’s iterate one final time.
In this version, color is used sparingly to emphasize the Home & Garden line, annotations (plus sparing data markers) have been added to highlight the key moments, and a clear recommendation clarifies what to do next. The meeting can now focus on discussing the proposed action rather than debating what the data means.
Slide titles framed as questions are more engaging than descriptive ones. Used deliberately, they can spark curiosity. But they can also divide attention, introduce confusion, or imply uncertainty. Instead, identify the main purpose of the slide and craft a takeaway title that states it clearly, ideally signalling the appropriate course of action. And in doing so, you can be confident the audience immediately understands the point of the communication and how to move forward.
If you are interested in learning more about the importance of words when creating slides, explore past articles for additional examples: transforming slide titles and vertical logic. Or if you’d like to practice crafting a compelling takeaway title, check out this community exercise.
2026-02-20 03:28:47
I was facilitating a workshop recently when someone asked one of my favorite questions about a graph on the screen: “So… what are we supposed to take away from this?”
Such a simple—and useful—question.
One challenge was that the graph was attempting to show multiple comparisons at once, so it wasn’t clear what mattered most. To further complicate things, the data in question spanned very different magnitudes, with one category dwarfing the rest.
Here’s an anonymized version of the original slide, showing new subscribers to a streaming service by plan type across two years:
Let’s revisit the simple question that was raised: what are we supposed to take away from this?
The story could be about one plan dominating the rest. Alternatively, the long tail of niche plans might be worthy of review and discussion. Or perhaps there’s something interesting about the year-over-year shifts across plans.
Each of those takeaways can be found in the original slide, but it takes work to get there. If this is part of a routine report or update and no action is needed, it might be fine as is. However, if we want someone to understand one or more of the nuances at play and drive a discussion or decision, we can approach things more strategically. Part of doing this means choosing different visuals for different takeaways. Another aspect to consider is whether to aggregate and how to treat categories so that we can see important points—without overwhelming with detail.
Let’s tackle one specific takeaway at a time, intentionally choosing a visual that supports it.
If our goal is to clearly communicate that one plan accounts for the majority of new subscribers, instead of burying that insight in a crowded bar chart, we can make it explicit.
This view is useful to show concentration (alternatively, stacked vertical bars or even two pie charts could have worked here). By collapsing many categories into just two—Premium vs. everything else—we remove noise and make the dominant pattern unmistakeable. This kind of aggregation is most appropriate when the goal is orientation or framing (not detailed plan-level decisions). The audience no longer needs to scan bars, compare labels, or reconcile magnitudes. The message is immediate.
I’ve kept the words descriptive here, but you can imagine how we might use a view like this to prompt a specific discussion, for example, whether the business is overly dependent on a single plan, how resilient new subscriber growth would be if Premium performance changed, or to consider whether it makes sense to invest in Premium or elsewhere.
Speaking of elsewhere—if those non-Premium plans are key, we’ll want to approach things differently.
It can be challenging to effectively visualize numbers of very different magnitudes in a single graph. In the instance where one or a few categories are large, but the story is in the smaller ones, removing the large categories from the visual can sometimes make sense. Key is to retain the context of the omitted categories so things aren’t accidentally misinterpreted.
In this instance, we could use words to comment on the size of Premium, then focus the graph on the non-Premium plans. See below. (If you were still afraid it might be missed, you could even play with positioning a bar under the premium note that extends off the right side of the page.)
Treating Premium separately and removing it from the graph allows us to see what’s happening in the tail. Low volume doesn’t mean low value—but it does suggest a different set of strategic questions.
A view like this could be used to drive discussions about portfolio rationalization (are there too many low-impact offerings?), whether small plans are strategically important or operationally distracting, or where simplification might free up resources without materially impacting growth.
A third—and entirely different—focus is change over time. Which plans are gaining momentum, and which are losing it? A slopegraph can make those changes easy and fast to see.
A view like this could help frame discussions about where recent investments are paying off (or not), which plans may need renewed attention or repositioning, or simply to better understand how subscriber preferences are shifting year over year.
None of these views is the correct one. Each highlights a distinct takeaway—and is useful to drive a different type of conversation. The mistake isn’t choosing the “wrong” chart; it’s asking one chart to answer every question at once.
By separating stories—dominance, the long tail, and change over time—we reduce cognitive load and increase clarity. We also give ourselves flexibility: the same data can support different discussions, depending on what is important and which decisions need to be made.
You could even imagine a progression through these three views, with an overarching narrative to tie it all together. One chart doesn’t have to do it all. A thoughtful set of charts can tell a more effective story.
Meta lesson: start with the question or takeaway. Then choose a visual (or combination of visuals) that will help you make it unmistakeable.
To learn more about various graph types, check out the SWD chart guide.