How you present data has a big impact on how people interpret it. In the run-up to this month's US presidential election, opinion poll averages showed a 1.5% lead in the popular vote for Harris. But it was misleading to present this as a simple line chart, argues John Burn-Murdoch, Chief Data Reporter at the Financial Times. Showing the full range of the polls would have made it clear there was a strong chance the result would go the other way -- as it did, in the event. This was a case of poll trackers, he goes on, leading people to "draw misleading conclusions from inherently uncertain data that is presented with false precision."
The point would have been well received at last week's DataFam Europe conference in London. Data analytics vendor Tableau, part of Salesforce, had gathered together some of its global community of four-and-a-half million data wrangling and visualization enthusiasts, all dedicated to conveying data as effectively as possible. Interest is soaring to new highs, fueled by growing access to an expanding supply of datasets, the potential of AI to accelerate analytics in various ways, and a corresponding hunger among business colleagues to consume data analysis and visualizations. Tableau is eager to meet that demand, as Ryan Aytay, its CEO, tells me:
Our mission over time is to make it that Tableau is just accessible for everyone. It just shows up wherever you work, whether it's embedded in Salesforce or some other application, and it helps you to get your job done. You don't have to be an expert. We're still going to have our experts, but we are broadening the aperture of who can use Tableau and just show up. It'll just work, versus you having to look for it.
While that means Tableau's user base is rapidly expanding beyond its traditional following among data geeks, that doesn't mean those business users need to become data scientists, nor that the data specialists no longer have a role. Southard Jones, Chief Product Officer at Tableau, likens them to chefs, working in the data kitchen to prepare ready-to-eat takeaways:
Not everyone needs to be a cook to feed the world. A certain number of people need to understand how to bring ingredients together and provide a meal that is healthy and going to not make you sick. But that doesn't mean every person who eats needs to be able to cook either.
Aytay adds:
How do you democratize data is not necessarily me [as a business leader] becoming data literate. I want to democratize data because I do think that every company has a real opportunity to be more data-driven in all their decisions. Most companies we speak to, they're like, 'Well, we're making decisions, but we actually don't have all the right information.' That's not great. Our job with Tableau is, how do we empower them to make the right decision? And for those people [the analysts] who need to be data literate, we're going to help them as well.
The toolbox that Tableau offers has been rapidly expanding over the past few years. Visualization has always been its core strength, and there were plenty of examples on display last week. Some visualizations entice the viewer to engage with the data, such as Annabelle Rincon's visualization of the letters Van Gogh wrote to family, friends and colleagues in each year of his life, which emulates the artist's own art. Others are designed to surface information as succinctly as possible, so that executives at businesses as diverse as telecoms company Virgin Media O2 and The Economist news magazine can quickly monitor trends or take decisions.
A more recent innovation to Tableau is the addition of a semantic layer that provides more flexible, automated connection of data from different sources. Once this layer has been set up to align data coming from disparate date sets, it's then much easier for the business user to simply ask for the data analysis they need, with AI helping to interpret their questions and fetch results. Jones explains:
A business person doesn't have to worry about how that raw data is structured. They can ask their question in natural language and get a response back that's trusted. That semantic layer can abstract away the challenges of the underlying raw data and give a trusted answer back. And then they go, 'How did you get to that?' Then it can explain where did it come from? Here are the table sources. Here's the actual raw data that it came from. I can show you how they got to that answer.
AI can also help create and maintain the semantic model. He goes on:
Leveraging AI, we can help build the semantic model, make suggestions. You can also have an agentic data steward that's looking across your semantic model, saying, 'Is everything right? Looks like two people created the same metric. They may have named it differently, but under the covers, it's analyzing the same data,' and would surface that to the person, the data steward and say, 'Hey, it looks like you might need to have some conversations to resolve this.' That's where AI, or an agent, can help a semantic model scale to large businesses, or cross between use cases and be usable, whereas today's manual semantic models are refined to small use cases.
Generative AI is more accurate when grounded in a semantic layer that puts the data in the right context, explains Aytay. Now combined with Salesforce's Agentforce platform, the data can be linked to follow-up actions. He sums up:
The new version of Tableau is now starting to come to life with the semantic layer. There's a lot of generative AI capabilities in the market, but what's most important is when you can get the context around the data, make sure it means something...
Tableau is way more than just a visualization tool now. See and understand, take action, become more efficient, and bring context to your data, I think, is a big part of where we are.
Bringing the Tableau community into the agent world was one of the primary goals of the DataFam event. Aytay goes on:
We need to show them that, just because you're an analyst building visualizations, you can still do that, but now you can do it better, you can do it faster, and your career path will include that. You'll also become agent builders. You'll be visualization builders, you'll be agent builders, and you will have a better path that way.
It means a lot of change and new learning, but on the whole the community is embracing the opportunity, says Jones. AI agents can also help them work faster by taking care of some of the repetitive chores that previously have had to be done manually. He says:
They want to deliver value to their business users, their business partners. If we, through an Agentforce platform, can enable them to skip some of those administrative things -- build dashboards in a more efficient way, build a viz that is embedded inside someone's flow of work, create a semantic layer so there's someone going to ask a natural language question -- they're happier. They'd much rather be doing that than, I'm going to spend so much time doing a bunch of administrative things that you had do two or three years ago as part of getting data ready for analysis...
Will people continue to create artistic dashboards? Yeah, because that's part of what makes Tableau great. A machine isn't going to create that. You need a human to be able to have some sort of artistic license to create that. It's not going to go away. It's just going to be their path to create and enable more business users can be accelerated -- and support more and more people with the same number of analysts.
Tableau is caught at the nexus of several powerful trends, which are bringing rapid changes that it has to lead its huge community through. One trend is the growing accessibility of various types of data as enterprises join together existing data stores and tap new data sources. At the same time, the latest developments in AI make it easier to interpret and wrangle the data, understand what people are asking for, and then connect it to the actions that people need to take.
While these trends make it easier to marshall, make sense of and present data without having to wrangle it through a jungle of spreadsheets, the role of the data analyst remains significant. They are the stewards of data quality, responsible for getting reliable, accurate and meaningful data in front of the people who need to see it, as well as tying it into next best actions where that's appropriate. Ultimately, tools like Tableau may well end up embedded in more everyday products that people use rather than them needing to visit a separate app, but they will still be there behind the scenes, doing the crucial work of ensuring the right data gets to the right people at the right time.
