How Salesforce Powers Marketing with Tableau Next | Dreamforce 2025

I'm going to hand it to my lovely co presenters real quick to introduce themselves.

Hi, my name is Emmy. I lead our data visualization team for marketing.

Hi, I'm Jennifer. I'm a lead data viz engineer on.

Our marketing team and my name is Casey. I help run our Tableau Next Customer Zero team here at Salesforce. And because this is all about Tableau Next Customer zero, Jen and Emmy invited me to come explain what that phrase means real quick and give you all some context. So let's start with what Tableau Next is. Hopefully you haven't made it this far into Dreamforce without seeing and maybe even getting a chance to play around with Tableau Next. But if it's new to you, it is the next version of the Tableau product. It is built on the Salesforce platform and deeply integrated with AgentForce and Data360, which means that it brings agent analytics and insights and action all into the flow of work and all into where all of your end users are already interacting with each other. And in fact, I don't have time to tell you everything about Tablet X right now, but I'll shout out my favorite feature, which is the Slack integration. I've been a tableau analyst for a long time. I've built a lot of dashboards in my day. And what I always hear from end users is there's too much stuff, there are too many links to remember, too many things to try and find. And the great thing about Tableau Next is it brings all of your key metrics, all of your important data and that agentic analytics experience all into Slack. So that's tableaux. Let's talk about the Customer Zero part. At Salesforce, we believe in drinking our own champagne, which means that we want to be the first and best user of every single feature and product that we roll out. And that's why I'm really excited about what Jen and Emmy did and what they're going to talk to you all about today. Because it's really a best in class example of being one of the first, earliest users of agentic analytics at Salesforce. So with that, I'll go ahead and hand it over.

Thank you, Casey. All right, so today we're going to just present to you how we answer questions. Cause that's our job as analysts, is we answer questions that users have. If you're also in the analytics world, all of these questions or some variation of them will sound very familiar to you. Usually what happens is when they become routine, we do what any good analyst does. We Build a dashboard. And then when we get variations of the same question, we add filters to the dashboard. But there always comes a point where questions get a little bit more complicated. Like this one. This one says which ous have the top MDP generation for AI and data this quarter. And if this doesn't mean anything to you, that's fine. It doesn't mean anything to most people. The reason that this question is very complicated is because you have to have a lot of prerequisite knowledge in order to try to answer it. You have to know what an OU is. OU is operating unit. And that's how we describe our go to market structure at Salesforce. You need to know that MDP is marketing driven Pipe, which is marketing's contribution to pipe generation. You have to know that when we said the top MDP we means that it's high. You also have to know that AI and data is one of our products. And you have to know that this quarter refers to one specific date and that Salesforce fiscal quarters and fiscal year is not aligned to the calendar year. That's a lot of information to know in order to answer all of this. So here's two ways that you can get an answer to this question. On the left hand side the you will see the traditional method. This is one of our dashboards. So you have to know a which dashboard to use and then it's a multi step process. You have to apply all of your filters, again knowing what each of them refer to and then find your answer all the way down. On the right hand side you have tableau next. In tableau next you just ask that question exactly how it was phrased and the agent will think about it, it will consult with its knowledge base and it will just give you the answer. It will even build you a graph. This is what we're presenting to you today. We're presenting to you how we build this and how it in the end is saving us a lot of time because you don't need a dedicated analyst or you don't need all of the pre training in order to use our dashboards to just get your answers. In order to present to you, I'm going to call Jen who actually built this for us. This is all live, these are all of our actual dashboards so she can tell us how she did it.

Thanks Ami. Does anyone here build dashboards in tableau cloud or in tableau desktop? Great. So I'll walk through our process, hopefully demystify it a little for you and call out where there are some similarities and differences. We're going to start with the data source and we're going to work all the way down to the agent force tableau analytics experience. Every good analyst knows you need quality data to get quality insights. So we'll first talk about data setup. On our demo we had a picture of how we do it today, which is our marketing cockpit on Tableau cloud that is supported by a snowflake data set that has been specially curated to support that dashboard. We didn't want to reinvent the wheel, we wanted to use the resources we had. So we took that snowflake table and we ingested it into data360. So that would enable us to be able to build in the Tableau next environment and utilize the semantic data model to create the Gentex experience. Once you have your data model objects loaded into data 360, you can build your semantic data model. I think of this as like training a new employee. It's where you set all of your business context and you do all of your knowledge, your business rules, maybe some jargon. In our semantic domain data model, you can join your different data model objects together, creating relationships. The more data you join in, the more questions you can answer the the better context the concierge has. To answer those questions, you need to enrich your data. That means cleaning up the raw data fields. As an engineer in the data source, we use a lot of abbreviations to identify what field to use. That doesn't make sense to our marketers. And so I need to go through and clean up what those field names are and add descriptions to explain how we get to those. We also need to add business preferences which are rules or logic that say, for example, mdp. What does MDP mean? Well, that stands for marketing driven pipe amount. And when someone types that into concierge task question, they need to know that they're getting the marketing driven pipe amount number. And that's a business preference that we've written in. So let's go back to that question that Emmy shared. Which ous have the top MDP generation for AI data this quarter? To set up our data, we have our operating unit field. We've set a clean name, it's text field. And I've added a description to say it's a combination of region and sales leader that we use to identify our OU for marketing driven pipe amount. That's our currency measure. When you sum it, it's a continuous metric and the more we generate, that's good. So we've set the sentiment up as Good. These are measures and dimensions in our data source in the semantic model. You can also build out your your parameters and calculations and metrics. Now let's talk about the visualization. This is how our users actually get the insights from that data that we've just curated and set up for them. What is a metric? So as I mentioned, in our semantic data model there are metrics. If you've built using Tableau Pulse, you'll be familiar with this concept. It's an intelligent KPI. It has your value, your measure. Marketing driven pipe amount charted over time with comparison time points and a natural language insight. We can say how we want to associate different attributes and dimensions to it. So here's our marketing driven pipe metric. I've associated operating unit and product, which is our APM1 cloud, to this field, to this metric and then I can set insights. So if I want to know that there are changes in trends or what the top contributors are, I can use my metric to find that. Once I've built it in the semantic model, it can be used by anyone who has access to building with that semantic model. You just build it once and now I can add it into my dashboards, multiple dashboards. My team can add it into their own dashboards and we can embed it in Slack and have canvases where people can access that data within their Slack environment. It's interactive. So once it's on a dashboard, like what's demoing here on the right, a user can drill into that metric and start filtering and slicing and dicing that metric for more insights to help them steer their analysis. Next is Vizzis. This is a chart or standard table, maybe a graph. It's a reusable element, much like a metric. So I can build a chart, e.g. mDP contribution by OU, the bar chart here. And I can now use it on multiple dashboards, just like a metric. And my team can also use it, which is a very different experience than a tableau cloud today, where you have to rebuild your visas if you need it in different dashboards. This is great for our marketers who don't want to go to multiple dashboards to see their metrics. They want to go to one place and get everything. And now our team can use the single source, viz, to get the insights on their dashboards. Here's an example of what it's like to build in Tableau. Next, it's actually built in Salesforce, but it looks a lot like the tableau desktop experience where you can drag dimensions and metrics to rows and columns. A bar chart's automatically created. You can customize with your marks card or select from the suggested charts that are available. So next we have the dashboard. We've built the design elements we need to have a dashboard. The dashboard is really just your canvas that holds those visual elements. And the experience here is different. It's more like CRM analytics than it is tableau. There's a grid canvas and you click an icon to add a metric. You drop the metric onto the grid and you select what metric you want to display. Like marketing driven pipe. Same for Vizz. You click an icon for the viz, you drop it to the dashboard canvas, you select what viz you've made that you want to show. And same with filters. Here's our marketing cockpit. So you can see that this marketing cockpit has our pipegen, our open pipe data. We also have attainment metrics. We've got contribution and marketing driven pipe. Every KPI here, every line chart is actually a metric that, that we've built in our semantic model. And every bar chart is a viz that we've built. All of these are interactive. A user can come here and click in and visit other, drill in more to the metrics or create copies of the visualizations here and customize them for their own view. And you might notice that we also have very few filters on this. We designed this in a way that we wanted to take advantage of this concierge experience. So we are addressing the highest most questions that we get from our users that the most users will have for this dashboard. Then anything that's like custom query that has very detailed niche requests that some people need, but not everyone will rely on the agent for that. And that's what this window is showing you here. You click the agent and it opens up the agent experience. So to do this we had, I would say that we had a lot of help from our marketers. We surveyed our marketers to understand how they were asking their questions and that helped us generate what language to use as we set this up. Because as an engineer I have different language for building my dashboards than someone who's asking for the data. And that's been really helpful to have us give use cases so we can build to it and also test to it. This is a very high level dashboard and our users were very excited to get their hands on it. And naturally the next question was, okay, now what I need to know, what are my top campaigns? What are my programs? How am I impacting this marketing driven pipe? And so we're continuing to do additions and enhancements to our semantic data model by adding in that data, joining it to our existing semantic model that we've built here. So so that we can have deeper conversations and insights with our data. We imagine that this semantic model development is going to be where we invest most of our time going forward and having to really enhance the context there. Okay, so we've built the foundation. Let's talk about the concierge experience. That experience of being able to ask questions. You saw a little demo of what a concierge can answer. Here are the different types of questions that you can ask. You can ask descriptive questions. The what, when, where questions. Which product is driving this? Where is my investment having the most return on my where am I getting the most return on my investment? What top campaigns should I be focusing on? You can do comparative questions looking for similarities and differences. Are there differences in how marketing drives pipe versus how we mature ACV deals? Where are their differences across regions or even different campaigns? We can also do exploratory questions so things like show me the top 10 campaigns that are driving changes or growth this year. Show me a breakdown of product for Amer specifically. And you don't have to only ask questions. You can actually request it to build things for you. So if you have a dashboard that doesn't have the vis you need on it and you know that you're going to have a conversation with a leader and you want to show a week over week trend of how your MDP is changing, you can request that concierge build that for you. And that's true for trend lines and it's true for bar charts as well. All of this makes it a lot easier and quicker for our marketers to get their insights and then start taking action on those insights and which drives our MDP growth.

Awesome. I have a few parting thoughts. So the first one is building the next generation of analytics for anybody who is currently in analytics is going to look very different from what you're used to. It's a complete mindset shift for anybody who builds dashboard because more than ever it's all about data prep. You will need to spend most of your time preparing the semantic layer and preparing the data, but also studying your stakeholders to understand how they're talking, study their language and design your semantic layer for conversation. It's very different from what we're doing today where you just have some fields with some names and they just kind of have to learn like how to talk to the dashboard. The second thought, I got it. Thank you. We are shaping our future at Salesforce and it's an amazing opportunity and we're also shaping yours. So if you have any feedback, you can come and talk to us. But we're working really closely with the product team while developing this proof of concept to influence the roadmap and to give them suggestions like, hey, we think it would be great if we could build this or here is one feature that's really nice in the current tableau environment that we would like to reproduce in Tableau next. And they've been really open to all of this feedback, so it's been a great adventure. Then finally we're just overall really excited about this product and we think you should be too. We're in this phase of the product growth where every month there's going to be new features that are released. The next thing that we're working on is this integration with SLOW that Casey presented at the beginning. That's going to be the next thing that gets released to our stakeholders and we think they're going to love it. And then finally we're excited about getting some time back. The agent capabilities as we presented at the beginning should free up time for our analysts because they will not have to answer those one off questions for leadership in particular. They will just be able to self serve. And we can keep our analyst time dedicated to either design work, which we know we're going to need to do more of, or deep dives like getting into the more needy, gritty questions and having their time better spent. And that's it for the time that we had today. Thank you everyone for joining us and we hope you all enjoy building on TNEXT very soon.