How to Use Einstein Discovery in Tableau

Bringing AI-powered predictions directly into your dashboards is no longer a futuristic dream. With Tableau's integration of Einstein Discovery, you can move beyond simply visualizing what happened and start understanding why it happened and what you can do about it. This guide will walk you through setting up and using Einstein Discovery within your Tableau workbooks, transforming your static reports into dynamic, predictive tools.

What Exactly is Einstein Discovery?

Before connecting it to Tableau, it helps to understand what Einstein Discovery is. It's Salesforce’s augmented analytics tool designed to bring an AI-driven data scientist to every business user. You don't need to know R or Python to use it. Instead, you feed it a dataset and tell it what business outcome you want to improve - like maximizing revenue or minimizing customer churn. Einstein Discovery then goes to work.

In minutes, it analyzes millions of rows of data to find statistically significant patterns, correlations, and insights. It delivers this information through four key capabilities:

• Descriptive Analytics: It explains what happened in your data, uncovering patterns you might have missed.
• Diagnostic Analytics: It tells you why something happened, identifying the key drivers behind the trends.
• Predictive Analytics: It uses statistical models to predict what is likely to happen in the future.
• Prescriptive Analytics: It suggests specific actions you can take to improve the predicted outcome.

Think of it as an automated workflow for statistical analysis that presents its findings in plain English. Your outcome is a "story" – an interactive collection of charts and explanations – and a deployed predictive model that you can use in other tools, including Tableau.

Setting Up Einstein Discovery for Tableau

Getting the two platforms to talk to each other involves a few setup steps. You'll need the right licenses and permissions in both Salesforce and Tableau to make the magic happen.

Prerequisites

Before you start, make sure you have the following in place:

• A Salesforce org with a Tableau CRM Plus or Einstein Predictions license.
• Tableau Desktop, Server, or Cloud (version 2021.1 or later is recommended).
• Admin permissions in both Salesforce and Tableau to configure connections and install extensions.
• A deployed Einstein Discovery model (more on that below).

Step 1: Create and Deploy a Model in Einstein Discovery

Your journey begins in Salesforce, not Tableau. The first step is to create a "story" to analyze your data and produce a predictive model.

Imagine you're trying to figure out why some sales deals have a longer cycle time than others. Your goal is to minimize 'Days to Close'.

1. Prepare Your Dataset: You’ll start with a dataset in Tableau CRM (formerly Einstein Analytics/Wave) that contains your historical sales deal information. This dataset should include your outcome variable ('Days to Close') along with dozens of potential explanatory variables (e.g., 'Lead Source', 'Deal Size', 'Region', 'Industry', 'Product Type').
2. Create the Story: Inside Tableau CRM Studio, you'll create a new story. You'll select your dataset, specify that your goal is to minimize the 'Days to Close' field, and let Einstein Discovery take over.
3. Analyze the Insights: Einstein will return a story explaining what factors have the biggest impact on deal cycle time. It might tell you something like "Deals originating from the 'Webinar' lead source tend to close 15 days faster than average" or "Deals in the 'Manufacturing' industry take 22 days longer." This diagnostic phase is invaluable for understanding the why.
4. Deploy the Model: Once you are satisfied with the insights, you can deploy the underlying model from the story with just a few clicks. This makes your predictive model a "live" asset that can be called upon by other applications, like your Tableau dashboard.

Step 2: Connect Tableau to Your Salesforce Data

Next, you’ll open Tableau Desktop. If you plan to analyze Salesforce data directly in Tableau visualizations (in addition to getting predictions), you’ll need to connect to it.

• In Tableau Desktop, under "To a Server," select "Salesforce."
• You'll be prompted to log in to your Salesforce account using your credentials. Authenticate the connection.
• Once connected, you can choose which objects (e.g., Opportunities, Accounts, Campaign data) to pull into Tableau to build your vizzes.

Important Note: You don't necessarily have to be visualizing data directly from Salesforce. Your Tableau workbook can be based on data from any source (like a spreadsheet or a SQL database), as long as its fields can be mapped to a deployed Einstein model.

Putting Predictions Into Action in a Tableau Dashboard

With a deployed model ready to go, you can now integrate it into a Tableau dashboard to get real-time predictions and recommendations. This is done using the Einstein Discovery Dashboard Extension.

Step 1: Add the Einstein Discovery Extension

Open the Tableau dashboard where you want to add predictive insights.

1. From the "Objects" panel on the left of your dashboard screen, drag the "Extension" object onto your dashboard canvas.
2. In the "Add an Extension" dialog box, you can search the Extension Gallery for "Einstein Discovery." Select it and add it.
3. You'll get an initial configuration screen for the extension. This is where you connect it to your Salesforce org and choose which model you want to use.

You’ll be prompted to log into Salesforce again here to authenticate the extension itself. You'll then see a list of your deployed Einstein models. Select the 'Days to Close' model you deployed earlier.

Step 2: Map Your Tableau Fields to the Model

This is the most critical part of the setup. You need to tell the extension how the fields in your Tableau sheet relate to the data the Einstein model was trained on.

Let’s say your selected Tableau worksheet contains fields like 'Deal Size' and 'Lead Source,' which the Einstein model also uses. You'll need to map them one-to-one.

• The extension's configuration pane will show you the fields required by the Einstein model on one side and the fields available from your Tableau worksheet on the other.
• For each model field, select the corresponding field from your Tableau worksheet. For example, map the model's "Deal_Size_Amount" to your Tableau worksheet's "Amount" field.

If you don't have a direct data match for one of the model's required fields (for instance, the model was trained with 'Region' but your current Tableau view doesn't have it), you can opt to use an average or default value. However, accuracy is always highest when the data matches what the model was trained on.

Step 3: Get Live Predictions and Recommendations

Once the mapping is complete and you save the configuration, the extension comes to life. Now, when you click on a mark in your Tableau view – like a specific salesperson's bar in a chart of open opportunities – the extension works instantly:

1. It sends the data from that single mark (e.g., salesperson 'Jane Doe', deal size '$50,000', industry 'Retail') to your deployed Einstein model in Salesforce.
2. The model runs a real-time prediction based on that specific combination of data points.
3. The prediction is sent back and displayed directly in the Einstein Discovery panel in your Tableau dashboard.

For our sales rep example, the panel might show:

• Predicted Days to Close: 68 days (12 days higher than average).
• Top Positive Factors: Factors driving a shorter cycle (e.g., "Lead Source is 'Partner'" has a positive effect).
• Top Negative Factors: Factors causing the delay (e.g., "Industry is 'Retail'" is the main negative contributor).
• Prescriptive Recommendations: What can be done to improve the outcome (e.g., "Add an executive sponsor to the deal" could potentially reduce the cycle by 8 days).

Suddenly, every data point on your dashboard is actionable. Your sales leaders aren't just looking at reports, they are diagnosing in-flight deals and getting AI-driven advice on how to accelerate them.

Best Practices for Using Einstein Discovery in Tableau

To really thrive with this integration, keep these practical tips in mind.

• Start With A-C-E: Your project should always start with a clear Action that you want to enable, supported by a Clarifying Question you need to answer and the Experience your users need to have. For example: "I want to enable Sales Managers to act on deals at risk of slowing down (the Action) by showing them which deals are predicted to have a long cycle and why? (the Question) in their existing opportunity dashboard (the Experience)." This A-C-E model keeps your focus on building solutions that have business impact.
• Train Your Model on High-Quality Data: The old adage "garbage in, garbage out" has never been more true. A predictive model is only as sound as the data it was trained on. Before a project ever starts, invest time in cleaning and preparing your source dataset in Tableau CRM or your source system. Make sure it's accurate, consistent, and has enough historical records to learn from.
• Don't Treat the Model as a Black Box: Take the time to review the insights from the Einstein story. The charts it produces for each variable are invaluable for understanding why the model makes the predictions it does. If a top factor seems completely counterintuitive, investigate your data. This builds trust in the model and helps you explain its output to stakeholders.
• Iterate and Improve: Your first model won't be your last. As business conditions change and more data becomes available, plan to revisit and retrain your models periodically. This ensures their predictions remain relevant and accurate over time.

Final Thoughts

By embedding Einstein Discovery's predictions directly into Tableau, you fundamentally change the nature of your dashboards. They evolve from static historical records into forward-looking, interactive tools that explain the 'why' and suggest the 'what’s next.' Business users can now make smarter, data-driven decisions on the spot, without ever leaving their familiar reporting environment.

Making sense of your data is becoming much more accessible. At Graphed you’re focused on this from the ground up, specifically for marketing and sales teams drowned in data from countless platforms. We let you connect sources like Google Analytics, Shopify, and your CRM in seconds and then simply ask questions in plain English to build real-time reports and dashboards. No more complex setups or steep learning curves - just instant answers to your most important business questions.