Bringing AI predictions to Tableau with Einstein Discovery
Editor's Note: the Tableau Extension Gallery is now Tableau Exchange.
With the upcoming Tableau 2021.1 release, we’re delivering the first integration of Salesforce’s artificial intelligence (AI) and machine learning (ML) capabilities in Tableau. This introduces an exciting new class of AI-powered analytics: Tableau Business Science. We’re bringing powerful data science techniques closer to the business, beginning with Einstein Discovery in Tableau. Users across the organization are now enabled with powerful predictions and recommendations to make smarter decisions faster. There are three ways to leverage the core ML technology of Einstein Discovery in Tableau—all with no coding required:
- Einstein in Tableau Calcs: Einstein Discovery is now integrated into Tableau’s calculation engine. Using the prediction capability is as simple as creating a calculated field with the new Einstein Discovery connection type. Drag and drop your predictive calculation onto your viz to see intuitive predictions.
- Einstein Discovery dashboard extension: Augment Tableau dashboards with predictions, explanations, and suggestions to improve predicted outcomes from Einstein Discovery, right within Tableau vizzes. Anyone viewing the viz can get a glimpse into the future, and even suggested actions to take.
- Bulk scoring and predictive insights in Prep Builder: Add a prediction step in your prep flow and bring prediction factors and improvements into your data set in bulk.
With these integrations, you’ll not only be able to understand what happened with historical data, but you’ll be able to bring the future into today’s decisions with predictions of what might happen and actionable recommendations you can take now.
Einstein Discovery brings AI-powered analytics to more decision makers
Einstein Discovery is Salesforce’s business-first ML platform that learns patterns from historical data which can be used to predict future outcomes. Analysts with domain knowledge of their business can build and deploy predictive models code-free. Users start by loading data into Einstein Discovery and then define a business metric they want to maximize (like sales or profits) or minimize (like patient waiting times). Einstein Discovery then presents automatically detected patterns in the data and a predictive model that can be evaluated for its performance. In fact, transparency and ethics are built into the product’s foundation, with bias protection, predictive factors, and live model monitoring so that users can trust Einstein’s predictions. There are no black box algorithms used here, ensuring safe AI and control in the hands of business experts.
The experience is designed to make analysis and predictions easily interpretable in the context of a business question. If the model fits your analysis and business context, users can deploy it to generate predictions on new data. Part of this deployment includes automatically generating an analytics extension calculation that can be copied into Tableau.
Bring predictions directly into calculated fields
Users can now create table calculations that add Einstein ML predictions directly into Tableau visualizations. Powered by the analytics extension, you can build calculations in Tableau that pass data from a viz to Einstein Discovery, and get results dynamically. These visualizations can be shared in workbooks or published to Tableau Server. (This capability is coming soon to Tableau Online.) It is as simple as creating a calculated field using the Einstein Discovery analytics extension and copying and pasting a script that is auto-generated in Einstein Discovery.
Delivering predictions dynamically in Tableau visualizations allows teams to use Tableau dashboards as tools to bring predictions into the context of their current analytics workflows. Teams building predictions have a quick, maintainable, and scalable way to create a visual front-end for their models that other business users can run with and explore. Users see results on the latest data every time they view the viz. Data can be filtered to view specific subsets and parameters can be passed to allow dashboard users to test different scenarios or assumptions with the model.
Available in Tableau 2021.1 Desktop and Server.
Create dashboards with dynamic predictions with the Einstein Discovery dashboard extension
Once you have built a model in Einstein Discovery, you can also bring it into your workflow within dashboards, giving users access to on-demand, interpretable ML predictions. Access this dashboard extension in the new in-product extension gallery, drag and drop it onto your dashboard, and then log into Salesforce when prompted.
You can then choose one of your deployed Einstein Discovery predictions and connect the model directly to a table of data, or to a set of parameters where you can perform what-if scenarios. Click on one or more rows of data from a Tableau visualization to get dynamic predictions and ways to improve the outcome based on the model.
Enrich data sets with powerful ML models with bulk scoring in Prep Builder
Arriving in the 2021.1.3 April release of Tableau Prep Builder, you will be able to write ML predictions (scores) from Einstein Discovery directly into your data sets. This integration allows you to score data in bulk including writing in key drivers of predictions and improvements. Combined with the automation of flows in Prep Conductor arriving in Tableau Online and Server in 2021.2, you’ll be able to operationalize and scale predictions across your organization.
Einstein Discovery predictions are integrated into Tableau Prep, providing users with ML-powered predictions right where they clean and shape their data sources. All users, including non-technical ones, can get top predictors and improvements written directly to data sources with just a few clicks. With Einstein Discovery in Prep, dashboard authors no longer need to set up prediction integrations and can avoid repetitive tasks.
Get started with Einstein Discovery in the latest version of Tableau
With Einstein Discovery in Tableau 2021.1, you can provide dynamic predictions and recommendations to anyone in your organization, helping them bring the future into today’s decision-making using ML. Integrated in Tableau dashboards, you can create powerful predictive models without writing code using the Einstein dashboard extension, and build table calculations with dynamic, contextual predictions from these ML models using Einstein in Tableau Calcs. Relevant predictions are also integrated in Tableau Prep Builder, offering insights during data preparation and giving people the power to write ML predictions (scores) directly into their data sets. You'll get transparent explanations of the key drivers behind predictions made based on your historical data, and suggested actions to iterate on and improve outcomes. These expanded AI capabilities provide richer analysis to analysts and business users, so that they can make smarter decisions in their everyday work and improve business outcomes.
Learn more about Einstein Discovery in Tableau and see it in action with a quick demo.
Thank you to our wonderful customers and community members. Your success drives our innovation and your feedback helps us build purposeful, thoughtful improvements. We appreciate your continued partnership and inspiration.