Recipe
A Recipe is a Data Prep tool in CRM Analytics (formerly Tableau CRM, originally Einstein Analytics) that prepares, cleans, and combines data into a dataset for analysis.
Definition
A Recipe is a Data Prep tool in CRM Analytics (formerly Tableau CRM, originally Einstein Analytics) that prepares, cleans, and combines data into a dataset for analysis. You build it on a visual canvas as a chain of connected nodes, where each node reads the rows from the node before it and passes its result onward.
A recipe starts with one or more input nodes that bring in source data. Middle nodes join, filter, aggregate, and transform that data. The recipe ends with an output node that writes the result to a dataset, a CSV file, Data Cloud, or an external system like Amazon S3 or Snowflake. Running the recipe produces or refreshes the data that dashboards and lenses then query.
How a Recipe shapes data inside CRM Analytics
Where Recipes sit in the CRM Analytics pipeline
CRM Analytics separates the work of preparing data from the work of visualizing it. Raw records live in Salesforce objects, external systems, or datasets that a connection already loaded. A recipe is the step that turns that raw input into a clean, analysis-ready dataset. Dashboards and lenses never query Salesforce objects directly. They query datasets, and recipes are the main way those datasets get built. You create and edit recipes inside Data Manager, reached from Analytics Studio or the Analytics tab. The recipe opens on a visual canvas where nodes appear as connected tiles. You add a node by clicking the plus button between two existing nodes or at the end of the chain. Because the canvas is declarative, you describe what each step should do rather than writing code for how to do it. This makes recipes approachable for admins and analysts who are comfortable with data but not with scripting. The trade-off is that complex logic can spread across many nodes, so naming and layout matter for anyone who has to maintain the recipe later.
The node types you actually wire together
A recipe is made of nodes, and each node has a job. The input node brings in source data from a dataset, a Salesforce object, or an external connection. The join node adds related columns from another input, with lookup, left, right, inner, full outer, and cross join options. The filter node drops rows that fail a condition, so you only carry forward the records you want. The aggregate node rolls rows up to a higher grain, like turning line items into per-account totals. The append node stacks rows from two inputs that share the same shape, which is handy for combining this year and last year. The update node swaps values in one column using values from another. The transform node is where most field-level work happens, holding one or more transformations that run in sequence. The output node ends the recipe and writes the result somewhere. You can branch a recipe so one input feeds several paths, then bring those paths back together with a join or append before the final output.
What a Transform node can do to a column
The transform node is the workhorse for changing individual columns. The most common transformation is a custom formula that creates a calculated column from an expression, using date, numeric, string, window, and multivalue functions. You also get type conversions, so a text field that holds a date or number can become a real date or measure that math and charts understand. Beyond formulas, the transform node offers bucketing to group values into named ranges, date math to add or subtract days and months or find the difference between two dates, and edit attributes to rename columns or relabel values. Several transformations bring machine intelligence into the prep step. Cluster segments rows into groups, detect sentiment scores the tone of text, predict missing values fills gaps, and time series forecast projects future measures. You can also pull Einstein Discovery predictions into the dataset. Stacking transformations inside one transform node keeps related cleanup together and easy to read.
Inputs, outputs, and where the data lands
Every recipe begins with input and ends with output, and both ends are more flexible than they first appear. An input node can read an existing CRM Analytics dataset, pull directly from a Salesforce object, or read from an external connection that has already been set up. Pulling straight from an object means you do not always need a separate load step before the recipe runs. The output node decides where the prepared data goes. The usual target is a CRM Analytics dataset, which dashboards and lenses then query. A recipe can also write to a CSV file, to staged data, to a Data Cloud data lake object, or to an external location such as Amazon S3, Snowflake, or Azure. One recipe can have more than one output node, so a single prep run can publish the same shaped data to a dataset for analysts and to a warehouse for other teams at the same time. Choosing the right output keeps you from rebuilding the same logic in two places.
Previewing, running, and scheduling a Recipe
A recipe is not useful until it runs, but you should look before you run. As you build, the canvas shows a preview of sample rows at the node you have selected. That preview lets you confirm a join matched the rows you expected or a formula produced the right values, all before you process the full data volume. Catching a mistake in preview is far cheaper than discovering it in a published dataset. When the logic looks right, you save the recipe, then run it. Running can be manual, on demand, or scheduled to repeat on a cadence that matches how fresh the data needs to be. Scheduled runs are what keep a dashboard current without anyone touching it. You watch run status and history in the monitor area of Data Manager, where failed jobs show errors you can act on. For deeper troubleshooting, the recipe inspector helps you trace how data moved through each node so you can find where a result went wrong.
Recipes versus the older Dataflow
Recipes are not the only way to prepare data in CRM Analytics. The older tool is the dataflow, a JSON-defined pipeline of transformations like sfdcDigest, augment, and computeExpression. Dataflows still run, and plenty of orgs have them, but the recipe is the modern visual experience that Salesforce builds new capability into. Many recipe nodes map to dataflow ideas. A join node does what augment did, a transform formula matches computeExpression, and a transform relative calculation matches computeRelative. The difference is the experience. A recipe shows the pipeline as connected tiles with live previews, while a dataflow is closer to editing configuration. If you are starting fresh, build with recipes, since they get the newest node types and transformations and are easier for a teammate to read months later. If you inherited dataflows, you can keep running them, and you can rebuild the important ones as recipes when you want the visual canvas and the newer features.
How to create a Recipe in CRM Analytics
Create a recipe in Data Manager to combine and clean source data into a dataset that dashboards can query. You build the pipeline on a visual canvas, preview as you go, then save and run.
- Open Data Prep
In Analytics Studio, go to Data Manager and choose Recipes, then click Create Recipe. The visual canvas opens with a prompt to add your first input.
- Add input data
Add one or more input nodes. Select a dataset, pull directly from a Salesforce object, or read from a configured external connection to bring source rows onto the canvas.
- Join, filter, and transform
Click the plus button to add nodes. Use a join to add related columns, a filter to drop unwanted rows, and a transform node to build calculated columns, convert types, or bucket values.
- Preview the result
Select a node to preview sample rows at that point in the pipeline. Confirm joins matched and formulas computed correctly before processing the full data volume.
- Add an output node
End the recipe with an output node that writes results to a dataset, a CSV file, Data Cloud, or an external location like Amazon S3. A recipe can have more than one output.
- Save, run, and schedule
Save the recipe, then run it on demand. Set a schedule so it refreshes on a cadence, and watch run status in the monitor area of Data Manager.
At least one source of data: an existing dataset, a Salesforce object, or an external connection that has been set up.
A target for the prepared data, such as a dataset name, so the recipe has somewhere to write its result.
A clear, descriptive label that tells teammates what the recipe prepares and which dataset it feeds.
- A recipe cannot save or run without an output node, so the pipeline must end somewhere.
- Joins can multiply rows when keys are not unique on the lookup side, which inflates measures downstream; check row counts in preview.
- Convert text that holds dates or numbers to real date and measure types, or charts and math will not work as expected.
- A dashboard reads the dataset, not the recipe, so the dataset only changes after the recipe runs again.
Prefer this walkthrough as its own page? How to Recipe in Salesforce, step by step
Trust & references
Cross-checked against the following references.
- Nodes for Data Prep RecipesSalesforce
- Clean, Transform, and Load Data with Data PrepSalesforce
Straight from the source - Salesforce's reference material on Recipe.
- Create a Recipe with Data PrepSalesforce
- Transformations for Data Prep RecipesSalesforce
Hands-on resources to go deeper on Recipe.
About the Author
Dipojjal Chakrabarti is a B2C Solution Architect with 29 Salesforce certifications and over 13 years in the Salesforce ecosystem. He runs salesforcedictionary.com to help admins, developers, architects, and cert/interview candidates sharpen their fundamentals. More about Dipojjal.
Test your knowledge
Q1. What is a Recipe in Salesforce CRM Analytics?
Q2. Where does a Recipe sit in the CRM Analytics workflow?
Q3. Which operations can a CRM Analytics Recipe perform on data?
Discussion
Loading discussion…