CRMA Academy

Run and Troubleshoot the Dataflow

Execute the default dataflow to build your dataset, monitor it in Data Manager, and fix the most common first-run error — objects that haven't been synced yet — by running Connect manually.

Run and Troubleshoot the Dataflow

You clicked Create Dataset, which executes the dataflow. Now let's watch it run, understand what it's actually doing, and fix the error that catches almost everyone the first time.

What "run the dataflow" really does

When Dataset Builder writes to the default dataflow, it doesn't just build your dataset — it runs everything in that dataflow. In a training org that's not empty: it comes pre-packaged with other datasets and instructions. So running the default dataflow executes all of it, not only the opportunities dataset you just designed.

This is exactly why experienced builders eventually create their own dataflows. A single shared default dataflow means every build re-runs everything. As you grow, you'll split work into named dataflows (or recipes) so each one refreshes a known set of datasets on its own schedule.

Head to Data Manager → Monitor and you'll see your default dataflow running.

The classic first-run error

On a brand-new org — or the first time you run this — the dataflow may fail with an error. Don't panic. The cause is almost always simple: the dataflow is asking for objects that haven't been pulled into the sync layer yet.

Remember the chain: the dataflow reads from Connect/Sync, not from Salesforce directly. If Connect hasn't populated Account, Opportunity, and User, the dataflow has nothing to read and it errors out.

The fix: run Connect manually

  1. 1

    Go to Connect

    In Data Manager, switch to the Connect tab where your synced objects live.

  2. 2

    Run everything once

    Open the dropdown and manually run the sync. This pulls the data from Salesforce into the staging layer — the purple layer on the landscape diagram.

  3. 3

    Let the dataflow re-run

    With the objects now synced, run the default dataflow again. It reads the freshly staged data and builds your dataset successfully.

When everything is scheduled properly, this is automatic: the dataflow asks the sync layer "is the data here?" and, if not, Connect kicks off and brings it in before the dataflow proceeds. You only do this by hand when an environment is out of sync — typically a fresh training org on day one.

Forgot a field? Edit the digest node

Because you can't reopen the Dataset Builder field picker, editing fields later happens in the dataflow's digest nodes.

  • Open the dataflow, click the relevant digest node (say, Account), and re-add the field via Select Fields.
  • Save, and when prompted, choose to propagate the field all the way down to the final register node — otherwise it stops partway and never reaches your dataset.
  • Update and run the dataflow again so the new field is materialized.

We cover digest nodes, field attributes, and dataflow editing properly in the Data Manager lessons — this is just enough to unstick you.

You've built your first dataset

Once the dataflow finishes green, hop back to Analytics Studio and there it is — your opportunities dataset, sitting in your private app, ready to explore. On the landscape diagram it now lives in the dataset layer, built by a dataflow that reads from sync.

Mental checklist when a dataflow fails: (1) Are the source objects synced? (2) Did Connect actually finish before the dataflow ran? (3) Does every field you need propagate to the register node? Nine out of ten first-run failures are answered by those three questions.

Dataset Builder is friendly, but it locks you into lookups and a dataflow. Next we build the exact same dataset with a recipe — more clicks up front, far more control, and the direction the platform is heading.

Discussion

No comments yet — be the first to start the discussion.