Fan-Out / Fan-In

Visual Flow

Rendering diagram…

When to Use This Pattern

Use fan-out/fan-in when:

• You have a collection of items to process (invoices, applications, line items)
• Each item can be processed independently of the others
• Processing takes significant time and you want to parallelize it
• You need to aggregate results after all items are processed (totals, summary, merged output)

How It Works

Rendering diagram…

Fan-Out: Split the input into individual items and launch parallel processing for each.

Fan-In: Wait for all parallel branches to complete, then combine the results.

Implementation Guide

Step 1: Prepare the Input Collection

Get the list of items to process:

• Query a database or list
• Parse a CSV or Excel file
• Receive an array from an API webhook

Step 2: Fan Out — Launch Parallel Processing

For each item in the collection, spin up an independent processing branch. Two approaches:

Approach A: Loop with child workflows

1. Iterate over the collection
2. For each item, start a child workflow (fire-and-forget)
3. Each child workflow processes the item and writes its result to a shared location

Approach B: Parallel branches (fixed count)

1. Use parallel branches in the workflow designer
2. Each branch handles a subset of items
3. Best when the number of items is small and known

Step 3: Process Each Item

Each parallel branch does the same work:

1. Validate the item
2. Call external services (API lookups, data enrichment)
3. Apply business logic
4. Write the result (to a list, database, or variable)

Step 4: Fan In — Wait and Aggregate

Wait for all branches to complete, then:

• Count successes and failures
• Sum totals (amounts, quantities)
• Merge outputs into a consolidated document or report
• Flag exceptions for manual review

Step 5: Handle Partial Failures

Not all items will succeed. Design for this:

Scenario	Strategy
1 of 50 items fails	Log the failure, continue with the other 49
Half the items fail	Complete what you can, send a summary of failures
All items fail	Abort and alert — likely a systemic issue

Example: Invoice Batch Processing

Fan-Out:

1. Query "Pending Invoices" from ERP → returns 50 invoices
2. For each invoice, start a child workflow:
   • Validate line items
   • Match to purchase orders
   • Check budget availability
   • Create approval task

Fan-In: 3. Wait for all 50 child workflows to complete 4. Generate summary:

• 45 invoices approved ($125,000 total)
• 3 invoices flagged for review
• 2 invoices rejected (PO mismatch)

5. Send summary email with links to each invoice

Tips & Best Practices

• Cap the parallelism. Don't launch 1,000 parallel branches. Process in batches of 10-20.
• Use a shared status tracker. Write each item's status to a SharePoint list or database table so you can monitor progress in real-time.
• Make each branch idempotent. If a branch fails and gets retried, it shouldn't create duplicate data.
• Time-box the fan-in. Don't wait indefinitely — set a maximum wait time and handle stragglers separately.