Why Steps?
- Modularity — Each step does one thing well
- Debugging — See exactly where issues occur
- Readability — Understand the workflow at a glance
Example: Lead Qualification
Step 1 — Fetch leads:
Get new leads from @HubSpot created in the last 24 hours
For each lead, get company info from @Apollo
Score each lead:
- Company size (1-10)
- Industry match (1-10)
- Title seniority (1-10)
If score > 25:
- Add to "Hot Leads" in @HubSpot
- Alert #sales on @Slack
If score <= 25:
- Add to nurture sequence
- Tag as "nurture"
Data Flow
Each step automatically accesses outputs from previous steps:
Using the leads from Step 1...
Take the enriched data from the previous step...
Filter the scored leads from Step 3...
Script wires up the data flow automatically.
Conditional Logic
If/then
If score > 25, add to hot leads
Otherwise, add to nurture
Early exit
If no leads found, skip remaining steps
Loops
For each lead:
- Enrich
- Score
- Update CRM
Error Handling
Per-step
If enrichment fails, log error and continue with next lead
Workflow-level
If critical step fails, alert #ops and stop
Retry
If API fails, retry 3 times with backoff
Tips
- One thing per step — “Fetch leads” not “Fetch, enrich, score, update”
- Validate early — Check required fields before processing
- Plan for failure — What happens if a step fails?
Advanced
Parallel processing
Process leads 1-50 and 51-100 in parallel
Combine when done
Checkpointing
Save progress every 100 items
Resume from checkpoint if interrupted
Next
