Marketing automation software should be tested with a real customer journey and real data constraints. Feature grids rarely reveal implementation effort, delivery risk, or reporting gaps.
Define one journey
Use a journey that includes acquisition, consent, segmentation, at least one behavioral trigger, several messages, a conversion event, and a handoff to another system. Recreate the same journey in every candidate.
Score the system boundary
- Data model: contacts, companies, events, custom fields, and deduplication.
- Journey control: triggers, branching, delays, suppression, and testing.
- Delivery: consent, unsubscribe state, sender configuration, and monitoring.
- Integration: ecommerce, CRM, analytics, warehouse, and webhook support.
- Reporting: campaign outcomes, lifecycle visibility, and exportability.
- Migration: import validation, history retention, and rollback options.
Distinguish creation from execution
Generating campaign copy is not marketing automation. A platform must reliably act on data, respect permissions, send or trigger work, and preserve an auditable history.
That distinction is why our Marketing Automation Agent lab is classified as a watch-stage hypothesis. We want to learn whether teams need a focused campaign operator before committing to a CRM-scale product.
Use a go/no-go rule
Do not migrate if the candidate cannot reproduce the representative journey, preserve consent and data history, or supply the reporting required by the team. A lower subscription does not compensate for unreliable execution.