Three marketing automation rollouts that failed, and what we'd have done differently.
Most marketing automation rollouts fail. Not catastrophically. They just never produce the productivity gain or the pipeline lift that was promised in the kickoff deck. Six months in, the team is using 30% of the platform and complaining about the cost.
Here are three rollouts we studied where the failure was specific and the fix was identifiable.
Case 1: A 40-person team that bought an enterprise MAP and got slower
A B2B SaaS company at $30M ARR migrated to a full marketing automation platform. The rollout took 7 months, cost $400K in implementation, and produced a marketing team that took 3x longer to ship campaigns.
What went wrong: They implemented the platform as a feature list rather than a workflow. Every team member was asked to learn every module.
What we would have done differently: Implemented one workflow at a time. Lead capture and routing first. Email lifecycle second. Multi-touch attribution third. ABM personalization fourth.
Case 2: An ABM platform deployed without account selection criteria
A mid-market software company bought a $180K/year ABM platform. After 9 months, sales hadn't closed a single deal attributed to the platform.
What went wrong: They never agreed on what makes an account a "target account." Marketing was running campaigns against one list, sales was prospecting from another, and the ABM platform was orchestrating engagement against a third.
What we would have done differently: Before the platform purchase, a 4-week alignment engagement to define the target account list. Same 200 accounts for marketing, sales, and CS. The platform comes in as the orchestration layer for an agreed list.
Case 3: A workflow automation rollout that produced 47 brittle integrations
A growth-stage SaaS company hired a contractor to "automate everything." After 6 months, they had 47 Zapier workflows, none documented, all of which broke when any of the connected tools updated their API.
What went wrong: Workflow automation was treated as a quick win, not an infrastructure decision.
What we would have done differently: Before any workflow gets built, define the data model. What entities exist. What events exist. What system of record exists for each entity. Then build workflows against the model, not against the urgent need.
The pattern across all three failures
Every failed rollout had the same structural problem: the technology was implemented before the workflow was designed.
The four questions we ask before any rollout
1. What is the workflow this technology will own? If you can't describe it in one sentence, the rollout will fail.
2. Who is the system owner? A single person with authority to say no to feature requests.
3. What is the success metric in 90 days? Specific, measurable, time-bound.
4. What do we stop doing when this goes live? If nothing gets retired, the team is just being asked to do more work with a new tool.
If you can't answer all four before the contract gets signed, the rollout will follow the pattern of the three cases above.