The strategy decks show the trough math, the migration percentage, and the recovery curve. They do not show what breaks. This is what broke at a $40M ARR SaaS 21 months into a SaaS-to-agents transition, on the day the per-seat tier actually sunset.
The five breaks
- Migration drift on accounts the dashboards said were fine. Twelve accounts had migrated on paper and were paying the new committed minimums, but their actual usage was below the minimum, so they generated no overage and nobody escalated because the dashboard read "on plan." When the legacy surface disappeared they did not notice for three weeks, then escalated angrily. Two churned. "Migrated" is not the same as "actively using the new pricing."
- A cohort you expected to migrate churned instead. Of 35 mid-market accounts on year-long contracts, 14 churned at renewal. Their common pattern: their primary use case sat on the edge of what the agent handled well, so their dispute rate ran high and a mid-year competitor look tipped them. Edge use cases need their own migration track and earlier outreach, not the long tail of Wave 2.
- A unit definition too loose at scale. "Resolved ticket" defined as no escalation in 48 hours worked for 98 percent of cases. At sunset volume the 2 percent that reopened related tickets after 60 to 72 hours produced 200-plus contested cases per week.
- Dispute volume at 12x projection. Projected 50 a month, got 600. CS analysts scaled 4 to 9 to 14, quality dipped during the ramp, and successor NPS fell 8 points.
- Quiet sales extensions. Two reps negotiated unauthorized 6-month legacy extensions for three accounts, without approval or CRM updates. Finance caught it in a reconciliation. Not malicious, but it created a fairness problem with customers who migrated on time.
What still worked
Five things did most of the heavy lifting: the lead customer reference (seven inbound logos in months 18 to 22), the migration tooling (78 percent self-service), the board narrative (the board never panicked because it always knew where it was on the curve), the legacy team's transition into migration engineering, and the quarterly reality checks that caught two problems three quarters early.
What I would do differently
Test the unit definition at 10x projected dispute volume before launch, not after. Hire dispute analysts to 5x the projection ahead of need. Audit sales rep commercial behavior weekly during the final 60 days before sunset and 60 days after. And run a parallel graceful-exit track for accounts that will not migrate, because some come back and all of them talk. For the six-quarter sequence this sits inside, see Pricing Migration: The 18-Month Quarterly Playbook.
If you are approaching a sunset, plan for these five specifically. The transition that landed at 91 percent outcome revenue and 71 percent blended gross margin was not the one that was planned. It was the one that emerged from honest reaction to what broke. Before your next sunset, write down which of these five you have actually staffed for.