← Back

PLG'd Up: How Vercel Overcame Scale Limitations

Jul 10, 2025 (8mo ago)

Our Marketo system had become a bottleneck. With over 20,000 new contacts hitting our system daily and Marketo's sync cap, we were looking at backlogs that stretched for weeks. Campaigns couldn't execute on time. Sales handoffs were delayed. The team was buried in manual data work.

We had two options: keep patching or rebuild from scratch. We chose to rebuild.

What was breaking

The problems were threefold:

  • Volume mismatch. Around 20K records per day hitting a system that couldn't keep up.
  • Data stuck. Unreliable scoring and lifecycle management from week-long backlogs. Contacts were created in Salesforce but didn't MQL until they synced into Marketo, ran through lifecycle workflows, and synced back. SDRs had to rely on "latest contact sales date" just to know who to reach out to.
  • Operational risk. Slow automations and bad data meant lost credibility with the business.

At one point we had an 811K contact backlog on pull updates alone. The sync was 3-4 hours behind on a good day. For upcoming events, we had to keep an eye on confirmation emails and send manual batches if needed.

How we chose the next platform

We evaluated Inflection, HubSpot, and Marketo against our core requirements:

  • No sync backlogs. Real-time product triggers, not hourly batch jobs.
  • Data and automation. Easy management of business logic for stakeholders who aren't engineers.
  • Flexibility. Modular architecture with native Snowflake and Salesforce integrations.

Inflection scored highest on integrations, automation, pricing, customer support, user experience, and scalability. We made the call.

All hands on deck

The best way to eat an elephant is one bite at a time.

Step 1: Inventory our evergreen assets. We preserved emails, key programs, and lists that were still working.

Step 2: Ensure the right resourcing. We partnered with the EMMIE Collective to supplement our team during the migration. They helped us move fast without cutting corners.

What we shipped

In 30 days, we migrated:

  • 8 nurture journeys
  • 19 audience lists
  • 50+ emails
  • A new lead routing system we call the "Sorting Hat"

The Sorting Hat is a nurture journey router that evaluates enterprise lifecycle stage, account status, plan type, and membership in other journeys to determine where each contact should go. It runs weekly and keeps contacts from falling through the cracks.

Beyond the email infrastructure, we also shipped a lot of Salesforce Flows to handle lifecycle progression, team member syncing, opportunity contact roles, and lead scoring.

Discovery: vibes are not a workflow

Platform migration was only part of the problem. During the process, we discovered our email approval workflow was broken.

The issues:

  • Overlapping feedback. Everyone commenting on everything all at once. One Slack thread had 101 replies.
  • No SLAs. Same-day build and approval requests were the norm.
  • Inconsistent approval chain. No clear sequence for who reviews what and when.

The fix: putting the train on the tracks

We migrated to Knak. The benefits:

  • Simplified building process. Drag-and-drop email creation instead of custom HTML.
  • Dark mode support. Greatly simplified custom CSS for dark/light mode-specific images.
  • Approval chains. Consistent, linear feedback that accelerates the review process.

One path, clear owners, no ambiguity. Campaign velocity improved and the team stopped dreading launches.

What's next

We're looking at four areas:

  • AI Agent. Automating campaign operations tasks to focus on quality and impact.
  • DWH Writes. Writing performance data to the data warehouse so we can cook from there.
  • Webhooks and Tokens. Firing webhooks to first-party apps we develop with v0. Build Vercel with Vercel.
  • Product Events. The true unlock of PLG potential. Using product events to inspire more ships.

The takeaway

Specialization beats parity.

Sometimes fixing a broken system means rebuilding it entirely. Incremental patches on a fundamentally constrained architecture will only get you so far. The 30-day sprint was intense, but we came out the other side with infrastructure that can actually scale with the business.