How we built a welcome flow system with 7.6× the original revenue per recipient.
At one weighted-vest company, the team was handling urgent work across the business. The welcome flow stayed live, but nobody had time to analyze and improve it. We built a statistical and creative system that turned performance data into the next test. In its first six weeks, Klaviyo-attributed revenue per recipient moved from $0.36 to $2.73.
The flow kept running. The team had no time to improve it.
The team was handling urgent work across the business. Nobody had time to analyze weak flow performance or plan the next test.
The existing welcome series generated $0.36 in Klaviyo-attributed revenue per recipient. That result did not show which email, offer, image, or timing decision to change.
The company had Klaviyo performance data and years of email creative, but no system connected them. The team needed a repeatable way to find the next useful test without adding another manual task.
We taught the analyzer when a pattern was worth acting on.
Python models compared performance week over week, month over month, quarter over quarter, and year over year. They also compared the flow with ecommerce benchmarks, but only surfaced a conclusion when the send volume was large enough to support it.
For every position in the cadence, the analyzer read open rate, click-through rate, click-to-open rate, unsubscribe rate, revenue per recipient, and CTA performance. We used CTA performance as a practical stand-in for scroll depth.
Purchase timing exposed a narrow window. Among subscribers who did not buy during their first week, roughly 90% still had not purchased six months later.
Measurement fed the next creative decision.
We built a wiki of the company’s strongest current and historical email content. Past winners were treated as candidates, not rules. If an old idea stopped working, the system could reject it or create a variation to test.
- MeasureRead every email position across comparable time windows.
- JudgeCheck sample size before promoting a result.
- ImproveTurn the finding into a branded copy or design direction.
- TestCompare the variation and feed the result back into the system.
We encoded the brand’s visual rules so the improvement system could produce consistent email designs. Its rubric covered image assets, hero treatments, offers, subject and preview text, cadence length, and character count. It also compared emoji use in subject lines, previews, and body copy.
That made the revised welcome flow one output of a reusable process. Each recommendation had a performance reason, a creative brief, and a test that could prove it wrong.
The first iteration established a stronger baseline.
During its first six weeks live, the revised welcome series generated $2.73 in Klaviyo-attributed revenue per recipient. The original series generated $0.36. The observed flow-level result was 7.6× the original.
This was the first live iteration, not the finished version of the flow. It gave the team a measurable baseline. The system could keep reading performance, preserving what worked, and generating the next improvement. No creative choice became permanent.
Scope: DTC revenue attributed in Klaviyo. Amazon sales were excluded. The comparison covers the revised welcome series’ first six live weeks against the original series.
Your flow should tell you what to test next.
Send us the flow you want to improve. We’ll analyze each email, find where performance falls away, and give you a specific next step to test.