Fivetran · 2019 — 2021
Getting the first sync finished
More than half of prospects who started Fivetran's onboarding never completed a data sync. I was the first designer on the app, and this is the flow we rebuilt.
Fivetran's managed connectors handled schema normalization and table structure, so data teams didn't have to. But getting a pipeline running still required going through IT or a data engineer — costly, slow, and squarely in Fivetran's way.
The brief was a system that let non-technical users self-serve. The measurable version of it was narrower: get more people through setup to a finished first sync.
Synthesized from customer interviews across key personas.
Reorder the steps, and put the documentation inside the step that needs it
I mapped the existing flow in low-fidelity first, marking every pain point so none of them got quietly lost in the redesign. Then we changed the order of steps, standardized the visual system, and rebuilt the layout to make room for embedded docs — education in place, at the moment someone is stuck on a technical form.
My PM and I agreed up front to measure it with the metrics our analytics team already tracked: time to value, and dropoff at each key step. No new instrumentation, no argument later about whether it worked.
We added a Welcome step and made things worse, then took the commitment out of it
The embedded docs did what we hoped: more completed syncs during onboarding. The new Welcome step did not. Asking people to choose a source and a destination before anything else created more dropoff earlier in the flow than the version we replaced.
We shipped a follow-up quickly, making that first choice non-committal — you could move without deciding. That release produced an overall 36% improvement in sync rates at onboarding, and we kept tuning individual steps against ongoing usage tracking.
Treat PII anxiety as a design problem, not a documentation problem
The most emotional thing in the interviews was not knowing whether personal data was about to be copied into the warehouse. People hesitated at the schema step, and hesitation at the schema step means no first sync.
Phase 2 gave users an explicit review of the schema and its PII risks before syncing — the goal stated as reducing PII anxiety and raising sync rates, in those words, so the team knew what we were buying with the extra step.
Research turned up problems across the entire journey and we could not take them all on. Enabling self-service meant starting where confidence broke down — setup — and phasing the rest behind it.
Beyond setup I worked across the roadmap — transformations, the sync chart UI, and more. The lesson that stuck was about words.
Coming into data integration new, I had to learn the domain properly before I could write a single useful label, and I had to pull the right stakeholders into copy decisions rather than making them alone. You cannot design an educational experience for technical and non-technical users at the same time until you actually understand the thing being explained.