A 365-day analysis of San Francisco Fire Credit Union's deposit application funnel: where 85% of prospective members were being lost, why, and the page-by-page fixes that followed.
The short version
Pull-through when I started
15%
Average completion time
17.4 min
Application processing time
-19%
The problem, drawn
As Data Scientist at San Francisco Fire Credit Union (August 2024 to August 2025), I owned the analysis of how new members join. Deposit accounts are the majority of new membership applications, so if the credit union wanted to grow, this funnel was where growth lived.
I analysed 365 days of deposit-application funnel data in BigQuery. The pull-through rate was 15%: of everyone who started a deposit application, 85% never finished it.
Time was the clearest culprit. The average completion took 17.4 minutes for people who reached the final page, and a quarter of prospects took more than 20 minutes. Industry research says that is far past the danger line: The Financial Brand reports abandonment rising to 60% or more once account opening takes over five minutes, and McKinsey found every additional 10 seconds of application time correlates with a 5% increase in abandonment.
The funnel data showed where people left, but not why. So I walked the existing application page by page, the way an applicant would, and logged the friction.
The personal information and ID screens were one long form with no progress indicator, which is punishing on a phone. There was no way to save and continue later, so anyone interrupted mid-application lost everything. The ID upload step had a low click rate and produced applications stuck in pending for lack of a document. Eligibility questions and product selection were tangled together on one screen, asking people to make two kinds of decision at once.
The analysis pointed to two levers: make the process faster and easier, and stop treating an abandoned application as gone forever.
The recommendations were specific to the friction we found. Collect email first, which makes save-and-resume possible and lets us follow up with people who drop out. Simplify the eligibility questions (published research suggests removing three selections lifts pull-through by around 11%) and split them from product selection so each screen asks one kind of question. Prefill forms from answers already given. Make the ID upload an explicit, obvious step. Give the funding screen plain descriptions and a clear submit.
Each recommendation shipped as a mockup of the redesigned screen alongside the data that justified it, so the design conversation started from evidence rather than opinion.
The funnel work contributed to cutting application processing time by 19%. Just as importantly, capturing email early turned abandonment from a dead end into a follow-up list, and the page-by-page evidence gave the digital team a prioritised backlog rather than a vague sense that the application was too long.