The channel
paid couldn’t
buy.
In the depth of crypto winter, paid acquisition stopped working. I bet that growth wasn’t a budget problem, it was a channel problem, and built the foundational referral system Gemini still runs on today.
In direct revenue attribution since inception. A compounding channel built when the rest of the industry was cutting acquisition spend. Organic by design.
Refer on either surface, earn when the referee transacts or is approved.
Late 2023 was the trough of crypto winter. Trading volumes had collapsed, exchange revenue with them, and across the industry every growth team I knew had its paid budget cut to the bone.
Gemini’s acquisition mix was overweighted on paid. CPMs kept rising while ROAS kept falling. Inside that math, the answer wasn’t to spend smarter, it was to build a channel that didn’t depend on spend.
What was missing was unmistakable: every healthy consumer fintech had an organic acquisition motion that paid for itself. We didn’t. So I proposed building one.
You out-build.”
The thesis was simple: every existing user is a latent distribution channel. In a market where you can’t outbid Meta or Google, the cheapest user is the one your users bring you.
I framed referrals as performance-based CAC paid to users, not a marketing campaign. The economics had to be ROI-positive on a per-referral basis from day one. No incentive paid out unless the referred user generated real, qualified revenue: a funded trade or an approved card.
And it had to be infrastructure, not a one-off promotion. A foundational system Gemini could compound on for years, across whatever surfaces we launched next.
Exchange
Built the foundational referral system on the core exchange surface. $50 reward, paid only when the referee completed a qualifying trade. Tracking, attribution, payouts, and abuse controls all designed once, designed to last.
Credit Card
Extended the system to the credit card surface. Same $50, same referrer, new qualifying event: an approved card. Reused the v1 plumbing rather than rebuilding, which is what made the multi-surface story possible at all.
Tuning
Reward economics, eligibility windows, and abuse controls re-tuned as market dynamics shift. Paid acquisition appetite returns and recedes; the referral channel keeps compounding underneath.
Direct revenue attribution from the referral program has been in the $35–40M range since inception, and it is still being added to every month. By construction, every dollar of that is paid for by real qualifying activity, not speculative spend.
The bigger win was structural. We came out of crypto winter with a permanent organic channel sitting alongside paid, sized large enough to matter, and architected to extend to whatever surface Gemini launches next. The credit card extension was the proof: the system was infrastructure, not a campaign.
The program continues to evolve with the market. The bones I built in Q4 2023 are still what it stands on.