Case study · Oxygen · 2021 to 2023
Turning a hundred thousand free accounts into a membership business
Oxygen Elements, a four-tier program for freelancers and one-person businesses, priced from free to $199.99 a year.
- Timeline
- 2021 to 2023
- Role
- Director of Product Management
- Scope
- Consumer, business, formation, invoicing
- Type
- Membership, growth, monetization
Highest ranked banking app on the US App Store Finance chart in launch week, ahead of Capital One and Chase.
Growth in platform users following the relaunch and rebrand.
Reduction in processor API calls after redesigning the balances infrastructure.
Of members who formed a business entity also opened a personal account.
The problem
Oxygen earned revenue one way, through interchange on card spend. No lending book, no spread. I owned the whole suite, personal and business.
Roughly one in a hundred of our hundred thousand personal accounts carried a positive projected lifetime value. The five hundred business accounts were almost all profitable and transacted about ten times as often, at barely higher acquisition cost. The brief was to grow the business line. The hundred thousand unprofitable personal accounts were mine as well, and nobody had a plan for them.
The insight
I interviewed five of our most engaged business customers and five who had spent heavily and then gone quiet. Both groups wanted the same thing, and it was not a business feature. They wanted what a premium travel card gives you. Lumpy self employed income made them ineligible for those cards no matter what they earned.
When I put a prototype of the business benefits in front of them, they asked, unprompted, whether they could have it on their personal account. A one person business does not have separable business and personal finances. The person is the business. One benefits package, built once, could serve both books. A paid ladder on the personal side, bundled by default on the business side.
The ladder
| Tier | Annual fee | Qualifies on, per 30 days | Cashback | Savings APY |
|---|---|---|---|---|
| Earth | $0 | No minimum | 2% | 0.50% |
| Water | $19.99 | $1,500 in or $500 out | 4% | 1.00% |
| Air | $49.99 | $3,000 in or $1,000 out | 5% | 2.00% |
| Fire | $199.99 | $6,000 in and $2,000 out | 6% | 3.00% |
Read the Fire row twice. Water and Air admit a member on deposits or spend. Fire requires both. A member who parks money without spending costs interest and earns no interchange. A member who spends without depositing never becomes primary. Only the tier that costs the most to serve refuses to admit either one.
Savings rates applied to the first twenty thousand dollars and paid zero above it, so the cost of serving a member was fixed and knowable regardless of how much they deposited. The protections stayed free at every tier, including Earth, which is what made the free tier worth opening in the first place.
Elements was the personal product. Business accounts did not climb it. They received the Air benefit stack by default, plus the things only a business needs, formation and invoicing. That asymmetry was the point. A business account already produced the deposit and spend behavior the ladder exists to induce, at roughly ten times the transaction frequency of a personal account. Charging it to climb would have taxed exactly the customer we most wanted.
Launch
Elements launched on 23 June 2021 with a new brand identity. The app reached fifth on the Finance chart during launch week.
The four names above it were payment apps rather than banks. On that chart, that week, Oxygen was the highest ranked banking application in the United States, ahead of Capital One and Chase.
Three decisions worth defending
Micro caps instead of a monthly cap
Six percent cashback is several times what debit interchange earns, so an uncapped program loses money on every qualifying swipe. A monthly ceiling protects the average month and still lets one person drain the pool buying gift cards at a six percent grocer in an afternoon. We set ceilings per transaction and per day per merchant instead, in the range of seven to ten dollars, with frequency limits by category. That caps qualifying spend at roughly a hundred and twenty dollars a day at any one merchant, which no ordinary shopper reaches and no arbitrageur can work around. The headline was six percent. The realized blended rate was a fraction of it, by design.
A referral that bought activation
Introducing a friend who opened an account paid five dollars at Earth and ten at the paid tiers. A further hundred released only when the referred member took a payroll direct deposit of five hundred dollars or more and made at least five card purchases within sixty days. Ninety percent of the payout sat behind the two behaviors that actually made an account profitable, on a clock long enough to prove the behavior had stuck.
Formation as the wedge
Rather than argue the business roadmap, we tested it. Landing pages and in app fake doors ran against four candidate products. Formation won, specifically the S corporation, because an owner paying themselves above roughly eighty thousand dollars keeps materially more of it. That is a number a person can feel, unlike an interest rate. Around a hundred business members I had written off as dormant came back to use it, it drew about four hundred new customers from a more valuable segment, and eighty percent of them also opened a personal account. Formation was the wedge, and the personal ladder was waiting on the other side of it.
What I got wrong
We built business invoicing first, on the founder's one stop shop thesis, and I made the call to release it. Within hours a member could not open the app. The fault was in our Stripe authentication flow, where a rejection placed a hold on the entire account rather than only the new feature. We were mobile only with no branches, so an affected business owner could not check a balance or move money at all.
I turned off the feature flag on launch day and repaired the accounts by hand. The launch was gone. A customer locked out of their bank for a day is a worse outcome than a failed launch, and that trade was not close. The fake door tests came next because I had spent the credibility to argue for anything.
The larger miss was calibration. The insight held. The reward structure was not tightly enough conditioned on revenue generating behavior, and on a business earning about one percent of transaction volume that gap compounds. Build the eligibility rules with the ladder, not after it.
The rest of the work
A rewards ladder is a machine for increasing engagement, which on this model is dangerous. Revenue arrived only when a member spent. Processor cost arrived every time the app asked the processor a question. Engagement scaled cost continuously and revenue only occasionally. I investigated the balances infrastructure and redesigned it with our system architect, cutting API calls by ninety percent with no change to what the member experienced. The ladder made people engage. That work made engagement affordable.
Two thirds of business applicants who did not arrive through the formation funnel were rejected on incomplete documentation. I shadowed a support agent, compared accepted against rejected applications, and interviewed rejected users. Clearer document descriptions, instant upload validation and a review team specialized by document type lifted completed applications by half and doubled approvals among them. At peak the platform served 250,000 monthly active members and more than 1,750 businesses a day.
Figures here are drawn from Oxygen's published rates and terms, contemporaneous trade coverage, and public App Store rankings. Internal financials are not included. Happy to go deeper on any of it in conversation.