Chief Product Officer / United States
Decisions that gave up revenue or a client to remove a surprise
- Credit unions on one shared platform
- 74
- Lanternfield Cooperative Technology clients, June 2026
- Members served by client credit unions
- 2.6m
- About 1.5 million sign in to digital banking each month
- Banking tasks completed by screen-reader users
- 91%
- Member testing panel, 2026; 58% in 2022
Decisions & consequences
01Charging for overdrafts only when customers could see them coming
In 2019 Corriston Bank earned US$96 million from overdraft and returned-item fees across about 1.3 million consumer checking accounts. Eight per cent of those accounts paid 79% of the fees. Customers paying more than ten fees a year closed their accounts at twice the rate of others, often while overdrawn. Our sampling showed that most fees followed a moment when the app's available balance looked sufficient, but pending card holds, held deposit funds or the overnight posting order meant it was not.
I proposed a projected end-of-day balance in the app, listing pending items beneath the available figure, and four fee changes: no returned-item fee at all, no overdraft fee when an account was overdrawn by US$50 or less, a window until the end of the next business day to bring the account back before any fee, and a maximum of one fee a day instead of four. Our chief financial officer, Raymond Szostak, opposed the fee changes in a written memo. He expected fee income to fall by about US$55 million and proposed a lower fee under the existing rules instead. Our head of branch banking worried that the next-day window would be hard to explain at the counter. Both memos went to the executive committee with mine.
Our chief executive, Celia Monteverde, approved the full proposal in August 2020, and we took it to the board's risk committee with the revenue effect stated on the first page. We gave up more than half of a revenue line in a year when interest margins were already falling. The fee changes took effect in November 2020.
In 2021 overdraft-related fee income was US$41 million. Fee complaints fell 64%, account closures among the heaviest fee payers fell 31% and losses written off on overdrawn accounts fell from US$9.8 million to US$6.1 million. Checking accounts receiving a direct deposit rose 6%. The bank now tests every proposed fee by asking whether a customer could have foreseen it from the app.
What I would do differently is the sequencing. The projected balance did not reach every customer until April 2021, so for five months the policy was more generous than anything customers could see. I should have shipped the display first, and given branch staff a way to check a customer's position rather than a written note.
Ending points expiry on a rewards card, and paying for it
In 2014 Quillmore's rewards cards carried a rule that points expired thirty-six months after they were earned. About 3.2 million accounts were active in the programme. Our finance team estimated that roughly US$24 million of points a year would expire unredeemed, and that estimate reduced the rewards cost the business recorded. Rewards complaints had doubled in two years, and the complaints team could show me who was losing points: cardholders who used the card mainly for occasional large purchases, many of them over sixty-five, and people whose balance sat just below the redemption threshold.
I proposed ending expiry for every open account in good standing and making the cash redemption value equal to the travel value, which had been about 20% richer and favoured our most active cardholders. Bernard Tollefsen, the card business's chief financial officer, opposed the proposal. He pointed out, correctly, that it would remove US$24 million of annual income and require a larger rewards liability on the balance sheet. Our head of card marketing argued that expiry notices were our most effective reactivation tool, because dormant cardholders spent in order to save their points.
Everett Danvers, president of the card business, approved the change on condition that we tracked its cost against retention every quarter. We gave up the expiry income permanently and increased the rewards liability by about US$57 million when the change took effect in April 2015. We kept an expiry rule for closed accounts and for accounts more than ninety days delinquent, a compromise with finance that I thought fair.
In the following twelve months rewards complaints fell 72%, and voluntary attrition among active rewards accounts fell from 9.4% to 7.9%. Travel redemptions fell less than finance had modelled. Net of retained balances and interchange, our analysts put the annual cost at about US$15 million. Reactivation mailings without the expiry threat drew about half the response the old notices had.
The business learnt to read unredeemed points as a sign that the product was not working rather than as a source of margin. What I would do differently is the framing. I presented the change as fairness to cardholders, which invited a debate about generosity. I should have presented it as correcting income that depended on customers failing to use what they had earned, and asked finance to help me measure it.
Moving members' data connections off their passwords
When we measured it in 2024, 29% of all sign-ins to Lanternfield's platform came from data aggregators using members' stored usernames and passwords on behalf of budgeting, payment, lending and tax apps. About 1.1 million members had shared their credentials with at least one app. Our client credit unions could not tell those sessions from criminal attempts to test stolen passwords, which imitated the aggregators' traffic, and confirmed account takeovers across our clients reached 212 that year. A member who wanted to cut off one app had to change her password and break all the others.
I proposed building token connections with scoped permissions, adding a Connected apps page to the shared app and then refusing password sign-ins from each aggregator once it offered a token route. Our chief technology officer, Mitchell Haddix, argued that we should wait for the federal rule on consumer data access to settle rather than build twice. Four owner credit unions objected that members of smaller apps would be cut off, and one aggregator warned that its clients would blame the credit unions. The board's product and risk committee approved the plan in January 2025, with a cutover date set separately for each aggregator and an exception route for apps that had no token option.
We gave up the option of waiting, and about US$2.3 million of development that a settled rule might have shaped differently. The Connected apps page launched in April 2025, and password access ended for the largest aggregators in March 2026.
By June 2026, 94% of third-party data traffic used token connections and automated sign-ins had fallen 81%. Confirmed account takeovers in the preceding twelve months had fallen to 64, a fall to which improved fraud controls also contributed. Members ended 186,000 connections. Contact centres at our client credit unions took 9,400 calls about broken connections in the month after cutover.
The change gave our clients a record of who reads their members' data, which several of their examiners have since asked to see. What I would do differently is twofold. The page first listed the aggregator's name rather than the app's, so members disconnected apps they still wanted; we should have tested the labels with members. And we gave aggregators nine months' notice but credit union contact centres only six weeks, an order that should have been reversed.
A cash-flow view for small businesses instead of invoicing
In 2023, 31 of our client credit unions offered business accounts through Lanternfield to about 52,000 business members, mostly sole traders, trades firms, farms, dental and veterinary practices and small shops. Our business module was the consumer app with a few extra permissions. Our clients estimated that a fifth of their business members kept their main operating account elsewhere, and lenders at three credit unions heard about cash shortfalls only when an owner asked for an emergency loan.
We interviewed 64 owners. Their most common question was whether they could cover payroll, rent and suppliers over the next few weeks. Our sales team and two large owner credit unions wanted us to build invoicing to compete with online business accounts, and a fintech offered to supply an invoicing product under our name with a revenue share we estimated at US$1.9 million a year. I recommended against it. Invoicing duplicated accounting software most owners already used, and the revenue share depended on fees charged to members. Instead I proposed a thirteen-week cash-flow view built from account history, scheduled payments and receipts the owner entered, with dual approval for payments and a referral to the credit union's business lender whenever the projection showed a shortfall.
Our head of sales disagreed in writing, arguing that cash-flow tools did not bring in new accounts and invoicing did. Our chief executive backed my proposal. We gave up the partnership revenue and a feature prospective clients had asked to see. We launched in September 2024 at all 31 credit unions.
By June 2026, 41% of active business members used the cash-flow view each month. They had 27% fewer overdraft events than comparable non-users, and business deposits at the 31 credit unions grew 11% in a year, against 4% at clients that had not switched the module on. The shortfall prompt opened 1,140 credit conversations, of which 380 led to approved lines of credit.
What I would do differently is the testing season. We tested in spring and launched in autumn, and our projections failed seasonal businesses such as landscapers and lakeside resorts badly that first winter. We added seasonal patterns in 2025, after some owners had stopped trusting the numbers. I also underestimated how many referrals lenders would receive and should have planned their capacity with them.