Signals: The Silicon Valley Bank that never was

Like many readers, I chose to open my first business bank account at SVB. SVB is the bank that gave me my first mortgage. Uniquely, it's also the bank that bought my first company.

This essay explores my experience selling to SVB, working there, and our disagreements with them over our vision for what the bank could be. My story is unique. I got to work there but was always an outsider. I was a technologist in a place filled with bankers. And I had a clear idea of what I wanted to build in an institution unfamiliar with shipping new ideas.

In many ways, this directly relates to where SVB finds itself today.

The bank recently went into receivership because it mismanaged the duration of its bond portfolio, its communications around attempts to fix that, and the ensuing liquidity crisis from a concentrated deposit base.

But in my experience, all of these topics were actively talked about nearly a decade ago: a cyclical deposit base and the need to keep the duration of assets (which SVB didn’t originate itself) short duration.

When we sold Standard Treasury to SVB, we found ourselves up against a culture that sought consensus at the expense of decision-making and that focused its attention on the comfort zone of relationship banking while ignoring adjacencies that could have allowed it to diversify earnings through fee and payments income and increase its homegrown loan book. We weren’t the only ones advocating for strategic reorientation, and if SVB had pursued it, it's possible that the past couple weeks would have turned out differently.

The scene is in early 2013. Stripe is only three years old, and many of us still remember when it was called /Dev/Payments. Marqeta is starting to tackle card issuing, but Simple is the only neo bank going, and they are struggling to get a bank partner. Plaid would get founded in May. It’s early, but fintech is in the air. Those of us who grew up against the backdrop of the great financial crisis figured there must be a better way, and we would build it.

My co-founder Dan Kimerling and I started Standard Treasury in March 2013. I had left Stripe (oops?), and he’d left a company called Giftly when we got into the S13 batch of Y Combinator with a simple concept: API banking. White label payments and bank accounts distributed to and through fintechs. When we started the company, we didn’t understand the bank’s perspective, but our theory was that there would be many more fintechs in the next decade, that banks struggle with distribution (have you seen the CAC numbers?), and that there had to be something there.

We spent two years trying to sell to banks. We did proofs of concept with Morgan Stanley, Goldman Sachs, and JP Morgan. None of them live in production. We got frustrated. We had built great technology. An excellent API. Great documentation. Connective tissue to ACH systems and core banking platforms. But we couldn’t find banks that were willing to do it.

We heard that something was happening in the UK. That new banks were getting charters, like Starling and Atom. We went over there and talked to a bunch of lawyers and accountants, and we met with folks like Tom Blomfield (who had recently left Starling but hadn’t quite founded Monzo yet). We devised a plan to get a bank charter in the UK for the “AWS of banking,” sell into the fintech market there and then return to the US using a branch.

We pitched everyone. A Series A startup with only vision and no traction. But what a vision! (Read our whole deck here). 2015 was a big year for venture capital, but not for us. We struck out completely, getting farthest with two visionary investors who saw the world as we did— Chris Dixon at A16Z and Keith Rabois (then) at Khosla. Both brought us to full partnerships meetings multiple times, but they couldn’t get it through (or didn’t want to do it in the end).

We were right about fintech’s growth. We were right about the need for API banking. We were right about the company that needed to be built, but just a few years too early to market. And it’s fascinating to watch VCs and entrepreneurs think more and more highly over time of a company that failed. It’s been good for me, obviously.

A bunch of our team went on to found or work at Treasury Prime. It’s named as an iteration of Standard Treasury and has expertly run the playbook of working with banks to launch APIs. Some days I think Chris and Jim were better at executing than we were. But most days, I give myself the benefit of the doubt and believe Standard Treasury timed the market wrong.

Another one of our engineers, David Jarvis, went on to be the founder of the UK take on our idea: Griffin Bank. They’re also doing very well and just recently got their license. Jarvis has been persistent to keep moving forward there (he also had British citizenship, which helped).

I will skip most of the story of our sale process – a harrowing tale of making payroll week after week as we were selling the company – in the end, we had acquisition offers from Coinbase (oops?), Airbnb (oops?), and SVB. SVB was the only buyer that wanted the entire team (both Airbnb and Coinbase wanted four of our six teammates, though a different four in each case) and also wanted to keep the product going. We closed the purchase in August 2015.

Banks primarily make money on interest income. They borrow cash short-term from their depositors, and they lend long-term through things like term lending, mortgages, and credit facilities, or, mainly in SVB’s case, capital call loans and venture debt. The difference in the rates between those two— paying depositors, say, 1% and lending out at 3% – is called net interest income. Because of this “maturity transformation” between deposits and loans, banks can never simultaneously pay out all their depositors. Bank regulation and supervision are guidelines to keep banks solvent and hopefully prevent a bank run. Sometimes that doesn’t work.

Banks have another way to make money, called non-interest income. For SVB, the point of the acquisition was to increase non-interest income. That is, the money they make in every other way than making loans. Non-interest income can be earned through annoying things like overdraft, account, mortgage, and wire fees. But it can also be earned via interchange on card acquiring and issuing, asset and wealth management, warrants on venture debt deals, cash management and transaction banking, ACH facilitation, and more. In short, most everything in banking over API is an opportunity for non-interest income.

Markets tend to value non-interest income more, and while banks often trade at a slight premium to book value, they get a lot of credit for differentiating products and clients (which SVB did) and for non-interest income. The more your revenues look like SaaS revenues, the better it is for an organization’s market valuation because SaaS revenues tend to imply repeatable, scalable, and steady earnings. It’s the goal of many banks right now, and why, for example, Morgan Stanley trades at a higher premium– they focus on wealth management.

SVB also knew then (in 2015) that its deposit base was incredibly cyclical. They had seen steady and significant drawdowns of their deposits after the dot-com bust and the great financial crisis. The theory of the case for the acquisition was that we would build out non-interest income so that the bank could smooth its earnings (and hopefully its ROE numbers) in the face of needing to keep the asset side of the balance sheet safe and at a short duration.

Now my theory of the case also extended far beyond payments and bank accounts. You could imagine an entirely white-label bank that is amazing at managing its fintech partners and focused entirely on embedded payments, lending, wealth management, asset management, and more. Such a bank could distribute checking and savings accounts through neobanks and commercial accounts through vertical SaaS companies. That bank could also distribute loans through consumer and SMB credit platforms (i.e., lending-as-a-service).

I envision SoFi, Square, and Lending Club never becoming banks because they would have had a good, scaled partner in SVB. SVB would be the partner bank of choice for card issuing programs. It would be “headless”. Do everything that Treasury Prime, Marqueta, and Kanmon do combine but turbocharged with a banking charter.

But that comprehensive vision did not fit where SVB management saw the bank heading – they bought us to integrate our API platform into the bank for white-labeled payments and bank accounts. Even in 2015, the bank already had more deposits than assets (loans) it could originate independently, so we brought up lending-as-a-service a few times.  It was dead in the water from the start. SVB was very uncomfortable doing anything in lending that was outside their complete control – or their perceived core competencies like venture debt, capital call loans, and lending to entrepreneurs and VCs directly. There may have been compliance concerns about third-party originations, but the primary objection was cultural: SVB’s bankers thought of themselves through the lens of old-school banking driven by relationships and personalized knowledge.

Ultimately, our job was to be the center of a new focus on non-interest income. Steady SaaS-like income would amplify returns on equity (ROE) in the good times and protect the bank in the down times.

Operating in big companies is always a tough transition for startups. Transitioning to a bank can be even more complicated. Our challenges were both tactical (day-to-day) and strategic (big picture for where the bank should go).

Tactically, we were slowed down for reasons both reasonable (risk management procedures) and cultural. Our team’s reporting structure was split. Our engineering had to report to IT, I had to report to Product, and Dan had to report to Strategy because of organizational rules about where specific types of employees could roll up to in the organization.

SVB’s culture was one where front-line ICs and low-level managers were not expected to make decisions independently, but instead you would be expected to escalate all decision-making. That led to paralysis wherein only the most senior managers could make even the most mundane decisions. It was difficult to launch our product coherently against our differing managers’ various requirements and incentives and what decisions they wanted to bring to their bosses and use political capital.

Beyond that, the organization was incredibly consensus-driven. The default is the status quo if not everyone can agree. And though getting buy-in and sign-off is often the hallmark of a sound risk management culture, many colleagues would privately gripe that any one dissenter could often prevent change.

We were asked to divert our engineering attention to any number of unrelated projects as soon as the bank realized that ours would ship. We were asked to support the (still not complete) relaunch of commercial online banking. When it turned out we had continuous deployment and integration (versus the bank’s quarterly release cycle), we were asked if we could fix the bank’s entire existing deployment processes. I was “promoted” to take over more product areas (against my wishes) and put on the firm-wide architecture and technical risk committee, so I had less day-to-day time to just be the PM for API banking.

But the real delays were strategic. Despite the acquisition – which we later learned was driven by one visionary C-level executive – there was firm-wide disagreement on technology and payments investments and priorities. And where there was dissent, there was no movement or progress because no one (even the CEO) was willing to make decisions independently.

Our work ended up splitting the bank into two factions: one that wanted increased focus and investment in non-interest income and one that wanted to keep the status quo (LaaS was never seriously on the table). One that thought there should be a strong focus on payments and non-interest income more broadly, and one that just wanted to keep lending to the tech sector based on relationship banking. One that thought the bank needed to not only say it was for the innovation economy but to be an innovative company itself, and one that thought our sector focus and access were differentiation enough. One that thought technology would animate the bank's future, and one that thought it didn’t matter to what the bank did daily.

These factions did not fall neatly based on organization charts or age or tenure at the bank — the Head of Relationship Banking (the head banker and salesman) wanted diversification in revenue. His up-and-coming direct report thought keeping our heads down and focusing on loans made sense. The technology organization split between folks who thought the only thing that mattered was updating online banking and other internal systems (primarily but not exclusively the people who focused their reputations and political capital on those projects) and others who also wanted to focus on revenue-generating activities.

But in the end, the bank decided not to launch a robust API platform. Or better put, the bank could never choose to act and defaulted to not launching. We got through much of the muddle around things like deployment and architecture and programming language and risk approvals – but launching an open API platform to onboard partners was ultimately not something the bank’s most senior leaders could positively decide to do. By default, they wanted to focus their technical energies on online banking and their cultural powers on lending through relationships.

The bank must diversify away from interest income. We were stubborn, and they couldn’t fire us (that would require someone making an affirmative decision!), so we fought for a decision to be made. We made decks. We wrote strategy memos. We presented to board committees. But the bank’s leadership deprioritized the idea and decided to focus on the relationship-driven banking that was the core of the business model historically. So we lost, and we never went into GA.Ultimately, one API banking opportunity came along that the bank couldn’t pass up precisely because it (theoretically) drove clients to build a relationship with SVB:  being the infrastructure behind getting a bank account on Stripe Atlas. SVB’s management let our team make that platform, not because of a strategic vision around payments or non-interest income but because it expected that every startup built on Atlas could graduate into a relationship with SVB’s Early Stage Banking team. SVB eventually turned even that off because the graduation theory never panned out, and Stripe found new partners. I’m not sure who maintained it after we all left.

SVB has decreased its payment capabilities and partnerships over time and done less and less tech-driven banking, and everyone on our team was gone after the two-year earn-out.

Much has come out over the last few weeks that doesn’t surprise me about the tactical decisions of SVB over the previous two years. Executives pushed out the bank’s loan portfolio duration to increase ROE. The bank didn’t have a CRO for much of last year. Risk committees discussed duration risk, but executives couldn’t reach a consensus to change the portfolio. They sold the entire available-for-sale portfolio in one go because they couldn’t foresee a liquidity crunch. Goldman Sachs and SVB botched the capital raise and, more importantly, completely botched the communications (why announce it the day the FDIC is at Silvergate!). Blackrock’s Financial Market consulting group gave outside advice about financial controls that SVB ignored. And on and on. Many have written about these mistakes and more.

But I keep returning to the bank's long-term strategic orientation. Not what happened over the last two years but what the bank focused on over the previous two decades.

SVB was uniquely situated at the epicenter of venture capital and venture-backed businesses. With this came what many have pointed out as their core strategic challenge: concentration risk in their core customers. All their customers came from the same industry and were exposed to the same economic cycles. Though it's easy to point to the Twitter-driven bank run as the proximate cause of their downfall, the root cause was that they would always have liquidity challenges from all their deposits coming from one industry and that industry having the same cycle. At the same time, the bank could never originate as many loans as it had deposits.

The solution to flighty deposits is to keep a very short-duration loan book. We openly and frequently discussed this idea when I was at the bank — and one of the critical arguments for the increased non-interest income camp: “We need to keep a short duration loan book which depresses ROEs, so we need to increase earnings through non-interest income.”

They had a front-row seat to the financial technology revolution of the last decade. They could have been its chief sponsor and enabler instead of folks being pushed to smaller community banks like Evolve Bank or Radius. They had an opportunity to build out their technology and compliance partnerships with, and perhaps look more like, tech companies.

This strategy would not have been mere convenience; it would have allowed them to hedge their core challenge of flighty deposits.

Now does what happened to Standard Treasury matter for what happened over the last few weeks? No, not exactly. What matters, though, is that SVB never fully internalized the hard truth about their business as indicated by their failure to integrate our company successfully: If you bank highly cyclical companies that draw down deposits EXACTLY when the economy goes through a rough patch (and interest rates rise), you should hold a very short duration portfolio with your assets. My argument was then and remains now that such a bank could primarily increase its ROE through tech-driven non-interest income.

I just haven’t found the bank that will do it yet.