I’m A Founder Again: Why I Started Rivet

I’m a founder again and I’m so excited to share what I’ve been quietly working on for the past year.

This is the longer version of that story: how a litigator ended up in payments, what I learned running payments infrastructure from the inside, and why I think the next few years are going to change how every company that takes cards thinks about its payments.

How a lawyer ends up in payments

I never planned on working in payments. I’m the child of two attorneys, one who started and grew his own firm, the other who ran legal departments for large corporations, modeling for me that there was no such thing as a glass ceiling in our household. I started my career the same way: litigation, then in-house as a member of Google’s legal department.

Then I noticed that the worst part of running a legal practice was getting paid. Clients still mailed paper checks. And when I dug in, I found out why: most of the major payment platforms wouldn’t even touch lawyers’ trust accounts, because of all the rules around money that’s held for a client but doesn’t yet belong to the lawyer. So in 2014, I left Google to start Headnote and bring payments for law firms online.

Learning payments from the inside

Headnote had two private equity exits in two years (another story for another post, hat tip to Alpine and Francisco Partners). Along the way, my team and I ended up running the payments infrastructure behind a whole portfolio of software companies, basically an internal Stripe or Adyen, for a PE roll up of companies with tech stacks and engineering cultures as different from one another as you can possibly imagine. As such, Headnote’s technology had to be cloud-based, multi-tenant, highly scalable, and able to onboard new software acquisitions as fast as possible so it could start capturing payments and maximizing payments revenue.

That’s where I learned the part of payments nobody talks about, buried deep in the Fraggle Rock that is fintech infrastructure. Every month, part of my job (along with a small army of analysts, product managers and engineering resources) was going through our customers’ transactions one by one, analyzing card mix and processing costs with more granularity than I thought possible, trying to figure out what had changed and why. It was a never-ending, Sisyphean exercise: as soon as we finished analyzing, investigating and making changes to the transaction coding at the payload level, it was time to start again. The treadmill of monitoring and optimizing our processing costs was more work than I had anticipated.

If I’m being honest, I have the earnout structure that was put into place after Headnote’s first PE exit to thank for this learning. Had I not had a personal stake in getting really great at optimizing payments revenue, I probably would have decided that “good is good enough”, and wouldn’t have dug deeper and deeper, as I went further into the subterranean layers that are card processing infrastructure. But I was a dog with a bone, on repeat: why did our fees go up or down, and what could we do to make payment revenue more profitable, predictable and reliable?

Sometimes the answer was volume, other times it was card mix that led to repricing decisions. Often it was something much harder to see: a rule had changed, a field had stopped making it through, a payment that used to qualify for one rate now qualified for a worse one. Nobody sent us a notice. There was no cheat sheet or playbook. Fees showed up after the monthly close process, and it was up to us to decode it.

The problem almost nobody can explain

US merchants paid nearly $200 billion in card processing fees last year, up about 70% since 2019. For most companies it’s their biggest cost after payroll, and almost nobody can tell you why it costs what it does.

The fees are a slippery beast, in part because the card networks change the rules that decide what each swipe costs twice a year. The short version: it’s all based on how well you send contextual transaction data, data that will likely get lost, dropped or degraded somewhere between your systems and your customer’s bank.

And the only companies that manage it well are the ones big enough to hire entire teams to do it. The largest platforms in the world have dedicated payments optimization teams whose whole job is squeezing basis points out of every transaction. Most companies can’t hire that team, and honestly, most shouldn’t have to.

Why now

After Headnote, I spent a few years on the other side of the table coaching founders, which I loved. (I also had a quick but impactful breast cancer speed bump along the way.) But I kept coming back to the same idea.

The tools we’d built to run payments worked well, but they only worked because of the people running them. Payments optimization has always been expert labor. Someone has to read the networks’ rate sheets, hundreds of rows each, twice a year. Someone has to decode the data each processor sends back, which never looks the same twice. Someone has to check every transaction against rules that keep moving, then figure out what to fix. That’s why only the biggest companies do it well: it takes a team of specialists, and that team never gets to stop.

Then AI got good enough to do the reading. It can take in a 300-row rate sheet and a messy processor export and work out what changed, what’s missing and what it’s costing you. That turns expert judgment into something software can do continuously, for every company, not just the ones who can afford a payments team. And it means a fix can run on every single payment, in milliseconds, before it ever reaches the processor.

It also changed how we build. The old way took weeks or longer. You’d watch transaction volume pile up, have an analyst go transaction by transaction to work out where the money was leaking and where to start, spend precious roadmap time building the fix, push it to production, and then start the whole cycle over again. Now that loop runs on every payment, in milliseconds. Our own insight engine reads a customer’s payment data and tells them in plain English what’s going on. A small team can now do what used to take a department, and that’s exactly what we want for our customers too.

That was the light bulb. That’s Rivet.

What Rivet does

Rivet sits between companies and their processors. We show them what their card payments really cost, keep them current with network rule changes automatically, and fix the data behind every payment so it qualifies for the rate it should. It’s the payments optimization team you’d hire if you could, without the headcount or the giant budget.

We do that in three steps:

• Audit: we look at a sample of a company’s transaction data and show them where they’re leaking money and why.
• Monitor: a live source of truth for every fee, downgrade and rule change, broken out however their finance team needs it.
• Optimize: an API that sits in front of the processor, enriches and corrects each payment in under 50 milliseconds, and sends it on its way. When the networks change the rules, our customers are already covered.

Our first year

At the start of this year, we had no office, no website, no product and no paying customers. Here’s where we are now:

• We raised an oversubscribed pre-seed round from Precursor Ventures, MSIV and others. I’m especially honored that so many investors from my previous companies chose to back us again.
• We’ve analyzed nearly 20 million transactions and roughly twice as many fee records, with live data flowing in from Stripe, Adyen and Fiserv.
• We’ve found money left on the table in every single audit we’ve run. Most companies we look at have somewhere between 5 and 50 basis points of savings available. In one audit alone, that added up to more than $2 million a year.
• Every one of our customers came from my network or a customer referral. Not one came from marketing, mostly because there wasn’t any. They range from vertical software platforms to fintechs and marketplaces, they all share the same problem, and they’re happy to share Rivet with others.
• We have 100% customer retention. We’re honored by the partnerships we’ve formed with each and every one.
• We’re seeing real results, not just findings. After fixing its business-card data, one customer increased its share of Visa-verified transactions by about 7% in a single month, which meant nearly $20,000 in savings actually realized in thirty days.
• We became an official Stripe partner, built a Stripe app, and have multiple patents pending on our core technology.
• We built a team of veterans, nearly all of whom have worked together before. After years apart, coming to work every day feels a lot like coming home. How many people can say that?

What we’re learning

The thing I keep coming back to is the data itself.

Every card payment carries a little packet of information. Most of it is bare bones: an amount, a card number, a merchant. But the context around a payment is worth much more. Knowing that someone paid $15 for parking on a Tuesday is useful. Knowing that I paid $15 on a rewards card, from Marin, to park for a Weezer concert at Chase Center is a completely different level of information. It tells the bank the payment is real, it tells the merchant something about its customer, and it tells everyone in between how to treat it.

Right now, no single company can see how the quality of that data connects to what happens next: the rate a payment gets, whether it’s approved, whether it turns into a chargeback. Each processor, network and bank sees its own slice. We’re starting to see across all of them, and I think that view is where the long-term value is.

Where this is going

Here’s what I believe, and why it matters now more than ever.

Every company with real volume is becoming a payments company. A marketplace, a software platform and a distributor all end up running two businesses: the one they set out to build, and a payments operation they never planned on. That’s not a niche. Cards moved more than $50 trillion worldwide in 2024, and that number is on track to pass $70 trillion by 2030. Every one of those payments carries data, and every one of them is priced, approved or declined based on it.

Fees are just the first line item. US merchants paid nearly $200 billion to accept cards last year, and that’s the part everyone can see. The same data that decides what a payment costs also decides whether it gets approved, whether it gets flagged as fraud, and whether it turns into a chargeback. Card fraud alone cost the industry more than $33 billion in 2024. Lower fees are where we start, because they’re the easiest to measure. Higher approval rates, fewer false declines and fewer disputes all come from the same fix.

The rules will keep changing, everywhere. The card networks rewrite their rules twice a year, and every region, card type and processor adds its own wrinkle. Nobody finishes complying. You either have a system that keeps up automatically, or you have a team that falls behind a little more every April and October. That’s true for a company processing $100 million a year and for one processing $100 billion.

AI agents are about to make the data problem much bigger. Soon, a growing share of purchases will be made by software acting on our behalf. When an agent books the flight or reorders the supplies, the bank loses most of the signals it uses today to tell a real purchase from a risky one. The data riding along with each payment becomes the whole story, and the companies that send clean, complete, trustworthy data will get approved and priced better than everyone else.

And the data is valuable to everyone in the chain, not just merchants. A merchant wants lower costs. A platform wants better margins on embedded payments. A processor wants happier, stickier merchants. A bank wants fewer bad approvals and more good ones. A network wants its rules followed. They all need the same thing: better data at the source.

Why this has to be independent. Your processor only sees its own slice, and most companies with real volume use more than one. Processors also have little reason to fix it: interchange passes through to them, so a downgrade doesn’t change their economics. It changes yours. The company that fixes payment data has to sit on the merchant’s side, work across every processor, and learn from every payment it touches. Every payment we fix teaches us something about the next one, and that compounding is something no single processor, network or bank can build on its own.

So we’re starting with the merchants and platforms at the very beginning of the payments chain, fixing their data before it ever reaches a processor. But the long-term opportunity is to become the data layer that sits underneath all of it, so every payment, made by a person or an agent, anywhere in the world, carries the information it needs to be priced, approved and trusted correctly.

The payments teams that win will be small, run on great data and lean on AI for everything tedious. We’re building the infrastructure for that future.

Come build it with us

If you run payments or finance at a company with meaningful card volume, I’d love to show you what we’re finding. Please book a time with me here or email me at sarah@rivetbps.com. We’ll look at your data for free.

If you’re an engineer or payments nerd who wants to work on hard, real-world payments problems, let’s talk. We’re hiring.

And if you invest in payments infrastructure, I’d love to get to know you now.

Fourth company. It’s just as fun and just as hard as it ever was. I’ve missed it.

Still Riveted,
Sarah