How Ramp data works

When I joined Ramp last year, the idea of a six-year-old fintech startup starting an economics function was, internally, seen as highly experimental: a product of Ian Macomber’s vision to have Ramp, already known for its engineering, product, and design, known for its data.

The big banks had all started their own think tanks covering American consumers, in part as a brand-building operation following the financial crisis. Some firms went further. Zillow became where people actually go to make decisions about housing. ADP's payroll data became an essential complement to government labor statistics as federal agencies faced budget cuts and survey response rates declined.

No one had done the same for business spend, despite it being a leading indicator of broader economic shifts. The limited datasets that did exist were often survey-based (like ISM Manufacturing) or reported on long delays and lacking granularity (like BEA’s Supply-Use Tables). These datasets are useful for broad strokes, but not much help if you're a business trying to benchmark your software costs or figure out whether your peers are pulling back on marketing.

Over the past year, we've started to achieve what we set out to do: make Ramp the most credible source on how companies spend and operate. Our data has been cited by the Federal Reserve, The Wall Street Journal, The New York Times, the Financial Times, CNBC, and Apollo's chief economist. We’re even in the Bloomberg Terminal.

As more readers see Ramp data for the first time, it's worth stepping back and covering how our research works, what our data captures, and where it leans.

Our key research projects

When we pick topics, we ask three questions: do we have the data, are we well-suited to write about it, and does publishing it serve American businesses? That's led us to everything from AI adoption to work culture to corporate sports spending. Our work is anchored by two flagship datasets, with more on the way:

• Ramp AI Index — Monthly tracking of AI spending across U.S. businesses.
• Top SaaS Vendors — Market share and trends in business software and categories using Ramp Rate.

Later this year, we’ll start publishing more macroeconomic data, including a near-real-time index of U.S. business spending trends, broken out by sector and spending category.

Our dataset

First, Ramp data covers more than $100 billion in annual spend across 50,000-plus U.S. businesses. Unlike other transaction datasets, which often only encompass consumer spend, Ramp captures business spend, using both corporate card and invoice-based payments (the majority of our payment volume comes from invoices paid on Ramp, actually). When our research relies only on cards, we are explicit in our chart notes (it's usually because adding bill pay would not meaningfully change our results). And because every transaction on our platform is tied to an itemized receipt or bill, we can see not just how much businesses are spending, but exactly what they're spending on.

Second, our dataset covers a broad range of U.S. businesses that use Ramp. It’s a mix of large enterprises (like Stripe, Shopify, and Visa) and SMBs (construction firms, dentists offices, schools, churches, and everything in between). Our data is always anonymized and aggregated, and we apply standard econometric methods to ensure no one business throws our results.

Third, our dataset covers a broad range of sectors. Software startups make up less than 10% of Ramp’s active customer base. Our largest customer segments are professional services, health care, software, manufacturing, construction, and finance.

Where we skew

No private dataset will perfectly reflect all firms in the U.S. — our dataset is no exception (most U.S. firms are smaller than the typical Ramp customer). For this reason, we break out Ramp AI Index by business size and sector, and make the data available for direct download, so that researchers can make their own estimates using their own weighting.

While Ramp customers don’t skew toward tech startups, they skew high-growth and tech-forward. We’ve observed they are growing faster than the average U.S. business, and they are, after all, adopting a tech-forward financial operations platform. My preferred framing is not that Ramp businesses are reflective of the average American business, but that Ramp businesses are reflective of high-growth, fast-moving, tech-forward businesses. The ones other businesses should and will end up learning from. And therefore, if you want to see where business is going in the future, look to trends in Ramp data.

In my role, I often speak to hedge funds and quant traders who want to acquire Ramp data to assist their own financial models (we decline, Ramp doesn't sell data). I often ask them, if not for Ramp data, what would you use to track spend and AI trends among American businesses? They will usually show me a messy dataset with no methodological information and no source information, sold for high-dollar amounts by a shady data aggregator. I am routinely shocked at the lack of available data in this market, even to the firms willing to pay millions.

Like all researchers, we do the best with the data we have. We make no misrepresentations, and we don’t make claims if we’re not confident about the work that got us there. We welcome your thoughts, ideas, and questions, and believe that better data, from multiple sources, will produce better decision-making, better businesses, and a better economy.