One graph. The whole economy.
The Spend Graph is our low-latency, queryable model of the American economy. It indexes 2 billion transactions per month, resolved to business, location, and day.
Economic intelligence, built from the ground up.
A single card payment seems like one event. It’s actually several: a hold, a settlement, sometimes an adjustment or a refund, each recorded by a different system under a different name. What comes out is a pile of records no one can read, let alone query.
The Spend Graph makes sense of it all. Every record is isolated, then de-duplicated, mapped to the correct merchant, and finally structured by spending group, industry, location and time. Our mission is to build the most accurate model of American spending, allowing everyone to understand the economy they participate in.
Six steps. One record.
How every transaction joins the Spend Graph.
01
Sourced
The Spend Graph begins with billions of transactions from across the country, alongside census data and macroeconomic benchmarks, refreshed daily. At this stage, every record is de-identified and aggregated.
02
Cleaned
Before anything is counted, duplicates are reconciled and refunds are paired with the purchases they reverse. Each transaction is then standardized to one format, regardless of which bank, network or payment app carried it. What remains is one clean line per purchase.
03
Enriched
Each alias, store number, and payment-app wrapper is enriched by matching it to the business behind it in a process we call resolution. Each match is a statistical call made against everything else the Spend Graph sees, so it sharpens as more data comes in.
04
Structured
Next, each purchase is placed in the economy around it: by industry, location and customer. A merchant’s category reflects what the business is. Location is tracked two ways, where the store is and where its customers live, counted only in aggregate. Metros follow standard statistical areas, so answers line up with public data. Spending cohorts, such as household income, show who’s buying, not just how much.
05
Refreshed daily
New transactions arrive daily. A day’s first read lands within hours, most of it is in by 2 days, and it is complete by 7 days. We keep a record of each revision, so a figure that moves can be traced. History runs back to April 2025.
06
Queryable
Finally, the full record can be queried by merchant, category, place or window, from any AI tool that connects over MCP. Every answer returns transactions, unique purchasers, spend and average ticket.
American spending, free and open to anyone.
Use Claude or your favorite AI tool to explore $1.2T in consumer spending. Connect it in a minute, and ask your first questions free.
You asked: How fast is OpenAI growing with consumers?
2.3M
accounts paid OpenAI in September. Up from 1.5M in April. Monthly spend rose from $40M to $73M.
Try asking
From the American Spend model.
How close is it?
Three studies compare the Spend Graph with reported revenue for 53 companies, data included.
More questions?
What the Spend Graph covers, how we check it, and the questions we hear most.
Citing us?
The press kit has the source line, logos, boilerplate, and who to ask when a number needs checking.
What’s next?
Spending between businesses, item-level data and more small merchants. Here’s what we’re adding.


