Bayesian Surprise is a risk and data science practice in Atlanta. We sharpen the scoring systems small-business lenders already run on, starting from what's in place rather than tearing it out, and we've spent long enough on the other side of the table, as the risk officer and as the capital, to know what the score has to do and what it has to survive.
Three jobs that usually get handed to one person: risk management, credit, and data science. We split them the way they should be split and bring all three. The practice was founded by Dennis Carlson. His record follows, in the order a stranger can check it.
This whole site is asking you to trust people you haven't met. So here's the part you can verify in thirty seconds, then the part you'll have to take our word for, then the part a sales page would leave out.
From 2012 to 2016 Dennis was Deputy Chief Economist at Equifax. He co-wrote the subprime auto papers with the chief economist, Amy Crews Cutts, argued the bubble question in public, and got quoted for it in the trade press, sometimes approvingly and sometimes not. It's all still out there.
Equifax economic commentary on subprime auto, the paper that started the argument.
Auto Finance News: "Equifax economist defends subprime auto", from the Consumer Bankers Association conference.
American Banker author page: Cornell statistics, American Express, First Data, Equifax.
None of this is public, which is how it works inside big companies. It's the part he's proudest of, so it's here anyway.
Head of Consumer and Small Business Credit. Thirty people, credit policy and risk systems across fourteen billion dollars a year in originations. He redesigned the credit structure from a percentage deposit to a fixed-dollar limit, rebuilt the unit economics underneath it, and walked the change through Verizon's infrastructure for two years. It produced $350 million in incremental revenue and $170 million in EBITDA in its first two years. He also built the first-party fraud model that cut "no intent to pay" losses in half, and worked with Treasury to keep credit policy inside the device-payment securitization's covenants.
Director of statistics and fraud product development. Fraud models for forty banks. Found and broke a merchant-processing fraud ring, eighty percent less fraud loss within six months. Built the common-point-of-compromise system that traced bad transactions backward to the terminals they came from. Bayesian fraud models in R, back when that was an odd thing to do.
Director of quantitative analytics. Built the automated modeling platform in SAS, sampling through validation through scoring, before "data science" was a job title. Eighty-plus predictive models. The transaction triggers that found small-business owners hiding inside the consumer card base and quintupled prospect identification.
From 2019 to 2023 Dennis was Chief Risk Officer at Credibility Capital, a small-business lender making fairly priced term loans up to $500K. He took the risk function from a credit policy and a spreadsheet to the thing this site describes: gradient-boosted models, a cash-flow underwriting system built on raw bank transactions, a proprietary score blending personal and commercial bureau data. The bank-statement model alone reached a KS of 0.59 on holdout. Predictive accuracy went up one and a half times. Origination time came down 28 percent and conversion went up 13. They mined tax returns, parsed P&Ls off paper statements, tested keystroke analysis for fraud. They tried everything.
And here's the part a sales page would leave out. The company closed in January 2023, for a legal reason that had nothing to do with the models. He walked away convinced of something he'll say to your face: you cannot underwrite unsecured small-business term loans past a year or so without a very tight niche. Either you go niche, or you go short and price for it. That's your business. It's why we're standing in your doorway and not a bank's.
Mathematics and statistics at the University of Florida, valedictorian. The Ph.D. program in statistics at Cornell, left for American Express because building things beat proving things. Twenty years later, a master's in counseling, on purpose, because every system he'd ever built was run by people and he wanted to understand them as well as he understood the models.
Since 2023 he has been Chief Credit Officer at a private credit fund that lends to small-business lenders, which means two years of doing diligence on underwriting shops like yours from the capital's side of the table. We know what the people who fund you are looking for, because we've been the ones looking.