About
Built by people who’ve had to close the month.
Mise exists because the bookkeeping software for independent operators is either a spreadsheet, a $1,500-a-month restaurant suite, or QuickBooks with a hopeful expression. We’re building the thing in the middle.
The story
Four months of untouched books, a week before federal taxes.
My wife and I co-run a small DTC brand called The Store. We pour a candle, take a photo, list it on Shopify, ship it from our basement. It is a small business in the most literal sense.
What it is not, is simple. A single Shopify payout settles gross sales minus processor fees minus refunds, with a one- or two-day lag to the bank deposit. A casual purchase on the business card is sometimes a real expense, sometimes a personal item paid back as an owner draw. The math is solvable. The tools were not.
I sat with my laptop at midnight, four months behind on the books, with a week until federal taxes. I tried QuickBooks bank feeds. I tried a popular AI categorizer. Both gave me two-line journal entries that ignored the difference between gross and net, dated the revenue to the bank deposit, and left the in-transit clearing account quietly broken.
I closed the month by hand. Twice, to check it. And then I started writing Mise.
Mise is built to do the thing I had to do that night, but correctly, on every transaction, every month, without anyone touching a journal entry by hand. Not categorization with a cleanup tax. Bookkeeping that ties.
What we believe
Five principles that fall out of having actually done this.
Accounting-correct, not QBO-matching.
Mise does the bookkeeping a good accountant would, even when your existing QBO history disagrees. Owner draws stay equity. POS deposits are transfers, not revenue. The chart is simple but tax-correct.
Show your work.
Every categorization comes with a confidence score and a reason. Every journal entry shows its debits and credits. There is no black box. If Mise gets something wrong, you can see why and fix it once.
Correct once, learn forever.
The second time you fix the same merchant, Mise writes a permanent rule. Future transactions skip the LLM entirely — faster, cheaper, deterministic. Your review queue gets shorter every week.
Properly, or not yet.
Today Mise stays narrow on purpose — money in, money out, reconciled correctly, pushed to QuickBooks. Inventory, AP, AR, payroll are real parts of running a business and we’ll get to them. But we’d rather wait until we can do each one the way we’ve done the rest — correctly, on real books, without a cleanup tax — than ship five half-implementations that look complete in a feature matrix.
The customers decide what we build next.
Every quarter we publish a short ballot of candidate features and Mise customers vote. Suggestions roll in from inside the app. The roadmap belongs to the people running the books.
Where we are
Built on our own books, first.
Mise was built first on a single set of books — The Store LLC, the DTC brand we co-run. Real Plaid bank connection, real Shopify settlements, real journal entries pushed to QuickBooks Online. The categorization agent runs at 97.9% accuracy against human ground truth on our most recent benchmark. The balance sheet ties.
We’re now bringing on founding customers — a small group of independent restaurants and DTC brands who want clean books and are willing to help shape the product as we grow. Direct line to the team, lifetime founder pricing, and a real say in what gets built next. If that sounds like you, we’d like to talk.
Who’s building this
A team of two, in Saratoga Springs.
Kevin Merck
Engineering & product
Senior Financial Systems Analyst at Liberty Mutual, where he owns a global S/4HANA billing and settlement platform. Builds Mise on nights and weekends. Has closed enough months to know which corners not to cut.
Sophie Deshaies
Operator & first customer
Co-owner of The Store LLC. Mise’s first user, harshest critic, and the reason every workflow that ships gets tested on real books before anyone else sees it.