$30B is the largest data set we know of that has been trained specifically on startup accounting. And it’s growing fast. Like, really fast.
Why does Puzzle’s $30B training set matter to your startup?
A bigger training set reduces the time it takes to get accurate books.
At a high level, there are two foundational concepts in early stage accounting: completeness and accuracy. Having access to a large amount of data, like Puzzle’s $30 billion in transactions, helps in achieving better accuracy. This large dataset is tailored to meet the specific needs of startups. It helps in categorizing transactions accurately, handling unusual situations and edge cases, and setting up effective processes and rules that are important for the accounting of startups.
This results in more accurate financial reports. Since Puzzle also built the underlying accounting core (general ledger), you can get drafted financial statements every day just by using our proprietary software.
In QuickBooks, or at companies that use their general ledger software, accounting accuracy is only as good as the bookkeeper and their attention to detail. QuickBooks adopts a data quarantine model, a bookkeeper manually reviews each transaction categorization, manually finalizes the reconciliations, and then manually generates the financial statements from the general ledger. Consequently, financial statements are prepared and closed weeks after the month ends.
Puzzle’s $30B training set gives your startup a major advantage:
Scalability + Accuracy + Automation = Daily Drafted Financials and Insights
Why can’t QuickBooks generate daily financial statements like Puzzle?
It would require rebuilding the underlying software architecture and user experience focused on automation. Here’s what we did differently at Puzzle:
- Chart of accounts standardization leads to more accurate categorization: QuickBooks does not enforce standardization of the charts of accounts (i.e. categories), so every company builds their chart of accounts in a silo. Puzzle assigns every company a chart of accounts template by industry that can be customized without losing automation today, or as you scale.
- More accurate categorization enables automated journal entries (debits and credits): QuickBooks serves a wide range of business models, from dentists to donut shops, and processes trillions of dollars in transactions. Large diversity combined with unstandardized data structures leads to low confidence in categorization accuracy. As a result they must adopt a data quarantine model so bookkeepers can review or manually create rules for each transaction type. By focusing on startups, Puzzle can create (accounting) journal entries from categorizations with higher confidence, and tag the remaining low-confidence items for manual review.
- Automated journal entries enable daily drafted financials: Service providers like Pilot or Kruze which work in QuickBooks do not release financials until the “books are closed”, typically weeks after the last day of the month. In contrast, a high-confidence system that can generate draft financial statements with software, assigning a real-time completion status, which enables founders to use daily drafted financials in real-time to make business decisions.
- Even with accrual accounting: Puzzle is a native accrual accounting platform, so journal entries that involve accrual schedule management (bills, invoices, credit cards, payrolls, prepaid expenses, revenue recognition, etc.) are done inside the application. In QuickBooks, accrual schedules are handled manually in spreadsheets, making them tedious and prone to errors.
Examples which bring this to life
- A startup receives a large, unlabeled deposit into their bank account. That deposit could be revenue, but it also could be a SAFE, a convertible note, debt, or a priced venture round. QuickBooks is not trained on these types of deposits, and typically guesses ‘revenue’. Puzzle recognizes that large, unknown bank deposits need additional context, automatically notifies users of their existence, and allows users to describe the transaction in natural language before selecting a category.
- An invoice could be a customer paying you via an annual prepaid contract, monthly subscription, implementation fees, transaction fees, usage fees, or a combination of multiple revenue types. Depending on which it is, it could be recognized revenue on the income statement or deferred revenue on the balance sheet. As most small businesses don’t require revenue recognition, this is typically done in a spreadsheet or third party tool.
- An unknown purchase at Apple could be many things. For startups that are buying expensive MacBooks for engineers, most will need to be depreciated over the next 24-36 months, performed manually your bookkeeper each month.
The Puzzle rules engine is trained on data from thousands of startups, including those from Y Combinator, Village Global, Techstars, On Deck, and 500 Startups, as well as portfolio companies from Sequoia, A16Z, General Catalyst, and Felicis, and many, many others. In addition, startups that use Stripe, Ramp, Brex, Mercury, Meow, BILL, and others provide additional metadata which allows your startup to benefit from all their transactions.
Upgrade from QuickBooks to Puzzle in just a few minutes while preserving your historical books.
Learn more about why growing companies choose Puzzle over QuickBooks here.
_____
Important notes
- Yes, of course Puzzle does double-entry accounting, which was first documented in 1494.
- The simplification of “complete and accurate” is for the non-professionals reading above, ignoring core accounting concepts of “occurrence”, “classification” and “rights & obligations” for simplicity. We acknowledge the term “accuracy” in typical management assertions with respect to financials isn’t actually about the accuracy of the categorization of transactions; it’s about the accuracy of amounts.
- Manage expectations → We automate the “automatable”, and tee up the bookkeeper for the rest.
.jpeg)
