Puzzle makes startup accounting look much less intimidating. A founder can see cash, spending, burn, runway, revenue, margin, and financial statements without building a workbook.
That is useful. It is not the same as proving the books are correct.
Puzzle is the general ledger. That means every accounting entry lands there and the statements are built from it. A wrong category, duplicate payment, missing liability, or bad revenue date can flow into the clean founder dashboard.
We would not replace an existing ledger because Puzzle looks easier. We would let the accountant close two difficult months in both systems. Every important number should trace back to its source.
What Puzzle actually handles
Puzzle connects to bank accounts, cards, payment processors, payroll, invoicing, and other startup systems. Current examples include Mercury, Stripe, Ramp, Brex, Gusto, Rippling, Rho, Deel, and services reached through Plaid.
It suggests a category for each transaction. It can match the books to bank and card statements. It supports journal entries, changes to the chart of accounts, period locks, and cash or accrual books.
A chart of accounts is the list of places where money is recorded. Examples include cash, subscription revenue, payroll, loans, and prepaid software. A journal entry records a change between those accounts.
Cash accounting records money when it moves. Accrual accounting records income when it is earned and a cost when it is owed. A company can pay $12,000 for a year of software. Its books may then record $1,000 of cost each month instead of the full amount on day one.
Revenue recognition applies the same timing idea to sales. A customer may pay for a year in advance, but the company earns that revenue over the service period.
Puzzle also shows founder measures. Burn is the cash the company uses over a period. Runway estimates how long current cash will last at that rate.
These views are the strongest part of the public feedback. Founders like seeing the business without waiting for a custom spreadsheet. The modern integrations and automatic work can also remove a lot of copying.
The accountant writes the test
The founder should not create the trial checklist alone. The accountant or bookkeeper who will sign off on the books should list the hard transactions.
Use two real months. One should include the most difficult normal activity. Import the opening balances and history needed to understand them.
The test pack may include:
- Bank and card transactions.
- Payroll and payroll tax.
- Stripe sales, fees, refunds, and disputes.
- Annual customer prepayments.
- Annual software prepayments.
- A loan payment split between principal and interest.
- Founder expenses and reimbursements.
- Equity or convertible financing.
- Accounts payable and unpaid bills.
- Foreign-currency activity.
- A transfer between two bank accounts.
- A duplicate or reversed transaction.
- A customer credit and write-off.
The exact list depends on the company. This is why a generic demo cannot prove the ledger.
For every case, the accountant should state the right accounts, date, amount, supporting file, and treatment. Import it into Puzzle and the old ledger. Compare the entry and the final statements.
Do not use this review as an accounting policy. Ask the company's qualified accountant and tax advisers to decide the correct treatment.
Automatic categories need an exception process
Puzzle says connected context can support very high automatic category rates. Even a 98% rate means 20 items need attention in a month with 1,000 transactions. The two wrong entries may matter more than the 980 easy ones.
Test similar vendors that belong in different places. A payment to Amazon could be office supplies, hosting, or inventory. A Stripe deposit is not simply revenue because it may include fees, refunds, tax, and sales from several periods.
Change a wrong suggestion. Check whether Puzzle learns the rule and whether the correction affects only the right transactions. Split one transaction across several accounts. Add a note and source file.
Then undo the correction. Review who changed it and when. Make sure the accountant can approve exceptions without losing the original evidence.
Some public feedback warns about incorrect automatic categories and frustrating overrides. A competing accounting vendor supplies part of that criticism, so its framing deserves caution. Similar concerns also appear in practitioner discussions.
The decision is not whether AI makes a mistake. It will. The decision is whether the mistake is visible, quick to correct, and prevented from returning.
Fast wrong books are still wrong.
Reconcile every balance
Reconciliation means matching an account in the ledger to an outside statement. The ending cash balance should match the bank. Every difference should have an explanation.
Close both test months. Reconcile every bank and card. Match Stripe gross sales, fees, refunds, disputes, and payout timing. Match payroll expense, cash, taxes, and liabilities to payroll reports.
Check accounts receivable and payable. Confirm deferred revenue and prepaid costs move over the right months. Match loan balances to the lender. Tie equity to legal and capitalization records.
Then compare the trial balance, profit and loss statement, balance sheet, and cash activity. The same accounting policies should produce the same result in Puzzle and the old ledger.
A mismatch may reveal an error in either system. Investigate it. Do not adjust a number merely to make the totals agree.
Lock the first month. Try to edit a dated transaction and post a journal into the locked period. The system should follow the company's control rule. Then test the authorized way to reopen and document a correction.
Check the audit history. It should show the old value, new value, person, time, and reason. Ask what remains after an item is deleted.
Puzzle's current pages include real accounting controls beyond founder charts. Practitioner feedback still suggests some controls and reports may feel less mature than established ledgers. This close is where that difference becomes visible.
Test revenue with source schedules
Puzzle can be especially interesting for a software company with prepaid subscriptions and deferred revenue.
Use several contract shapes. Include a monthly plan, an annual prepayment, a midterm upgrade, and a refund. Add a discount, a cancelled contract, and a start date that differs from the payment date.
Create the expected revenue schedule outside Puzzle and have the accountant approve it. Then compare each month's recognized revenue and remaining liability.
Tie total customer cash to the payment processor. Tie earned revenue to the profit and loss statement. Tie unearned revenue to the balance sheet. Test a change after a period is closed.
Revenue rules can have serious accounting and tax effects. The software can calculate a schedule. The accountant must decide whether the policy and inputs are correct.
Build the real packages
The founder dashboard should be checked only after the books close.
Recalculate cash from bank balances. Recalculate monthly net burn under the company's definition. Check how one-time financing, large annual collections, and transfers affect the runway chart.
Build the package used by the board or investors. Include the exact profit and loss, balance sheet, cash flow or cash activity, budget variance, revenue, burn, and runway needed. Check comparison periods and export quality.
Build the tax-preparer package too. Export the general ledger, trial balance, statements, chart of accounts, and journal entries. Then export the matching work, attached records, vendor details, and change history.
Ask the accountant to import the exports into the next tool or a blank work file. A download button is not proof that a complete accounting record can leave.
Puzzle offers a one-click tax report. The tax adviser still needs to confirm that it contains the right records for the company and jurisdiction.
Pricing depends on the accounting depth
Puzzle offers a 14-day trial with access to the Complete plan. It does not require the company to keep that plan afterward.
The annual list price is $25 a month for Starter and $60 for Core. Complete is $100 a month, while Scale is $300. Monthly list rates are $30, $72, $120, and $360.
Starter supports one person and the first books. It includes categorization, drafted statements, cash and accrual books, connections, a tax report, and basic AI help.
Core supports five people. It adds cash, burn, runway, margin, spending, revenue, variance, custom accounts, advanced reports, and a self-guided QuickBooks move.
Complete includes unlimited people, white-glove QuickBooks migration, better accuracy and reconciliation help, classes, departments, projects, and revenue recognition. It includes 100 AI credits per month.
Scale starts with higher volume, subledgers, backup and restore, and fewer limits. It also has 300 AI credits and dedicated setup and support.
The live page also shows introductory discounts. Starter may be free for two months. Core, Complete, and Scale show 50% savings for three months. Treat those as temporary. Budget with the renewal rate and applicable tax.
Puzzle advertises a close-time guarantee for eligible customers. It requires timely onboarding, supported live connections, a single entity, a full monthly close, and two months of use. The refund covers Puzzle subscription fees, not implementation, tax, or other costs. Time savings are customer-reported. The guarantee does not promise correct accounting.
Record the hours for migration, transaction review, reconciliation, corrections, reports, and accountant review. Compare the second close with the old process only after both are complete and correct.
Who should choose Puzzle?
Puzzle is worth testing for an early-stage startup with straightforward entities and modern finance tools. It may be especially useful when the founder needs clear cash and revenue views. The accountant still needs to support the setup.
It is harder to recommend for a company with complex entities, inventory, or unusual tax rules. Skip it if you need deep firm controls or reports that the trial cannot produce. The public independent record is thin for the responsibility involved.
We did not connect an account, move books, close a month, or verify a report. We also did not test a tax export or contact support. The strongest sample is founder-heavy. Accountant voices are fewer and more cautious.
Our rule is strict. Keep the old ledger active. Let the accountant close two real difficult months in both systems. Review every automatic exception. Reconcile cash, cards, payments, payroll, receivables, payables, debt, equity, deferred revenue, and prepaid cost.
Lock and correct a period. Produce the founder, board, and tax packages. Export the whole ledger and audit history into a usable form.
Choose Puzzle only when the accountant can sign off on the books and the founder can use the dashboard. The runway chart is the last thing to trust, not the first.
