Most closes do not run late because the team is too small. They run late because the work has dependencies nobody wrote down, so each period somebody rediscovers the order, gets blocked waiting on a document that was never requested, and finishes in the middle of the following month with the numbers already stale.
A checklist fixes the ordering problem. What follows is the sequence, what belongs in each stage, and the test that tells you a stage is actually finished rather than just worked on.
Why a day count is the wrong target
Close checklists are usually written as a calendar. Day one this, day two that. The trouble is that a calendar describes how long the work took at whoever wrote it, not what your file needs, and it gives you a target you can hit by skipping things.
Dependencies are the real structure. You cannot reconcile an account until every transaction that belongs in the period is posted. You cannot produce a balance sheet worth reviewing until the reconciliations are done. You cannot explain a variance until the statements are final. Each stage needs the one before it to be genuinely complete, and that is what makes a close slip: not the size of any stage, but starting stage three while stage one is still open, then reworking it.
So the checklist below is ordered by what depends on what, and every stage has an exit test. When the test passes, move on. When it does not, the stage is not finished, whatever the date is.
Before the month ends
Everything here is work that can happen early, and moving it earlier is the single biggest change most files need.
- Vendor bills entered as they arrive rather than in a batch afterwards
- Expense reports submitted and approved before the last day
- Sales invoices issued and synced from whatever system raises them
- Bank, card, and payment processor feeds confirmed as connected and current
- Recurring entry schedules updated for anything new: a fresh annual software payment to amortize, an asset bought this month, a contract signed
The exit test is simple. On the last day of the month, is there a pile of work that has been waiting, or not.
Categorization is the part that decides this. If transactions are categorized within a day or two of posting, the month ends with the population nearly complete. If they are left, the first stage of every close is a catch-up sprint against a month of backlog, and each subsequent stage waits on it. Our daily bookkeeping piece covers the cadence that keeps this from accumulating.
Stage 1: lock the transaction population
The goal is that no new transaction belonging to this period will appear after this point.
- Final sync from every bank and credit card account
- Final sync from payment processors
- Payroll for the period posted, including any run that lands after month-end but covers days inside it
- Any cash movement recorded by hand: owner contributions, transfers, wires that never appear in a feed
Exit test: the closing balance in the ledger for every cash and card account matches the closing balance on that account's statement. Not approximately. Exactly, or with a listed set of items that explain the difference and each of which you can point at.
This is the stage that decides everything downstream. A transaction discovered during stage three sends you back through stages one and two, and that rework is where most of the lost time actually goes.
Stage 2: recurring and adjusting entries
These are the entries that exist because accrual accounting puts costs in the period they belong to rather than the period the money moved.
- Payroll accrual. Where a pay period straddles month-end, the wages earned in the closing month but paid in the next one belong to the closing month. Reverses at the start of the following period.
- Prepaid amortization. An annual insurance premium or software subscription paid up front sits on the balance sheet and moves into expense a month at a time.
- Depreciation. Equipment and capitalized costs write down over their useful life, on the method in your fixed asset policy.
- Deferred revenue recognition. Cash collected ahead of delivery is a liability until it is earned, then it releases into revenue over the term.
- Stock compensation. Monthly expense from the grant schedule, which your cap table provider will export.
- Accrued expenses. Costs incurred where the bill has not arrived yet, which is common with legal, accounting, and contractors who invoice in arrears.
Each of these should be read off a schedule that carries forward, not recalculated from source documents every month. The schedule is the work; the entry is a lookup.
Exit test: every recurring entry on your list is either posted or has a written reason it does not apply this period. A blank is not the same as a zero, and the difference matters when somebody reviews it later.
Stage 3: reconciliations
Every balance sheet account holds a number that is supposed to represent something real. Reconciliation is the act of proving it does.
Work through them: bank and card accounts against statements, accounts receivable against the invoice detail, accounts payable against open bills, payroll liabilities against the payroll provider, prepaid and fixed asset balances against their schedules, deferred revenue against the contract terms, equity against the cap table.
Exit test: for each account, a difference column that reads zero. Where it does not read zero, the residual is written down as a number with the two sources named, and it stays open.
That second half is the part worth insisting on. When a balance will not tie, the tempting move is an entry that forces it, and the reason to refuse is that the entry hides the evidence. The difference is information about where something is wrong, and posting over it means paying for the same investigation again next month with the trail gone. Our walkthrough of accounts receivable reconciliation shows the mechanics on one account in full.
Stage 4: statements and variance
Now the numbers are worth reading.
Produce the profit and loss, balance sheet, and cash flow statement, then go through them looking for what moved. Compare against the prior period and against budget if you have one. For each line that moved more than you expected, find the reason before anyone asks.
Exit test: every material movement has a written explanation, and every balance sheet line is one you could explain to somebody who has never seen the file.
A variance you flag yourself reads as control. The same variance found by an investor in a board meeting reads as a surprise, and the number is identical in both cases.
Stage 5: review, then lock
Someone other than the preparer reads the statements and the commentary where that is possible. Then the period is locked in the accounting system so nothing can post into it without an explicit reopening.
Locking matters more than it sounds. Reopening a closed period in some systems reaches further back than people expect, and figures that have already been reported can move underneath a document that quoted them. Lock the period, and require a named approval to reopen it.
Exit test: the statements you shared and the ledger agree, and will still agree in six months.
The close memo
At the end, write one page: the statements, cash and runway, what moved and why, anything unusual, and the open items carrying into next month.
Almost nobody does this and almost everybody wants it later. It is what makes an investor update quick to write, what an auditor asks for, and what tells you next quarter why a number looks the way it does. The open items list is the most valuable part, because it is the only place a known unresolved problem is written down rather than remembered.
Where closes actually break
- Uncategorized backlog at month-end. Stage one becomes a week of catch-up and every other stage waits behind it.
- Schedules rebuilt from scratch. Prepaid, depreciation, and deferred revenue calculated fresh each month, usually by the one person who knows how.
- The checklist living in somebody's head. No shared list means the order gets rediscovered monthly, and the company cannot close at all when that person is away.
- Reconciliations treated as optional when time is short. This is the one that compounds, because an unreconciled month becomes the opening position of the next one.
- Differences plugged rather than parked. Fast this month, expensive every month after.
A note on the R&D credit
Startups often accrue an R&D credit benefit monthly, and the payroll tax offset is the version that produces cash rather than a carryforward. It is worth being careful about the eligibility test, because it is commonly stated as a revenue ceiling alone.
There are two conditions, not one. Gross receipts for the credit year must be under five million dollars, and the company must have had no gross receipts at all in any year before the five-year period ending with that year. The second is a hard cutoff on the first dollar of revenue the company ever received, and companies fail it without realizing because they assume being small is enough. The election also has to be made on an originally filed return, not an amended one. Our R&D tax credit guide has the detail, and it is worth confirming both tests with whoever prepares your return before modeling any cash from it.
Getting started
Map what your close does now. Write down each thing that happens and what it waits on, and the two longest waits will be obvious. Almost always one of them is transaction backlog and the other is a schedule being rebuilt.
Fix those two, then write the list down properly with an exit test on each line. The point of the checklist is not that it is clever. It is that it is the same every month, so the close becomes something the company does rather than something a person remembers.
Median keeps the books current underneath all of this, with transactions categorized and posted every business day, so the close starts from a complete month instead of a backlog.