What bookkeeping actually involves, month by month

Bookkeeping gets discussed as a single chore, which is why it feels endless and why it gets postponed as a block. It is actually three different kinds of work with three different natural rhythms, and separating them makes the whole thing tractable.

There is capture, which is continuous. A transaction happens and a record of it has to be created and kept while the information still exists. This cannot be batched, because the fact being captured decays: the reason for a payment is in someone's head on the day and gone by the quarter.

There is recording, which is periodic. Captured documents are entered against the right accounts, which is a task that batches well because it is mechanical and benefits from concentration.

And there is reconciliation and review, which is monthly and is the only part that actually checks anything. Everything before it is assertion; this is the step where the assertions get tested against something independent.

Businesses that find bookkeeping unmanageable have almost always collapsed all three into one annual event. At that point capture is impossible, because the moment has passed, so recording becomes reconstruction and reconciliation becomes a negotiation with a bank statement. The work has not grown; it has been moved to the point where it is most expensive.

Daily does not mean lengthy. For most small businesses this is a few minutes, and its purpose is to stop information from being lost rather than to produce anything.

Issue invoices for what was supplied that day, while what was supplied is still known precisely. An invoice raised at the end of the week from memory gets the quantity or the date wrong often enough to matter, and both are fields somebody will later reconcile.

Collect and keep every document that arrives: purchase invoices, receipts, delivery notes, payment confirmations. Keeping them means putting them somewhere defined rather than somewhere temporary, and a pocket, a car dashboard and a WhatsApp thread are all temporary. Photographing a receipt on receipt is the cheapest version of this and it works because it takes fifteen seconds and cannot be postponed.

Record cash movements the same day. Cash is the only category with no independent trail, so a cash transaction not written down on the day is genuinely irrecoverable rather than merely inconvenient. Petty cash vouchers belong here.

Note anything unusual while you still know why it was unusual. A payment to a new supplier, a customer part-paying because of a dispute, money moved between accounts. One line at the time removes a query that would otherwise cost a phone call in month eleven.

The concentrated work happens shortly after a period ends, and it exists to turn a pile of captured documents into a set of books.

Record the sales for the month completely, meaning every invoice raised is entered, in sequence, with none missing. Record purchases and expenses against the correct heads, with the supporting document attached to each entry rather than filed separately, because a document you cannot connect to an entry is not supporting anything.

Then reconcile the bank. Every account, for the whole period, matched line by line against what the books say. This is the step that finds what the other steps missed, and it is the one most often skipped because it is the only one that can fail.

Then check the GST position: what you have recorded as outward supplies against what you are about to report, and your purchase register against the portal data, so that mismatches surface while the supplier who caused them can still fix them.

Then prepare the summary for your accountant, and close the period, which means deciding that it is finished and that nothing further will be added to it. Closing is the part people omit, and without it the month stays editable, which means every figure derived from it is provisional and every check performed on it may need repeating.

Some tasks do not fit a monthly rhythm and get forgotten precisely because they have no regular slot.

Stock counting is the significant one. Closing stock directly determines reported profit, and it comes from a physical count rather than from any calculation, so a business that never counts has a profit figure resting on an assumption. How often to count depends on what you sell, but never is not an interval.

Fixed assets need a periodic look: what was bought, what was disposed of, what is still physically present. An asset register that has drifted from reality produces wrong depreciation for years, silently, because nothing about the calculation reveals that the machine was scrapped in 2023.

Receivables need a review that goes beyond the total. Which balances are genuinely collectible, which are disputed, which should be written off. Nothing in the ledger raises this question, so it has to be scheduled.

Statutory deadlines sit alongside all of this on their own calendar, which is set by law rather than by convenience and differs by registration and turnover. The practical approach is to write the year's dates down once, at the start of the year, with your accountant confirming which apply to you, rather than discovering each one as it arrives.

The sequence above is not arbitrary and reversing parts of it wastes work.

Reconciliation has to come after recording, because there is nothing to reconcile against until the books contain the period. But it has to come before the summary goes to the accountant, because sending unreconciled records means the accountant discovers the problems, and their time costs more than yours. It also means the problems are found by someone who cannot answer the questions they raise.

Closing has to come last, and it has to be genuine. A period reopened to slip in a forgotten transaction invalidates the reconciliation that was already done on it, and if figures have already been filed from it the books now disagree with the filing. Late transactions belong in the current period with a note, or in a formal correction, and which one applies is a question for your accountant rather than a judgement call for whoever is entering it.

The deeper reason to keep the order is that each step's cost depends on the previous one being complete. Recording from complete capture is quick; recording from partial capture involves chasing. Reconciling complete records is a mechanical comparison; reconciling incomplete records is an investigation. Each skipped step multiplies the next, which is why an unbookkept year is not twelve times a month's work but considerably more.

A well-run monthly cycle produces books that are complete, categorised, tied to documents, and agreed with the bank. That is genuinely most of what is needed and it is not everything.

It does not establish that the figures are right in substance. A transaction recorded consistently in the wrong head reconciles perfectly, because reconciliation tests totals against the bank rather than the correctness of classification. A stock figure that was estimated rather than counted flows through to profit without anything objecting. A sale that was never entered leaves no trace anywhere in the cycle, since every check in it compares records against records or against a bank statement that never saw the cash.

It does not decide anything requiring judgement. Whether a receivable is bad, whether an expense qualifies, whether a payment is capital or revenue: these are decisions, and a routine can only ensure they are put in front of someone rather than made by default.

That is also the honest boundary for any bookkeeping tool. Software can capture, categorise consistently, match a ledger to a statement, flag a missing document, flag a gap in an invoice sequence, and remind you of a deadline. Every one of those is real work removed. What it cannot do is know about a transaction nobody entered, count what is physically on your shelves, or judge whether a customer will pay. Those inputs come from people, and the entire cycle is only as good as they are.