The Reconciliation Binder Index, Account by Account
Every balance sheet account is somebody's job every month, and the index is where that gets settled once instead of monthly.
The hard part of reconciliation is completeness, not accuracy.
Almost every company I have looked at reconciles its bank accounts well. Most reconcile receivables and payables well. The accounts that cause the damage are the ones nobody reconciles at all, and they are not neglected because they are difficult. They are neglected because nobody ever wrote down that they existed. A prepaid account created for one insurance policy in a hurried March, a clearing account somebody added when a new payment processor was connected, a payroll liability that was split into two accounts during a provider change and now only one of them gets looked at. Each of those is a small error at the moment it starts and a material one two years later, and every one of them is invisible to a process that reconciles the accounts on the list.
The index is what makes the list provably complete. That is its entire job, and it is the reason to build one even in a company where the reconciliations themselves are already good.
The published close checklist says which day the reconciliations happen on and what evidence a reviewer expects to find. This piece is the register that sits underneath it: which accounts exist, which method each one takes, how often, and how you prove that none were missed.
The completeness proof
Do not build the index from what you currently reconcile. Build it from the trial balance, and keep it tied to the trial balance.
The mechanism is a two-way check, run at every close, and it takes a couple of minutes once it is set up.
Every account on the trial balance has a row in the index. Including the ones with a nil balance, because a nil balance is a result and not an exemption. An account that ought to be nil and is nil has been proved; an account that is nil because nothing has posted to it since it was created is dormant and should be closed. The index is where you tell those two apart.
Every row in the index maps to an account on the trial balance. A row with no account behind it means the chart changed and nobody told the index. That is usually harmless and occasionally it is the tell that a balance was moved somewhere else.
The failure this catches is specific: a new account created mid-year. Somebody sets one up on a Tuesday to solve a coding problem, it works, and it is never reconciled because the reconciliation list was written before it existed. Two-way ties are how that account gets found in its first month rather than at a year end when nobody can remember why it was opened.
One instruction that makes the tie possible: the index is keyed on the account code, not the account name. Names get edited. Codes do not, and when a code does change, the tie breaks loudly, which is what you want.
The four methods
At this size there are four ways to reconcile a balance sheet account and every account takes exactly one of them. Naming the method in the index settles a surprising number of arguments, because most disputes about whether an account is reconciled are actually disputes about what reconciled would mean for that account.
| Method | What it means | What proves it | The accounts that take it |
|---|---|---|---|
| Tie to a third party statement | The balance agrees to a document produced by somebody outside the company, with differences listed individually and aged | The statement itself, plus the reconciling items with dates and owners | Bank accounts, credit cards, merchant and payment processor accounts, loans, leases, and any account held with a finance provider |
| Tie to a subsidiary listing | The control account agrees to the detail in a sub-ledger that the system maintains | The aged listing as at the close date, agreed to the control account, difference of nil or a documented reason | Accounts receivable, accounts payable, inventory where the system holds a perpetual record, and any customer deposit sub-ledger |
| Tie to a computed schedule | There is no external document and no sub-ledger. The balance is what a schedule you maintain says it should be, rolled forward each period | The schedule showing opening, additions, releases and closing, with the closing figure agreed to the ledger | Prepaid expenses, accrued liabilities, deferred revenue, capital assets and accumulated depreciation, vacation and benefit accruals, provisions, shareholder and related party accounts |
| Prove to nil | The account is a conduit rather than a balance. It should be empty at the close, and if it is not, every item in it is listed with an owner and a date | A listing showing either nil, or every residual item individually explained | Suspense, every clearing and in-transit account, payroll clearing, undeposited funds, intercompany clearing within a single entity |
The third method carries most of the risk, and it carries it for a structural reason worth naming. A tie to a third party statement fails loudly: the bank says one thing, you say another, and the difference is unarguable. A tie to a computed schedule fails silently, because the schedule is produced by the same function that produced the ledger, and a schedule that has been rolled forward with an error in it agrees to a ledger that has the same error. So the test on method three is not whether the schedule agrees to the ledger. It is whether the schedule is right, which means somebody has to read it rather than tick it, and the movement in the period is what they read.
The fourth method is the one companies leave off the index entirely, on the reasoning that an account that should be nil does not need reconciling. Exactly backwards. A clearing account carrying the same three items for six months is how a small error becomes a permanent feature of the balance sheet, and a “prove to nil” row that keeps coming back with residual items is the single most reliable early warning in this whole exercise.
Frequency, and why monthly is not the answer for everything
An index that says monthly against every row produces one of two outcomes: a close that is longer than it needs to be, or a set of rows that are ticked without being done. The second is worse, because it puts a signature under work that did not happen.
Assign frequency on two axes. How much does this account move, and what is the consequence of it being wrong for a period.
| Tier | Frequency | Applies to | Rule of thumb |
|---|---|---|---|
| Every close, without exception | Monthly | Cash and cash equivalents, credit cards, processor and clearing accounts, receivables, payables, payroll liabilities, sales tax accounts, deferred revenue, shareholder and related party accounts | High movement, or high consequence, or both. If in doubt, it goes here. |
| Every close, but as a movement check | Monthly, lighter | Capital assets and accumulated depreciation, loans and leases | The schedule is authoritative and the monthly job is to confirm the movement was posted as the schedule says, not to rebuild the schedule |
| Quarterly, with a monthly glance | Quarterly full, monthly reasonableness | Low-movement prepaid accounts, small provisions, deposits held with suppliers or landlords | Movement is rare. A full reconciliation quarterly, plus a monthly check that nothing moved when nothing should have. |
| On movement | When it changes | Long-term deposits, share capital, contributed surplus, any account that changes only on a corporate event | Reconciling an account that has not moved in three years produces paper rather than information |
| Annually, before the year end file | Once | Anything above that is immaterial and stable, plus every account carrying a nil balance, to catch dormancy | The annual sweep is also where dormant accounts get closed |
Two rules attach to this table.
The shareholder and related party accounts go in the top tier regardless of how little they move, because they are the accounts that get asked about first by anybody looking at the file from outside, and a year of undocumented lumpy movements takes days to reconstruct.
And a frequency below monthly requires a written reason in the index, one line. The discipline is not bureaucratic. It stops the frequency column from drifting downwards over time, which is exactly what it does when a busy month meets a row nobody feels strongly about.
The index fields
| Field | Note |
|---|---|
| Account code | The key. Never the name. |
| Account name | For humans |
| Method | One of the four. Not two. |
| Frequency | From the tier table, with a reason if below monthly |
| Preparer | One named person |
| Reviewer | One named person, who is not the preparer |
| Evidence required | The specific documents, listed, so that “attached” has a definition |
| Where the evidence lives | The folder path or the system location, written once |
| Materiality note | The threshold below which a difference is accepted with a note, for this account |
| Status this period | One of four, below |
| Date completed and date reviewed | Two dates, not one |
The status field has four values and only four, and the fourth is the one that earns the index its keep.
Reconciled. Agreed, evidence attached, reviewed.
Reconciled with an accepted difference. Agreed within the materiality note for that account, with the difference stated and a date by which it will be cleared. A difference accepted without a clearing date is not accepted, it is deferred.
Not required this period. Per the frequency column. This is a legitimate status and it must be visible, because an invisible one looks identical to a missed one.
Overdue. The period ended, the row was due, and it is not done. This status must appear on the close status page that the reviewer sees, in the count. A close that issues with three overdue rows may be the right decision on the day; a close that issues without anybody knowing there were three is not a decision at all.
The count of overdue rows, tracked over months, is the most useful single number in this document. It tells you whether the index is being maintained or admired.
Building it, in about ninety minutes
Genuinely a single sitting, and the fact that it is cheap is the argument for doing it now rather than at the next year end.
- Export the trial balance at the last completed close, at account code level, including nil balances. Every account, not the summary.
- Delete the income statement rows. This index is the balance sheet.
- Put the four method names in a column and assign one to every row. Do not leave any blank and do not invent a fifth. A row you cannot assign is a row you do not understand, and that is a finding.
- Assign the frequency tier. Default everything to monthly, then demote deliberately, with a reason in the cell.
- Put a preparer name and a reviewer name in every row. Where both would be the same person, leave the reviewer blank for now and count how many rows that is. That count is the input to the ownership question, which is a separate exercise.
- For the top tier, write the evidence list. Be specific: not “bank reconciliation”, but the reconciliation report, the statement, and the outstanding items list.
- Set a materiality note per account rather than one number for the whole balance sheet. A difference that is trivial in cash is not trivial in a shareholder account.
- Run the two-way tie once, now, and see what it finds. On a first pass in a company that has never done this, it finds accounts nobody expected to still be open.
Step 8 is the one that sells the exercise internally. It usually produces something concrete on the first run.
What this is not
It is not an audit programme, and the distinction is not a technicality.
This index is a management artefact. It exists so that the company can prove to itself that every account was addressed, by whom, on what basis. It is designed around what a company at this size can sustain every month with the people it has.
An assurance engagement is a different thing with a different purpose, conducted by an independent practitioner against professional standards, and it is not work I perform. A practitioner reviewing or auditing your statements will form their own view of what they need, will set their own materiality, and may well want schedules this index does not produce. Having a complete index makes their engagement faster and cheaper because the file already exists, which is a real benefit and is not the same as having satisfied anybody’s requirements in advance. The preparation for that engagement, and what a practitioner actually asks for, is set out separately in the piece on getting a first review engagement through.
The practical version of the boundary: do not write “audit” anywhere on this document, do not describe the reviewer’s mark as sign-off in an assurance sense, and do not tell a lender that your accounts have been reconciled to an audited standard. They have been reconciled to your standard, which you wrote down, which is a defensible position and a better one than most companies of this size can claim.
Two working rules follow from that boundary. A reconciliation whose difference is described as small does not pass: the materiality note in the index says what the accepted threshold is for that account, a difference inside it is accepted with a stated clearing date, and a difference outside it is worked. And an index that has drifted from the trial balance is not an index, it is a document about a chart of accounts that no longer exists. The two-way tie runs at every close, where it is cheap. A year of not running it turns it into a project, and the monthly close is the only place it was ever going to be a two-minute job.