General Ledger Review

For the reliability of the financials, the two documents that matter first are the general ledger and the bank statements. Monthly profit and loss statements and monthly balance sheets come with them, but those are outputs. The ledger is where the transactions actually live, and reconciliations and tie-outs have to happen before anyone starts the analysis.

A common and frustrating data room state is a well-populated folder containing trial balances and no general ledger. Trial balances are useful and their presence is usually a good sign about the bookkeeping. They are also summaries, and you cannot reconcile anything from a summary. The first three or four days of an engagement are ledger work, so a missing ledger stalls everything else.

Direct read-only access to the accounting system, where the seller is willing, collapses a significant portion of the request list on its own. Instead of asking for reports and waiting, we spend two or three days inside the file and come back with targeted questions. We have waited weeks for general ledgers on deals where access was not available. QuickBooks Online versus desktop makes a real difference here.

Ledgers exported as PDF have to be converted before they can be worked with, and on a business of any size that means hundreds of pages of conversion before the first reconciliation can begin. It is not analysis time, it is mechanical, and it is invisible to everyone waiting on the report. Where the accounting system can only export PDF, that is worth knowing at scoping, not at day four.

The same applies to historical data after a system migration. We have run engagements where the only surviving record of the prior year was a set of PDF backups the bookkeeper had saved before the switch, which puts a hard ceiling on how deep the historical analysis can go.

The ledger is not read line by line. It is read for patterns, and a handful of them come up repeatedly.

• Description quality. A ledger where deposits post as “deposit” with no customer attached, or where entire runs of activity read “AR invoice post” and “AP invoice post”, tells you very little about what happened in the business. We typically go back and ask, and where the descriptions cannot be improved we rebuild from underlying records instead.
• Month-end adjusting entries and reclasses. These are among the first questions we take into the management call. Noticing in the ledger that a company books month-end reclasses and asking the seller to walk through why is a much better conversation than asking about accounting policy in the abstract.
• Year-end activity from the external accountant. Where the CPA does the annual cleanup, December often looks strange, with journal entries hitting cost of goods sold and subcontractor accounts in both directions, payable adjustments and rebates moving individual expense accounts. That activity is legitimate. It also has to be understood before the monthly profile of earnings means anything.
• Accounts that are not what they are labelled. A one-time non-operating expense account may contain ordinary operating costs sitting in the wrong place, or a mix of genuine business expenses and personal spending that the accountant has partially reallocated to owner draws. What sits in that account frequently drives the add-back argument, so the label is not enough.
• Entity separation. On multi-entity and carve-out situations, ledgers have to be split by entity and by year or the comparison is meaningless, apples against oranges. We have had to go back and get them reissued that way before any revenue analysis could be trusted.

When the ledger does not carry enough detail, the answer is not to accept the summary. It is to rebuild from the layer underneath it.

On one engagement the ledger recorded payables activity with almost no description, so we requested the full accounts payable subledger and reconstructed operating expenses line by line to see whether they matched the underlying invoices. On another, a services business recorded wages against property names in the ledger, so we rebuilt a wage summary by payroll period, attributed each payroll run back to the specific contract it served, and extrapolated forward against contracted revenue to get a defensible margin. Two of those attributions could not be resolved from the records at all and had to be reverse engineered, which we said plainly in the deliverable rather than presenting an estimate as fact.

Detail can also cut the other way. We have worked with ledgers so granular that a single customer invoice splits into six or more lines across labour, repair orders and outside labour, depending on what was done. That makes tracing slower. It is still better than the alternative.

The proof of cash is the one people underestimate. Cash receipts and cash disbursements both get tied out, with non-operational transfers and shareholder distributions adjusted for, and where the result lands within a fraction of a percent it is a strong indication the reported revenue is real. On invoice tracing we usually sample somewhere between five and ten, weighted toward high value and high risk items and anything that looks out of place, rather than sampling at random.

A proof of cash that does not tie is not automatically a problem. In our experience there is a good reason almost every time, and it is usually one of two things: an accounting system change during the period, or a cash receipt that never made it into the financials. Both are fixable. Both take another layer of review.

That is also the point where diligence stops being a checklist. If the ledger looks odd, we go where the deal takes us and dig into it. If the tax figures stop making sense, we go there instead. The scope is defined, but which areas get the depth is driven by what the records actually show. See quality of earnings red flags for the findings that most often justify going deeper.