Why Plugs and Unexplained Entries Hide the Problem

A reconciliation difference is the amount by which two records that should agree don't. Examples are the bank statement and the cash ledger, or a subledger and the general ledger. A plug is an entry that makes the two sides agree without identifying why they differed. It usually goes to an account called "other", "miscellaneous" or a suspense account.

Here is an example. In March, the cash ledger shows $1,240 more than the bank. You can't find the reason by deadline, so you credit cash $1,240 and debit "other". The reconciliation ties.

The real cause was a customer receipt imported twice into the ledger. That cause is still there. The customer's receivable is understated by $1,240, and the import problem is still active. In April, another receipt of $860 duplicates. If it is plugged again, "other" now holds $2,100. In May, a $1,250 duplicate brings it to $3,350.

Each month looks clean. Meanwhile the books get further from reality, and nobody has fixed the import.

A correcting entry is a journal entry that fixes an error in the ledger. A good one is specific. It names the error, points to the source document and carries an amount you can recompute.

An unexplained correcting entry looks like a fix but is really a plug with better paperwork. It moves the problem out of the reconciliation and into the ledger itself, where it is harder to find.

Watch for these warning signs:

• Vague description, such as "adjust to tie", "per recon" or "true-up"
• No reference to a source document, transaction ID or written explanation
• An amount equal to the unreconciled difference, down to the cent

For example, an entry dated 31 October reads "per recon, $4,812.37", with no attachment. That amount is exactly the open difference on the intercompany reconciliation. Six months later, an auditor asks what it fixed. Nobody can say, including the person who posted it. Nobody can say whether it was right either.

A reviewer or auditor looking at your reconciliation needs to follow one path:

difference → cause → evidence

The question is whether they can get from the amount that didn't tie to the specific reason, and then to the record that proves it, without having to ask you.

For example, a reviewer sees a $1,240 difference on the March bank reconciliation. The note says it was a customer receipt imported twice and gives both ledger line references and the single matching bank line. They open the records, see the duplicate and move on. The path is complete.

Now take the plug version. The reviewer sees $1,240 posted to "other" with the note "to tie". The path stops at the first step. They have to find you, and you have to remember. If you've left the company, or it's eighteen months later, nobody can rebuild the path.

A plug or an untraceable entry doesn't just look bad. It breaks the one thing a reviewer relies on.

Netting happens when unrelated items of opposite sign offset each other, so the total difference looks small or even zero.

For example, a subledger-to-ledger reconciliation shows a net difference of only $15. It's tempting to call that close enough. But underneath are two separate problems:

• An invoice of $3,400 was duplicated in the ledger, so the ledger is $3,400 too high.
• A rejected batch of $3,385 never loaded into the ledger, so the ledger is $3,385 too low.

The net is $15. The actual errors add up to $6,785, and both need fixing. Plugging the $15 would bury both.

The rule is that a small net difference still needs its components examined. The next lesson shows how to split a difference into its parts. For now, try the sliders and see how the net shrinks while the real errors stay large.

This course holds every difference to one standard. Each difference ends in one of two ways:

1. Resolved. The difference has an explanation the records support. You know the cause, you can point to the evidence, and any entry you post traces back to it.
2. Open and documented. You record it as an open item: an unresolved difference written down with its amount, its age, what you checked, what is still unknown and who acts next.

A plug and an unexplained entry are not on this list.

An honest open item is a legitimate result, not a failure. Suppose a $2,300 intercompany difference can't be explained because the other entity hasn't sent its detail. Logging it as open tells the reviewer the truth: "We know the amount, we checked our side, we're waiting on Entity B's subledger, and Priya will follow up by the 10th." A reviewer can work with that. They can't work with a $2,300 plug.

AI tools can make plugging easier. If you ask one to "write a journal entry to clear this $4,812 difference" or "give me an explanation for this variance", it will produce something that sounds plausible. The entry will look professional and the explanation will read well. But neither is based on your records. It is a more convincing plug.

This course uses AI differently. You give it the actual reconciling records, with sensitive identifiers removed. You ask it to propose candidate causes tied to specific rows, for example: "rows 14 and 15 have the same amount, date and reference, so this may be a duplicate." A candidate cause is only a lead. You then test each one against the source records and mark it as supported, contradicted or unverified.

The AI suggests where to look. You and the records decide what is true. Later lessons cover how to prepare the data safely and how to word these requests.

Two questions will come up quickly: how small a difference has to be before you can stop investigating, and when a difference can be written off.

This course doesn't answer either one. Investigation thresholds and write-off decisions come from your organization's policy, along with whoever owns that policy, such as your controller or accounting policy team. If your policy says differences under a set amount may be cleared to a specific account with approval, follow it and document that you did.

What this course does teach is the method for everything else. You break the difference down, propose causes with AI, test them against the records, classify each component, and document the result. Remember that a small net figure can hide large components, as you saw with the $15 example. So apply the threshold to what you actually find, not only to the net figure.

Put the steps below in order. This is the path the remaining five lessons follow.