The deferred revenue rollforward that ties but cannot be rebuilt

The deferred revenue rollforward is a simple statement. Opening balance, plus what was added this period, less what was recognised, plus or minus adjustments, equals closing balance. Closing agrees to the general ledger. The schedule is signed off and the close moves on.

Then someone asks which contracts make up the closing balance, and the exercise that produced the tie-out turns out not to have produced that.

Agreeing at the total is not the same as agreeing

A rollforward built at the account level is assembled from movements: what the recognition run posted, what billing added, what the manual entries did. Those movements come from the ledger, and they agree to the ledger because that is where they came from.

A rollforward built from the schedules is assembled from contracts: every active arrangement, its remaining unrecognised consideration, summed. That is a different calculation from different data.

Both should give the same number. When they do not, the difference is usually posted as an adjusting entry so the statement balances, and the statement is now correct in the only sense it is being checked.

Where the difference actually comes from

Rarely from arithmetic. Almost always from a schedule the ledger does not know about, or a movement the schedules do not.

Schedules held outside the system. A contract with terms the ERP could not express, so its schedule lives in a workbook and its journal is posted manually. The ledger has the entry. Nothing connects the entry back to a schedule the system holds.

Modifications applied to one side. A term change reflected in the schedule but posted through a manual entry, or posted correctly and never reflected in the underlying schedule.

Manual entries with no contract. Reclassifications, corrections, an entry from two quarters ago that nobody can now attribute. These sum into the balance and belong to nothing.

Recognition run against stale terms. The run posts what the schedule says. If the schedule was not updated for a change that happened in the period, the entry is internally consistent and wrong.

Each of these is individually explainable in the month it happens. The reason they accumulate is that explaining them is somebody's memory rather than a stored fact, and memory does not survive a quarter, a reorganisation, or a departure.

What rebuildable means

A rollforward is rebuildable when every line in it can be decomposed to the contracts underneath, and every movement points at what caused it.

That requires four things:

One source produces both sides. The schedules generate the journals. The subledger sums the schedules. The ledger receives the journals. Agreement is a property of the construction, not the outcome of a check.

Every movement carries its cause. A new contract, a modification with its effective date, a recognition run for a period, a cancellation. Not just an amount and a date.

Adjustments are attributed or they are not made. An entry that cannot be tied to a contract is a question that has not been answered yet, and recording it as a balancing figure closes the question without answering it.

Prior periods stay as they were. The balance reported last quarter should still be reproducible next year, under the terms and prices that were live at the time, not recalculated with whatever the fields hold today.

What this is worth outside the close

The immediate return is a shorter close and a schedule nobody has to defend from memory.

The larger one arrives later, when the question comes from outside the team. An auditor sampling contracts and tracing to revenue. A diligence process asking for the balance decomposed by cohort. A restatement question about a period whose owner has left.

At that point the difference between a rollforward that ties and one that can be rebuilt stops being a matter of housekeeping. One of them answers the question in an afternoon. The other starts a project.

Your accountants decide the treatment and your auditors test it. What the system should be doing is running that treatment consistently and holding the evidence, so that the answer to "how did we arrive at this balance" is retrieved rather than reconstructed.