Reconciling funding month by month

Every return sends the whole year again, so a change made this month can move money reported months ago.

That is why earnings move in ways that look inexplicable until you know the mechanism.

The common causes are few: withdrawals recorded this month that remove earnings from earlier ones, records rejected at validation, monitoring codes changed, band changes from amended hours, and genuine delivery change.

Rule them out in that order. The first two explain most movements.

Compare period on period at total level, then by funding model, then by curriculum area, before going to learner level.

Compare the funded record count alongside the money: same count with different money is a rate or code issue, different count is a population issue.

Four steps, same order, every month.

• Download the funding reports for the period.
• Compare against the previous period at total level.
• Explain the movement in writing before anybody asks.
• Update the forecast with the explained position.

Writing the explanation down is the step that converts the routine from a task into a control. It also means the same question next month takes minutes rather than hours.

Where the movement cannot be explained, that is the finding. An unexplained movement is a data problem that has not been located yet.

Retain the funding reports per period alongside the submitted file and the validation output. That triple answers most retrospective questions without re-derivation.

Keep a standing movement log: period, movement, cause, action. It becomes the evidence of control as well as the working.

A reconciliation tells you where you are. A forecast tells you where you are heading, and it is only as good as the data behind it.

Forecasting from earnings to date plus planned delivery is straightforward arithmetic. What makes it useful or useless is whether the earnings to date are real.

A forecast built on a return containing learners who have left forecasts the error too, and it does so confidently. That is why the data quality checks and the forecast belong to the same routine rather than to different teams.

The useful discipline is to state the forecast with its assumptions: what has been delivered, what is planned, and what is known to be uncertain in the data. A forecast presented as a single number invites false confidence.

For streams that reconcile, the forecast has a decision attached: if delivery is tracking below allocation, what is being done, and by when does it have to start to make a difference.

Forecast per stream rather than in total, because the mechanisms differ: lagged streams do not respond to in-year delivery, and reconciled streams have a practical point after which action is ineffective.

Track forecast accuracy over the year. A forecast that moves substantially each month is telling you something about data quality, not only about delivery.