How further education funding actually works

A college is funded through several separate streams, not one budget. Each one pays for a different kind of learner, has its own rules, and is worked out in its own way.

The main ones are funding for 16 to 18 year olds, funding for adults, and funding for apprenticeships. There are others, and they do not behave the same way as each other.

• 16 to 19: funded by a national formula, based on the programme a student is on.
• Adult skills: an allocation you have to earn, and it is reconciled at the end.
• Apprenticeships: a price agreed with each employer, paid as the apprentice progresses.

The streams differ not just in rate but in mechanism, and that is what catches people out. A 16 to 19 allocation is built from data returned in an earlier year, so this year's delivery does not change this year's income. An adult skills allocation is a ceiling you earn against in year, and you can lose part of it by under-delivering. Apprenticeship income is not an allocation at all: it is the sum of what individual employers have agreed to spend.

Because the mechanisms differ, the management response to a problem differs too. Under-recruitment in 16 to 19 is next year's problem. Under-delivery in adult skills is this year's, and there is a point in the year after which it cannot be fixed.

Each stream is selected in the ILR by the funding model recorded on the learning delivery, which determines the calculation applied, the rules that govern eligibility and evidence, and the reports the record appears on.

16 to 19 provision runs under funding model 25 and is calculated through the national funding formula: student numbers by band, national funding rate, retention factor, programme cost weighting and the formula-protected uplifts. Adult Skills Fund provision runs under funding model 38, earning against an allocation with reconciliation. Tailored learning runs under funding model 11 outside the formula. Apprenticeships run under funding model 36 against negotiated prices capped at band maximums. Skills Bootcamps run under funding model 37 against milestones.

The college is not paid because it taught someone. It is paid because it recorded that it taught them, in the return called the ILR.

That sounds like a technicality. It is not. If the record is wrong, the money is wrong, the published results are wrong, and an auditor will find it.

The ILR is submitted every month and restates the whole funding year each time. Everything downstream reads it: the funding calculation, the funding reports finance reconciles to, the published achievement rates, and the population an auditor samples from.

This is why a data problem in further education is never only a data problem. A learner who left in October and was not withdrawn until June has generated eight months of funding that will be recovered, and has sat in the achievement rate cohort the whole time.

Each collection period submission is a full restatement of the in-year population, not a delta. Funding is recalculated across all periods on each submission, so a retrospective data change restates earlier period earnings.

The same submitted data is the input to the qualification achievement rate calculation at the final return, and is the population from which funding assurance sampling is drawn. One record therefore carries funding, performance and audit consequences simultaneously.

The funding year runs from 1 August to 31 July. Rules and rates are set for each year.

So the right answer to almost any funding question starts with a year. An answer without one is incomplete, and last year's answer is not automatically this year's.

Funding rules, rates and formula documents, the ILR specification and the validation rule set are all published per funding year. Codes that were valid last year may not be valid this year. Entitlements change. Devolved boundaries move.

This is also why an aim is governed by the rules in force when it started, not the rules in force when somebody asks about it. A learner who started in July is on last year's rules for that aim.

The learning start date fixes which funding year's ruleset applies to an aim, which is why cross-year cohorts sit under different rules within the same submission.

Every page in this Academy states the funding year it applies to and names the source version behind it. Where a comparison across years is useful, it is labelled as a comparison rather than presented as a single rule.

Funding touches nearly every part of a college, and the money is usually decided long before anyone from MIS sees the data.

Curriculum decides what is taught and for how many hours. Admissions and registry capture who the learner is. MIS turns all of that into the return. Finance reconciles what it earns. Quality watches what it means for results.

• Curriculum: decides the programme, the hours and the aims.
• Registry and admissions: capture identity, eligibility and enrolment.
• MIS: generates, validates and submits the return.
• Finance: reconciles earnings to the ledger and forecasts the year.
• Quality: watches retention, achievement and what the data says about them.

The single most useful thing to understand is that MIS does not create the data. It inherits it. By the time a problem is visible in a funding report, the decision that caused it was made weeks earlier by somebody who had no idea it was a funding decision.

That is why most durable improvement in funding accuracy comes from changing what happens at enrolment and in curriculum planning, not from working harder at return time.

The ILR is generated from the student record system, which is populated by admissions, enrolment, curriculum and registry processes. Corrections therefore belong in that system rather than in the generated file: a change made to the file is lost at the next generation and the source stays wrong.

Mapping each ILR field back to the operational process that populates it is the most effective diagnostic exercise a college can do. It converts a list of data errors into a list of process owners.