Who owns the training needs analysis, and who pays for it

I sat in a plan review last quarter where the training calendar had 23 line items. Every line had an owner in the L&D column. Not one had a name in the budget column. The meeting spent forty minutes arguing about which two programmes to cut and never once asked who was paying for any of them.

That is the whole problem in one table. A training need is not a finding. It is a commitment of somebody's money and somebody's operating time. If the analysis is produced by people who control neither, the output is a list of requests dressed as evidence.

The question is not who runs the survey

Most teams answer "who owns the TNA" by pointing at the person who designs the questionnaire. That is a clerical answer to a governance question. Designing the instrument is a skill. Deciding what gets funded is a right. They are different things, and confusing them is why the analysis keeps producing documents nobody acts on.

A recent systematic review of the field makes the pattern plain. Nunes, Ghedine and Llopis reviewed 101 studies published between 2009 and 2024 and found that most TNA research is conducted in a non-systematic way, focuses only on the individual level, and disregards alignment with organisational strategy. Individual-level analysis is exactly what an L&D team can do alone. Strategy alignment is exactly what it cannot do alone, because strategy lives with people who hold profit and loss.

The failure is not analytical laziness. It is that the analysis was scoped to the part of the problem the analyst had authority over.

Three rights, three holders

Stop asking who owns the analysis. Ask who holds each of three rights, and write the names down.

HR holds a fourth, quieter right: consistency. It is HR's job to make sure the same standard of evidence is applied in sales as in the plant, so that a need in one function cannot be waved through on a director's enthusiasm while an identical need elsewhere is refused.

The reason this split matters is that each right has a different failure mode when it lands in the wrong hands.

When L&D holds the right to define the problem, you get a wish list. L&D can describe a skill gap, but it cannot say what the gap costs, because it does not see the margin, the rework, the escalations or the customer churn. So the gap gets described in the only currency L&D has, which is course attendance.

When the business holds the right to define the method, you get anecdote. A plant head who has decided his supervisors need a communication programme will design a survey that finds exactly that. This is not cynicism, it is how confirmation works. Research on TNA in Saudi Arabian higher education found the primary obstacles to a working process were inadequate HR processes, insufficiently experienced HR directors, poor engagement, and favouritism. Favouritism is the polite word for an assessor with unchecked discretion over a decision that spends other people's money.

When nobody holds the right to release money, you get the meeting I sat in.

The method, in the order the work happens

Step one: the business writes the performance statement first

Before any learning language enters the room, the business owner writes one sentence with three slots filled.

"We are seeing [observable event] at a rate of [number] against a standard of [number]. It costs us [amount or unit] per month. I want it at [target] by [date]."

If the sentence cannot be completed with numbers, the need is not yet defined. It may still be real, but it goes on a watch list, not a plan. This single rule removes about a third of a typical request list, and it removes it without an argument about training philosophy.

Note what the business owner is not asked to do: diagnose. They are not asked whether the cause is skill, process, tooling or incentive. They are asked what is happening and what it costs. Diagnosis is the next step and it belongs to someone else.

Step two: L&D runs the diagnosis and is allowed to say "not training"

L&D takes the performance statement and tests the cause. Two independent sources per need, at least one of which is not a self-report. Observation, output data, error rates, assessment results, customer records. A survey of what people feel they are weak at is a starting point, never a conclusion.

The output of this step is a routing decision, and only some routes end in a course. In our own practice the four routes are: skill gap, so train; information gap, so fix the job aid or the process; environment gap, so fix the conditions; and motivation gap, so fix the consequences. If you want the fuller version of how a gap becomes a one-page plan, the method is set out in how to turn a competency gap analysis into a plan.

L&D needs explicit permission to route a request away from training. Without that permission, the function becomes an order taker, and order takers are judged on delivery, not on whether the order was right.

Step three: name one budget owner per line

One need, one budget holder. Not a shared cost centre, not "L&D budget, allocated". A named person whose unit absorbs the cost and whose number is supposed to move.

This is the test that does the most work, because it is mechanical. Take the current plan. Add a column. Fill in a name for every line. Every blank line is not a need. It is a preference with a course attached.

In most organisations this exercise removes more line items than any prioritisation workshop, and it does so in about twenty minutes.

Step four: L&D sets the evidence standard, and it applies to everyone

L&D's real authority is not over the money. It is over what counts as proof. Publish the standard: what evidence is required to move a need from watch list to plan, what sample size, what recency, what source types.

Then apply it without exception, including to the request from the most senior person in the room. The moment the standard bends for seniority, the function loses the only lever it has.

Step five: needs expire

A TNA is a snapshot with a date on it, not a standing entitlement. Put a review date on every need, twelve months at most, and re-sign it. A need that nobody re-signs is a need nobody owns.

This matters more in fast-moving roles than in stable ones. A capability gap identified for a process that has since been automated is not a gap, it is a memory.

The part most ownership models miss

Even a perfectly allocated need can fail after the money is spent, and the reason is usually not the course. It is the conditions the learner returns to.

The largest meta-analysis of training transfer, by Blume, Ford, Baldwin and Huang, pooled 22 studies covering 2,085 people and found a sample-weighted mean correlation of .23 between work environment and transfer. Within that, transfer climate correlated at .27, general support at .21, and supervisor support alone at .31, the strongest of the three. Support and climate are not things L&D controls. They are things the line controls.

Read that against the ownership table and the conclusion is uncomfortable. The person who funds the need is also the person who determines whether it converts. Which means ownership cannot stop at signing the cheque. The budget owner has to own the post-training conditions too: the practice opportunity, the follow-up conversation, the tolerance for a slower first month while someone uses a new method badly.

If a business leader is not willing to own those conditions, the honest answer is to not fund the programme. You would be buying a correlation you have no intention of honouring.

A worked example, with illustrative numbers

The figures below are illustrative, built to show the arithmetic rather than to report a real programme.

A distribution company with 1,400 staff runs its annual TNA and receives 31 requests. Under the old process, L&D scores them on a priority matrix and funds 18.

Under the split-rights process:

• 31 requests enter. The business owners write performance statements for 12. The other 19 have no numbers, so they go on the watch list.
• L&D diagnoses the 12. Four turn out to be information or process problems, and are routed to operations. Two are environment problems, mostly a scheduling conflict that makes the required behaviour impossible on the late shift.
• Six remain as genuine skill gaps with a named budget owner each.
• Two of those six are deferred by their owners because the unit cannot release people in the current quarter. That is a legitimate business decision, made by the person who bears the cost of it.

The plan that goes to the board has six line items instead of eighteen. It costs less. It also has a name against every rupee and a performance number against every name, which is why it survives the first review meeting.

The 19 watch-list items are not discarded. They are dated and revisited. Some will come back with numbers. Most will quietly disappear, which is what should have happened to them in the first place.

Where this breaks

Three honest limits.

First, the model assumes budget sits in business units. In a centralised L&D cost centre, where the function is given a pot and told to allocate it, there is no business budget owner to name, and step three collapses. In that structure the closest equivalent is a joint sign-off: a business sponsor and the L&D head both sign each line, and neither can add one alone. It is weaker than a real budget owner because the sponsor does not feel the cost, but it is better than nothing.

Second, it needs a business leader willing to write a number. In organisations where performance data is genuinely unavailable, not hidden but absent, step one cannot be completed honestly. Then say so and run a measurement build in parallel, rather than inventing a baseline.

Third, in firms under about 200 people the roles collapse. The person who owns the problem, the budget and often the analysis is the same person. The value there is not the split but the discipline: write the performance statement anyway, and write the routing decision down, so that a decision made from instinct is at least visible and revisable.

What to do this week

Open the current training plan. Add one column to the right of the last one. Title it "budget owner". Fill in a name for every line, using the person whose unit absorbs the cost, not the person who requested the course.

Leave the blanks blank. Count them. That number is the size of the problem, and it is the only slide you need for the next conversation with your CFO.

If you want the plan itself to carry the routing decision and the evidence standard in one place, that is what the training needs and assessment workflow in BolderBrain is built to hold. The method above works on paper too. The difference between the two arguments, and why a request is not a need, is worth reading before you start: training needs analysis versus training needs identification.