Your Training Needs Analysis Is Probably Asking the Wrong Question

There is a ritual still taking place in organisations everywhere.

Usually towards the end of the year, someone in HR sends out a Training Needs Analysis. An Excel sheet appears. Managers are asked what training their team needs next year. And within a few weeks, the organisation discovers that everyone apparently needs leadership, communication, presentation skills, AI, strategic thinking and, naturally, Excel. Again.

Then L&D consolidates the answers, produces a colourful training calendar, and spends the following year delivering precisely what everybody asked for.

There is only one small problem. That is not a Training Needs Analysis. It is a training wish list, and the two should never be confused.

The Chartered Institute of Personnel and Development makes the same point rather more politely than I do: learning needs analysis should be a systematic and ongoing process connected to performance gaps, not a rigid annual box-ticking exercise.

The distinction matters, because organisations do not have unlimited training budgets, unlimited employee attention or unlimited time. Every day somebody spends sitting in a programme that solves the wrong problem is a day the organisation has paid twice: once for the training, and once for the lost productivity.

Why doesn’t asking people what training they want work?

Because it asks the wrong question, of the wrong people, at the wrong moment.

Imagine going to your doctor with stomach pain and being asked which medicine you would like. You would find another doctor. Yet this is effectively what we do when we hand employees a course catalogue and ask them to tick.

The problem is not that people are being dishonest. It is that people cannot reliably diagnose their own capability gaps. A metasynthesis of 22 meta-analyses found only a moderate relationship between how people assess their own abilities and objective measures of their performance. Self-evaluation got more accurate when the skill was specific, familiar and measurable, which is precisely what a generic training survey is not.

There is also a structural problem underneath it. The people most affected by a capability gap are frequently the least able to name it. Someone who has never been taught to run a difficult conversation does not experience that as a training need. They experience it as a colleague who is hard to work with.

So employee perception is useful evidence. It is one piece of evidence. It is not the diagnosis.

Where should a training needs analysis actually start?

With a business problem somebody is already worried about.

Not “what programmes should we run,” but “what is stopping this part of the business from performing better.” Questions like these:

• Why are customers abandoning the process at this particular touchpoint?
• Why does this department have twice the turnover of the one next to it?
• Why are managers escalating decisions that should be made two levels lower?
• Why do projects keep missing deadlines at the handover stage?
• Why are new managers technically brilliant and struggling to lead former peers?
• Why are complaints rising when everyone has completed customer service training?
• Why are people reluctant to challenge senior leaders in an organisation that announces it has a speak-up culture?

Now we are somewhere interesting, because each of those contains three things: a situation, a population, and an observable consequence. That is the beginning of a diagnosis.

Most of the data you need already exists inside the organisation. Customer complaints. Exit interviews. Sales conversion rates. Safety incidents. Quality defects. Project delays. Help-desk tickets. Mystery-shopper reports. Audit findings. And that wonderfully revealing organisational database known as what everybody complains about at lunch. The last one is less scientific. It is occasionally more accurate.

Start from the problem and the population becomes obvious. Start from the course catalogue and you will train whoever volunteers.

Is training even the answer?

This is mildly inconvenient coming from someone who has spent nearly three decades in this business, but here goes. Sometimes the answer should be: don’t train them.

Suppose employees are not making decisions quickly enough, and HR concludes there is a decision-making training need. Perhaps. But suppose they already know perfectly well how to make the decision, and simply know that if they get it wrong their boss will take it apart in public. That is not a decision-making skills problem. That is a psychological safety and leadership problem, and no workshop will touch it.

Or people keep bypassing a process. Training problem? Maybe. Or the process requires fourteen approvals, three forms, two signatures and the patience of a Buddhist monk. That is a process design problem.

One question saves an enormous amount of training expenditure:

If these people already knew exactly how to do this, would they be doing it?

If the answer is no, be careful before prescribing training. Something else is blocking the behaviour, and it is usually one of these: incentives, leadership behaviour, unclear accountability, workload, inadequate tools, lack of authority, conflicting KPIs, poor process, absence of feedback, or a culture that quietly rewards something different from what the company says it values.

A needs analysis that cannot tell a skill gap from a system problem is not doing its job. It is generating demand for training.

And that may be the biggest flaw in the traditional approach. It assumes training before it diagnoses performance.

What questions produce a real diagnosis?

Here is a working set. It is deliberately short, because the value is in the follow-up rather than the coverage.

About the problem

1. What is happening now that should not be, or not happening that should?
2. How do you know? What are you actually seeing or measuring?
3. When did it start, and what changed around then?
4. Who specifically is affected? Not the department. The roles.

About the cause

5. If these people already knew how to do this well, would they be doing it?
6. What happens to someone who does it properly today? What happens to someone who does not?
7. Has anyone been trained on this before? What happened afterwards?
8. What does the best performer in this group do differently?

About the outcome

9. What would you see six months from now if this worked?
10. Who would notice first, and what would they notice?
11. What would you be willing to stop doing to make room for this?

Question 5 disqualifies more projects than any other, and that is a feature. Question 7 tells you whether the last attempt failed because of the content or because nothing reinforced it. Question 8 is the cheapest piece of consulting available to any organisation: find someone unusually good at the job and watch what they actually do, what they notice, what they say when it gets uncomfortable, what judgement they have developed that is nowhere in the official process document. Question 11 tells you whether the organisation is serious.

How do you know the gap is real rather than assumed?

Measure before you train.

One of the strangest things about corporate learning is how often organisations spend money improving something they never measured in the first place. Then six months later somebody asks whether the training worked, and everyone looks at the evaluation form. Four point seven out of five. Excellent lunch. Trainer was engaging. Room slightly cold.

None of that tells us whether anything changed.

Before designing anything, establish a baseline. What behaviour is happening now? How often? At what standard? With what business consequence? Then measure the same thing afterwards. Without a baseline you end up defending the investment with attendance and satisfaction scores, which is why so many training functions struggle to make their case to a finance director.

This is also where a proper assessment earns its place at the front of a programme rather than as a standalone exercise.

What should a finished analysis actually say?

For every proposed intervention, you should be able to complete these six statements:

These people currently [behaviour today]. The organisation needs them to [required performance]. This matters because [business consequence]. The evidence is [data or observation]. We believe the primary cause is [capability, system, or both]. Success would look like [observable change].

If you cannot complete those six, you do not have a training need yet. You have a topic.

And organisations already have plenty of topics. Leadership. Culture. Innovation. Agility. Digitalisation. Customer centricity. They look magnificent on a slide. None of them tells me what somebody needs to do differently at 10.15 on Monday morning.

Anything you cannot complete because the cause turned out to be a system problem should still be written down and sent to whoever owns that system. It is frequently the most valuable output of the whole exercise, and it is the part that usually gets quietly dropped because it is not a training deliverable.

Why does this matter more now than it used to?

Because the ground is moving faster than the annual cycle can track.

The World Economic Forum’s Future of Jobs Report 2025 found that employers expect 39% of workers’ core skills to change by 2030. AI and big data, cybersecurity and technological literacy are rising fastest, alongside very human capabilities including creative thinking, resilience and leadership.

LinkedIn’s 2025 Workplace Learning Report found 49% of learning and talent development professionals agreeing that their executives are concerned employees do not have the right skills to execute the business strategy. Nearly half the profession is looking at a skills crisis.

And Deloitte’s 2025 Global Human Capital Trends found that 66% of managers and executives say most recent hires were not fully prepared, with experience the most common failing. Deloitte’s conclusion is worth sitting with: the hardest gulf to close is not the skills gap, it is the experience gap.

That distinction is powerful, and it has a direct consequence for how you read your own analysis. You cannot put someone through a slide deck and give them experience. Sometimes what people need is practice, exposure, a stretch assignment, coaching, simulation and feedback. Training may be part of the answer. The classroom is rarely the whole answer.

So the answer cannot be to send out the form once a year. By the time you consolidate the spreadsheet, approve the budget, appoint the provider and find three dates when everyone is available, the business has developed a different problem.

What replaces it is not a prettier questionnaire. It is triangulation: employee perception, manager observation, performance data, customer evidence, direct observation and behavioural assessment. If five signals point at the same issue, you have something worth investigating. If employees say communication is excellent and customer complaints say otherwise, you have found something more valuable than a survey score. You have found a perception gap.

What happens after the analysis?

The analysis is only worth the effort if the design that follows respects it, and this is where most programmes lose the thread.

Decades of research into training transfer make one point very clear: what happens around the training matters enormously. Reviews of the transfer literature consistently identify realistic practice, behavioural modelling, managerial support, genuine opportunity to apply the learning, and follow-up as the factors that decide whether anything survives contact with Monday. The work environment does more to determine transfer than the quality of the session does.

So during the analysis, do not only ask what people need to learn. Ask what has to happen back at work for the behaviour to survive. That question changes the design.

In our own professional development work this is the difference between a programme that reports attendance and one that reports change. One frontline service programme reached 6,597 personnel across 74 batches, built as a reinforcement loop combining capability building, real-world validation, leadership coaching and internal trainer development rather than a single workshop. Another ran five two-day tracks with teachbacks and workplace application planning, and recorded 95 percent workplace action-plan completion. In both cases the design started from a defined operational problem rather than a course title. You can see more of how that plays out in practice across the engagements we publish.

Worth knowing if budget is the obstacle: a properly diagnosed programme is no harder to fund than a badly diagnosed one, and most of this work is HRD Corp claimable in the same way any other registered programme is. The same principle applies there too, which is that what qualifies for a claim comes down to how the programme is designed rather than how it is described.

The short version

A modern training needs analysis does not begin with a course catalogue. It begins with performance.

Do not ask only what employees want to learn. Find out where the work is breaking. Do not rely on one source. Triangulate. Do not assume every problem is a skill problem. Investigate the system. Do not wait until December. Sense capability gaps continuously. Do not measure whether people enjoyed the training. Measure whether anything changed.

The question a CHRO should be asking L&D is not how many people we trained, or how many programmes we ran, or what the satisfaction score was. It is this: what business problems did we diagnose, what behaviours did we change, and what evidence tells us they changed?

That is a much tougher question, which is precisely why it is a better one.

At its best, L&D is a performance function. And occasionally its most valuable contribution is telling the organisation that this is not a training problem at all. That is not L&D talking itself out of work. That is L&D finally behaving like a business partner.

Diagnose first. Prescribe second. Doctors worked that one out a long time ago.

Frequently asked questions

What is a training needs analysis?

A structured process for identifying the gap between current and required workplace performance, and determining whether learning, organisational change, or a combination of the two is required. It is a diagnosis, not a course selection exercise.

Is a training needs analysis the same as a training survey?

No. A survey can contribute data to a needs analysis, but it should not be the whole diagnosis. A strong analysis combines employee and manager perspectives with performance data, direct observation and customer evidence.

How often should an organisation run one?

Rather than relying on a single annual exercise, monitor capability continuously and run a deeper analysis when strategy, technology, roles or performance requirements change. CIPD describes learning needs analysis as an ongoing process linked to organisational performance, not an annual event.

How do you know whether training is actually the solution?

Ask whether people would perform correctly if they already had the knowledge and skill. If the answer is still no, because incentives, management behaviour, process, workload or authority prevent it, training alone will not solve it.

Can AI improve training needs analysis?

Potentially. AI can help analyse skills data, feedback and job requirements at a scale humans cannot. But it does not remove the need for diagnosis, and it will analyse enormous quantities of bad data much faster than anyone could before. Technology should improve the diagnosis, not merely automate the old questionnaire.

References

• Blume, B. D., Ford, J. K., Baldwin, T. T. and Huang, J. L. (2010). Transfer of Training: A Meta-Analytic Review. Journal of Management, 36(4), 1065 to 1105.
• Chartered Institute of Personnel and Development. Learning Needs Analysis.
• Deloitte (2025). Global Human Capital Trends: Closing the Experience Gap.
• Grossman, R. and Salas, E. (2011). The Transfer of Training: What Really Matters. International Journal of Training and Development, 15(2), 103 to 120.
• LinkedIn Learning (2025). Workplace Learning Report 2025.
• Salas, E., Tannenbaum, S. I., Kraiger, K. and Smith-Jentsch, K. A. (2012). The Science of Training and Development in Organizations: What Matters in Practice. Psychological Science in the Public Interest, 13(2), 74 to 101.
• World Economic Forum (2025). The Future of Jobs Report 2025.
• Zell, E. and Krizan, Z. (2014). Do People Have Insight Into Their Abilities? A Metasynthesis. Perspectives on Psychological Science, 9(2), 111 to 125.