How to run a training needs analysis (step by step)

The short answer

To run a training needs analysis (TNA), you: (1) define the business decision the analysis must inform, (2) measure what the target group can currently do — with an assessment, not a self-rating survey, (3) analyse the gaps at group level and separate the ones training can close from the ones it can't, (4) design the programme from the ranked gaps, and (5) re-measure the same capabilities after training so improvement is a number rather than an impression.

The order matters. Most TNAs that fail skip step 1 (so the analysis answers no particular question) or soften step 2 into a survey (so it measures confidence instead of capability). The steps below cover what each stage looks like when it's done with measurement rather than opinion.

Step 1: Define the decision the TNA must inform

A TNA is not a census; it exists to inform a specific decision — what should next quarter's programme contain, is this team ready for the new tooling, which of three proposed workshops deserves the budget. Write that decision down first, because it dictates who you measure, what you measure, and what the report must be able to say.

This is also where you fix the audience: the specific cohort (a sales region, a graduating batch of team leads, a client's operations floor), the capability question in one or two sentences, and the constraint the answer must respect — budget, time-off-floor, or a programme date already in the calendar.

Step 2: Measure capability, not confidence

The standard TNA instrument is a self-rating survey — "rate your negotiation skills from 1 to 5" — and it is the weakest link in most analyses. Self-ratings measure self-perception, which is unreliable in both directions: novices overrate because they don't know what good looks like, and experts underrate because they do. A survey tells you what people think they need; a budget should rest on what they can demonstrably do.

Measurement means putting people in front of realistic tasks and scenarios for their actual work and scoring the judgement they show — the approach a situational judgement test takes. Historically the blocker was cost: a bespoke diagnostic took consultant weeks. AI composition removed that blocker — on AssessAll's TNA Studio, you describe the cohort, context and capability question in plain language and AI composes a bespoke instrument and analysis framework for that exact requirement, delivered by share link or QR with no participant accounts.

One design rule at this step: no pass marks. People who believe they're being graded give defensive answers, and defensive answers poison the data. A needs analysis should read as developmental — priorities and stages, not verdicts.

Step 3: Analyse gaps at group level — and separate trainable from not

Individual results tell each participant where to focus; the programme decision lives at group level. A useful group analysis ranks the cohort's gaps by consequence to the business objective (not just by average score), and surfaces sub-group patterns — the gap that only exists in one region, the capability the senior half has and the junior half lacks.

The most valuable sentence a TNA can produce is "this is not a training gap." Some low scores trace to process, staffing, tooling or environment problems that a workshop cannot fix, and a report that can't say so will happily let you buy training for a staffing problem. TNA Studio's group report makes this split explicitly — trainable gaps on one side, everything else routed to the ops conversation where it belongs.

Step 4: Design the programme from the ranked gaps

With ranked, trainable gaps in hand, programme design becomes selection rather than invention: the curriculum covers the gaps that are high-consequence and trainable, skips what the cohort already has, and sets entry points per sub-group where the data shows different starting lines.

This is also the step where the analysis pays for itself commercially. For an internal L&D team, it turns the budget conversation from "we propose this workshop" into "the data says these three gaps, in this order." For training providers and consultancies, a measured baseline replaces the generic proposal — the engagement is scoped on the client's evidence, which is considerably harder to say no to.

Step 5: Re-measure after training to prove the gain

A TNA that ends when the programme starts proves nothing about the programme. The full loop re-measures the same competencies after training — and ideally mid-programme — so improvement is a measured delta on the baseline rather than a feedback-form impression. On AssessAll, Learning Journeys sequence exactly this: baseline, during-programme and outcome waves, with an Impact & ROI report showing which capabilities moved, by how much, and for whom.

Because the measurement is metered per participant rather than licensed per seat, the re-measure wave costs a defined, budgetable amount — which is what makes the proof-of-gain step affordable enough to actually run.

What this looks like in practice

A concrete example: a consultancy is asked to build a train-the-trainer programme for a client's 40 internal trainers. Step 1: the decision is what the six-week programme should contain. Step 2: all 40 take a trainer-readiness diagnostic covering needs-diagnosis, explanation craft, facilitation and practice design (AssessAll ships this as a ready instrument — the Technical Trainer Readiness TNA, 18 credits per participant). Step 3: the group report shows explanation craft strong, practice design weak, and a facilitation gap confined to the newest cohort. Step 4: the programme weights practice design, adds a facilitation stream for the new cohort only, and drops the planned explanation module. Step 5: an outcome wave re-measures the same four areas and the delta goes in the sponsor's report.

Total measurement cost for that engagement: two waves × 40 participants at 18 credits each — visible up front, and small against the programme it de-risks. A new organisation's 250 free credits cover a pilot baseline before paying anything.