What Is a Training Needs Assessment?

A training needs assessment is a structured comparison of the knowledge, skills, and abilities a job requires against those the people doing it currently have, carried out at organization, task, and person level. It produces a ranked list of gaps, states which gaps training can close, and gives each one an owner, a method, and a measure.

On June 24, 2025 the National Transportation Safety Board adopted its final report on Alaska Airlines Flight 1282, the Boeing 737-9 whose left mid-exit door plug separated at 14,830 feet on January 5, 2024. The probable cause was Boeing’s failure to provide adequate training, guidance, and oversight so that manufacturing personnel could consistently follow its parts removal process.

Four bolts were removed on September 18, 2023 and never reinstalled; the staff who closed it were not the specialists the process assumed, and no record was raised. Among 19 new recommendations the Board asked Boeing for a structured on-the-job training program defining the tasks a qualified worker must master. That is a training needs assessment, ordered afterwards.

Boeing’s case is unusual only in scale. The same missing step shows up in OSHA citations, in compliance programs that train everyone on everything, and in risk teams that hire for a skill nobody measured. The method below applies equally to a factory floor, a compliance team, and the risk analysts a risk function is trying to develop.

Training Needs Assessment: Key Takeaways
A training needs assessment measures the gap between what a job requires and what the people doing it can currently do, at three levels: organization, task, and person. Its output is a gap register with owners, priorities, and evidence, and a decision on which gaps training will close and which it will not.
The NTSB’s final report of June 24, 2025 on Alaska Airlines Flight 1282 named Boeing’s failure to provide adequate training, guidance, and oversight to manufacturing staff as the probable cause of the door plug separation. Four bolts were never reinstalled; the assessment that should have caught the on-the-job training gap did not exist.
Five of OSHA’s ten most-cited standards in FY2025 carry a training clause: hazard communication (3,010 citations), lockout/tagout (2,562), respiratory protection (2,294), fall protection training (2,216), and powered industrial trucks (2,150).
ATD’s 2025 State of the Industry puts direct learning spend at $1,054 per employee, formal learning hours at 13.7 (down from 17.4), and cost per learning hour at $165 (up 34%). Fewer, dearer hours make targeting the right gaps the main lever.
The World Economic Forum estimates 59 of every 100 workers will need training by 2030, and North American employers put their own figure at 67%. 11 of the 59 are unlikely to receive it.
Four data sources feed the assessment: records and performance data, observation and work sampling, surveys and self-ratings, and tests or certifications. Each has a known bias; use at least two per gap.
Not every gap is a training gap. Missing tools, unclear standards, and workload cause performance problems that training cannot fix, and the assessment must say so before budget is committed.

 

What a Training Needs Assessment Is and Is Not

The term dates to McGehee and Thayer’s 1961 book Training in Business and Industry, which set the three levels still used today: organization, operations (task), and person. ISO 10015:2019, the guideline for competence management and people development, frames the same work as determining competence needs, comparing them with current competence, and planning to close the difference.

Activity	Question it answers	Output	How it differs from a training needs assessment
Training needs assessment	Which required capabilities are missing, at which level, and will training close them?	Gap register with priorities, owners, and non-training causes flagged	The reference activity; the others feed or follow it
Skills gap analysis	Which skills does the workforce lack against a future plan?	Skills inventory and heat map	Usually organization level only; rarely tests whether training is the fix
Job or task analysis	What does this role actually do, to what standard?	Task inventory with frequency, criticality, and difficulty	Describes the requirement; the assessment measures people against it
Performance review	How did this person perform against objectives?	Rating and development plan	One person, one period; causes beyond skill are usually not separated
Competency framework	What behaviors and knowledge define each level of each role?	Competency library	The yardstick; useless without the measurement step

The distinction that matters most in practice is the last column. A review that reports low performance has not yet shown a training need. The CIPD learning needs factsheet makes the same point: the analysis must separate gaps in capability from gaps in resources, process, or motivation, because training only fixes the first.

For a risk or compliance function the assessment is also a control. The DOJ’s Evaluation of Corporate Compliance Programs asks whether training is tailored to the risks employees face and whether the company measures its effect.

A documented training needs assessment is the evidence that answers both questions, and it belongs in the control measure library.

Why Skipping the Assessment Costs Money: The 2025 Evidence

Regulators count training failures, and the counts are large. OSHA’s ten most frequently cited standards for fiscal year 2025 include five with an explicit training requirement, and together they drew 12,232 citations between October 1, 2024 and September 30, 2025, per the final data published by Safety+Health.

Figure 1. Hazard communication, lockout/tagout, respiratory protection, fall protection training, and powered industrial trucks each require documented training; together they drew 12,232 citations in FY2025.

Each citation is an unassessed training need with a dollar figure. On December 5, 2024 OSHA proposed $288,299 in penalties against Mitchell Industrial Tire Co. after a forklift operator at its Elm Mott, Texas distribution center was killed by a falling pallet; the citations included failing to train forklift drivers as 29 CFR 1910.178(l) requires.

Spending now rewards targeting. ATD’s 2025 State of the Industry reports formal learning hours per employee fell to 13.7 in 2024 from 17.4, cost per learning hour rose 34% to $165, and direct spend held at $1,054 per employee, as Reworked summarised. Fewer hours at a higher price leave no room for the wrong gap.

Figure 2. ATD’s 539-organization sample used 13.7 formal learning hours per employee in 2024 at $165 an hour, against 17.4 hours at $123 a year earlier.

Source and date	Finding	Sample	What it means for the assessment
NTSB AIR-25-04, June 24, 2025	Inadequate training, guidance, and oversight named as probable cause	One accident, 19 new recommendations	Task-level assessment of on-the-job training is a safety control
OSHA FY2025 top ten, April 2026 final data	12,232 citations across five training-bearing standards	All federal OSHA inspections	Compliance training gaps are the most common finding in the country
ATD State of the Industry, August 2025	13.7 hours per employee at $165 an hour	539 organizations	Every hour has to close a ranked gap
Skillsoft IT Skills and Salary, November 2024	65% of IT decision-makers report team skill gaps; 72% plan to train existing staff	5,100+ IT leaders and professionals	Person-level assessment decides who gets the 72%
TalentLMS 2026 L&D Report	42% of US HR managers report skills gaps; 70% of employees multitask during training	1,000 employees, 101 HR managers	Relevance, not volume, is the constraint
WEF Future of Jobs, January 2025	59 of 100 workers need training by 2030; North America 67%	1,000+ employers	Organization-level assessment sets the five-year scope

We treat the OSHA list as a ready-made organization-level assessment for any US operation with a warehouse, a plant, or a construction site. If the operational risk register lists forklifts, chemicals, or work at height, the five standards in Figure 1 define five task-level assessments before any survey is sent.

The Three Levels of a Training Needs Assessment

McGehee and Thayer’s three levels are still the spine because each answers a different question with different data. Skip the organization level and you train for a strategy nobody set; skip the task level and you train to opinions; skip the person level and everyone sits the same course. OPM’s planning and evaluating guidance uses the same three.

Training needs assessment level	Question	Data sources	Output
Organization	Where does strategy, regulation, or risk require capability the workforce lacks?	Strategic plan, risk register, regulatory obligations, incident and audit findings, turnover and vacancy data	Priority capability areas with a business owner and a deadline
Task	For each critical role, what must be done, how often, to what standard, and what happens if it is done wrong?	Job analysis, procedures, equipment manuals, regulatory standards such as 29 CFR 1910.178(l), error and near-miss records	Task inventory rated for frequency, criticality, and difficulty; required knowledge, skill, and ability per task
Person	Which individuals or teams fall short of the task standard, and by how much?	Observation, practical tests, certifications, performance data, self-rating cross-checked by supervisor	Named gaps per person, with cause classified as skill, knowledge, resource, or motivation

The organization level holds the five-year numbers. The World Economic Forum’s Future of Jobs Report 2025 estimates that 59 of every 100 workers will need training by 2030: 29 upskilled in their current role, 19 upskilled and redeployed, and 11 who need it but are unlikely to receive it. Companies headquartered in North America put their figure at 67%.

Figure 3. Employers expect 59 of 100 workers to need training by 2030 and admit 11 of them will not get it; North American employers put their own figure at 67. The training needs assessment decides which 11.

The task level is where safety and compliance obligations become measurable. A compliance requirement such as hazard communication training under 29 CFR 1910.1200 already states who must be trained, on what, and when; the task analysis converts that into a checklist an observer can score. OSHA’s Training Requirements in OSHA Standards compiles every such clause in one document.

Most assessments go wrong at the person level, because self-rating is the cheapest method and the least reliable. In the TalentLMS 2026 report, 84% of US employees said they were satisfied with their training while 70% admitted multitasking through it. Satisfaction is not competence, and only observation or a test tells you which.

How to Run the Assessment in Six Steps

Most published versions of the process stop at implementation. The six steps below add the two that decide whether the effort was worth it: classifying the cause of each gap, and setting the measure before the training starts. ISO 10015 clause 6 describes the same sequence as determine, plan, provide, and evaluate.

#	Step	What you do	Evidence produced
1	Set the scope and the trigger	Name the business driver: new regulation, incident, system change, strategy, or turnover. Agree which roles and sites are in scope and who owns the result.	One-page charter signed by the sponsor
2	Build the task inventory	For each in-scope role list the tasks, rate each for frequency, criticality, and difficulty, and attach the standard (procedure, regulation, or specification).	Task inventory with ratings and standards
3	Measure current capability	Collect data from at least two of the four sources per critical task: records, observation, survey, test.	Scored capability per person or team per task
4	Classify each gap by cause	For every shortfall ask whether the person has done it correctly before, has the tools and time, knows the standard, and is rewarded for doing it. Only knowledge and skill gaps go to training.	Gap register with cause codes
5	Prioritize and choose the method	Rank by criticality times gap size; match method to gap (on-the-job for procedural skill, instruction for knowledge, coaching for judgment).	Ranked plan with method, owner, cost, and date
6	Set the measure and the review date	Define the Kirkpatrick level 3 behavior or level 4 result that will show the gap closed, and when it will be checked.	Evaluation plan filed with the gap register

Step 4 saves the most money, because it removes entries that training cannot fix before anyone prices a course. Four questions sort a shortfall in a few minutes, and the answers decide whether the entry goes to the training plan or back to the manager as a resource, standard, or consequence problem:

• Has the person performed the task correctly before? If yes, the gap is not knowledge or skill; look at tools, time, feedback, or consequences.
• Does the person have the equipment, access, and time the task needs? A missing tool is a resource gap and no course closes it.
• Is the standard written, current, and known? An unclear standard is a management gap; training people to an unwritten standard produces variation, which is what the NTSB found at Boeing.
• Does doing the task correctly carry a consequence the person notices? If nobody checks the removal record, the record stops being kept.

The worked register below shows six entries from a distribution operation after step 4. Two of the six are not training gaps, a normal ratio in our experience. Each entry cites the same standard the enterprise risk register cites, so the HR key risk indicators and the training plan speak the same language at the risk committee.

Task and standard	Level	Gap measured	Cause code	Action, owner, measure
Pre-shift forklift inspection, 29 CFR 1910.178(q)(7)	Person	Observation: 4 of 11 operators skipped brake and mast checks	Skill	Structured on-the-job training with checklist sign-off; warehouse manager; 100% observed compliance at 30 days
Rack loading limits and pallet banding, site procedure WH-07	Task	Records: procedure last revised 2019, limits not posted	Standard unclear	Rewrite and post limits; operations director; no training until standard is current
Hazard communication for new degreaser, 29 CFR 1910.1200(h)	Person	Test: 9 of 14 handlers could not locate the SDS or name first-aid steps	Knowledge	30-minute instruction plus SDS station; EHS lead; test score 100% before first use
Lockout/tagout on conveyor 3, 29 CFR 1910.147(c)(7)	Person	Records: 2 of 6 authorized employees past annual retraining date	Knowledge (lapsed)	Retraining and periodic inspection; maintenance supervisor; zero overdue authorizations
Incident reporting within one shift, policy IR-02	Organization	Records: median report time 3.4 days; 40% of near misses unreported	Consequence	Supervisor follow-up and recognition, not a course; site lead; median under 8 hours in 90 days
Sanctions screening exceptions, compliance procedure CP-11	Person	Records: 6 false negatives in Q2 review; 2 of 3 analysts new since March	Skill	Case-based coaching with the compliance officer; head of compliance; zero repeat findings at next quarterly review

The register depends on two design choices. Every task cites its standard, so a new supervisor can read the requirement without asking. And every action names a measure that will be observed, which is how the register later answers the question the three lines model puts to internal audit: did the control work?

Choosing the Method: Four Data Sources Compared

No single data source is reliable enough to rank a gap on its own. Records show outcomes without causes, observation is expensive, surveys measure confidence, and tests measure recall. The table sets out where each is strong, what it costs, and the bias to correct for; we recommend two sources per critical task and three where the task is safety-critical.

Training needs assessment data source	Best at	Typical cost and time	Known bias	Correct for it by
Records and performance data (incidents, errors, audits, LMS, HRIS)	Organization and task level; finding where problems cluster	Low; days if data is already collected	Shows the symptom, rarely the cause; under-reporting hides gaps	Pairing with observation on the top three clusters
Observation and work sampling	Person level for procedural and safety tasks	High; 30 to 90 minutes per task per person	Observer effect; people perform better when watched	Unannounced sampling; checklist scored against the written standard
Surveys, self-ratings, interviews, focus groups	Wide coverage; surfacing perceived needs and barriers	Low to medium; two to three weeks including analysis	Confidence, not competence; 84% satisfied while 70% multitask	Supervisor rating of the same items; never used alone for safety tasks
Tests, simulations, and certifications	Knowledge and regulated competencies	Medium; test design is the cost	Measures recall under exam conditions, not habit on the job	Practical component; 30-day follow-up observation

Surveys still have a place, and the 2025 LinkedIn Workplace Learning Report shows why: only 15% of employees said a manager had helped them build a career plan in the previous six months, and 50% named lack of manager support as the barrier to development. A survey finds that barrier; no observation does.

Figure 4. Eighty-four percent of US employees are satisfied with training and 70% multitask through it. A self-rating survey captures the first number; only observation reveals the second.

For technical and security roles the task inventory does not have to be built from scratch. The NIST NICE Framework lists tasks, knowledge, and skills for cybersecurity work roles, which turns a cybersecurity risk management team’s assessment into a mapping exercise, and the entry-level analyst interview questions we published double as a knowledge test.

Certifications answer a narrower question than managers assume. A risk management certification proves knowledge at exam date; it says nothing about whether the holder applies it to your register. Use certification as the knowledge test and observation as the skill test, and record both against the task, which is the approach our GRC certification salary analysis recommends.

From Findings to Plan: Prioritizing and Proving It Worked

A gap register with forty entries and one budget needs a ranking rule. We rank on two axes: how critical the task is (regulatory, safety, or revenue consequence of failure) and how large the measured gap is (share of people below standard times the distance below it). The result is a four-cell matrix that also sets the method.

Priority cell	Definition	Response and method
Critical task, large gap	Regulatory or safety task where more than a quarter of performers fall below standard	Stop-work or supervision until closed; structured on-the-job training with sign-off; measured within 30 days
Critical task, small gap	Regulatory or safety task with a few named individuals below standard	Targeted coaching or retraining for the named people only; certification where the rule requires it
Routine task, large gap	Productivity or quality task with widespread shortfall	Check the standard and tools first (cause code); then group instruction or e-learning with a practical check
Routine task, small gap	Isolated shortfall on a low-consequence task	Manager feedback and job aids; no formal training spend

Proof follows the Kirkpatrick model; set the level before training starts. Level 1 reaction and level 2 learning cost little and prove little. Level 3 behavior, observed on the job at 30 and 90 days, is the minimum for any critical-task entry, and level 4 results tie the entry back to the key risk indicator that triggered it.

Three measures belong on every plan, and most HR and safety systems already collect the underlying data. Report them quarterly beside the other key risk indicators so the committee sees training as a control with a trend, and apply the same amber and red thresholds it uses elsewhere:

• Gap closure rate: share of register entries whose level 3 measure was met by the review date, reported quarterly to the risk committee alongside other key risk indicators.
• Overdue regulated training: count of people past a retraining date on any standard with a fixed cycle, such as the annual lockout/tagout inspection or three-year forklift evaluation, with a zero tolerance.
• Cause mix: share of gaps coded resource, standard, or consequence rather than skill or knowledge; a rising share means the register is catching management problems that training would have hidden.

The evaluation date is the difference between an assessment and a wish list. The NTSB’s recommendation to Boeing asked for training results to be documented and archived so they could be analyzed later; that is level 3 evidence retained for the next assessment, and the same discipline our KPI dashboard guide applies to any control.

Common Training Needs Assessment Questions Practitioners Ask

What is a training needs assessment in simple terms?

A training needs assessment compares what a job requires with what the people doing it can currently do, then lists the differences, ranks them, and decides which ones training will fix. It works at three levels, organization, task, and person, and its output is a gap register with an owner, a method, and a review date for each entry.

What are the three levels of a training needs assessment?

Organization, task, and person, from McGehee and Thayer’s 1961 model. Organization analysis finds where strategy, regulation, or risk demands capability; task analysis defines what each critical role must do and to what standard; person analysis measures named individuals against that standard. All three are needed, because skipping one produces training aimed at the wrong target.

How is a training needs assessment different from a training needs analysis?

In practice the terms are interchangeable; CIPD and ATD accept either. Where a distinction is drawn, assessment is the whole process from scope to plan, and analysis is the data-gathering and gap-measurement step inside it. What matters is that the work classifies each gap’s cause, because a training needs analysis that assumes every gap is a training gap wastes budget.

How often should a training needs assessment be repeated?

Run a full assessment every 12 to 24 months and a targeted one on every trigger: a new regulation, a serious incident or audit finding, a new system or product, a reorganization, or turnover above threshold in a critical role. Regulated tasks with fixed cycles, such as the annual lockout/tagout inspection or three-year forklift evaluation, set their own dates.

Who should conduct a training needs assessment?

The line manager owns the result because they own the performance; learning and development or HR designs the method and runs the data collection; subject-matter experts write the task standards; and the risk or compliance function confirms the regulatory tasks are in scope. For safety-critical work, an observer independent of the team should score the observations.

What methods are used in a training needs assessment?

Four data sources: records and performance data, observation and work sampling, surveys and interviews, and tests or certifications. Each has a known bias, so use at least two per critical task. Records find where problems cluster, observation measures skill, surveys surface barriers and perceptions, and tests measure knowledge under exam conditions.

Can a training needs assessment show that training is not the answer?

Yes, and a good one does so often. If the person has performed the task correctly before, lacks the tools or time, works to an unwritten standard, or faces no consequence for doing it wrong, the gap is a management or resource problem. In the worked register above, two of six entries were not training gaps, a normal ratio.

Lessons from Assessments That Failed

Training needs assessment failure	How it showed up	Fix
Survey used as the only source	84% satisfaction reported while observed compliance was under 70%	Observation or test on every critical task; survey for barriers only
Every gap treated as a training gap	Courses bought for problems caused by missing tools or unclear standards	Cause code on every entry; resource and standard gaps go to the manager
Task standard never written	Trainers taught their own method; the NTSB found the same variation in Boeing’s removal process	Task inventory cites a written standard before any training is designed
Regulated cycles tracked in heads	Two of six authorized employees past retraining; citation risk under 1910.147	Overdue regulated training as a zero-tolerance indicator
Assessment done once at launch	Register five years old; new systems and rules unassessed	Full refresh every 12 to 24 months plus trigger-based reviews
No measure set before training	Completion reported at 100%; behavior on the job unchanged	Kirkpatrick level 3 measure and review date filed with each entry
Person level skipped	Whole department sent on the same course; 70% multitask through it	Named gaps per person; only people below standard attend
Owner is HR alone	Line managers ignore results they did not sign	Sponsor from the line signs the charter and the register

Where Training Needs Are Heading by 2030

AI skills will dominate the organization level through 2030. In ATD’s 2025 sample 55% of organizations trained staff on AI skills, and in Skillsoft’s IT Skills and Salary survey AI and machine learning ranked lowest of nearly 30 competencies while topping the investment list. A large gap on a critical task belongs in the first priority cell.

Expect regulators to ask for the assessment itself. The NTSB’s recommendations to Boeing specify a structured on-the-job training program with documented, archived results; OSHA’s training-bearing standards already require written certification of competence; and the DOJ asks whether compliance training is tailored to risk. A register with task standards, cause codes, and level 3 evidence answers all three.

The person level will get harder as workload rises. TalentLMS found 65% of US employees faced higher performance expectations this year and 53% said workload prevented adequate training. Gallup’s State of the Global Workplace series ties manager engagement to that pressure, which is why the risk culture a team works in belongs in the organization-level scan.

Internal mobility will supply more of the answer than external hiring. The WEF’s 19 workers per 100 who will be upskilled and redeployed need a task inventory for the destination role, which our GRC analyst versus risk analyst comparison sketches for two common moves, and the employee KPI examples article supplies the person-level measures.

If you train to a schedule rather than to a measured gap, we run the three-level assessment against your risk register and regulatory obligations, build the task inventory and gap register with cause codes, and hand over the evaluation plan the risk committee expects. See what we offer and get in touch to scope the first ten critical tasks.