Training Needs Analysis Work Sample for Learning and Development Teams and Human Resource Business PartnersBefore anybody builds the course: is training the answer, and what is the cause?
Forty-four exercises on the decision a training needs analysis is actually made of. Five performance problems arrive as they really do: the sponsor's request in the sponsor's own words, two or three lines of record, what the manager said, and the number. For each one you weigh six named causes: skill and knowledge, expectation and feedback, tools and environment, incentives and consequences, process and permission, selection and capacity. Twelve keyed exercises test what to find out first and whether the response you would propose is sized to the diagnosis, and two written exercises ask for the first question you would put to a manager and a two-sentence reply to a sponsor. The score is a chance-corrected profile match, and the report reproduces each case down the page with your own calls in the margin beside the lines they bear on.
The first needs-analysis assessment that scores the cause call, not the course
The Training Needs Analysis Work Sample is a fifty-minute applied skill assessment for learning and development teams and HR business partners. It measures the cause call: whether the weight you give to skill, expectation, tools, incentives, process and capacity on a real performance problem matches what its records warrant, scored as a chance-corrected profile match across five cases.
The expensive, commonplace failure in training needs analysis is answering yes when the question was never asked. A sponsor asks for a course; the records show a system default that fills a field wrongly, a rota that ends one shift at the minute the next begins, a bonus that pays for skipping a step, or a standard nobody ever stated. The course runs, is enjoyed, and the number does not move. The published performance-analysis literature has asked the same question for fifty years: could they do it if they had to? Courses in needs analysis teach the question; no assessment we know of measures whether an analyst actually answers it when the request arrives with the answer already in it. This one does, and only that.
Thirty of the forty-four exercises are weights. Five cases, each a request as it arrives: a claims desk whose error rate rose with a new system, a sales floor whose new reps convert at half the rate, a ward whose night handover takes four minutes, a field crew whose lockout checklists are missing on solo jobs, and an analyst team whose client reports keep coming back. For each case you give six weights on the same six causes, on a slider or a six-point scale. Twelve keyed exercises ask what to find out first, which source answers which question, what a manager's remark points to, what the sponsor's stated solution hides, and which response is sized to the diagnosis. Two written exercises are graded against a printed rubric and reported apart.
The six causes are the cells of the published behaviour engineering model and the branches of the performance-analysis flow, with the environment read before the person, and the vocabulary for the competencies is anchored on the generalised and detailed work activities for training and development specialists and human resources specialists published in O*NET, the occupational database the US Department of Labor's programme releases under a public licence; this instrument is not affiliated with or endorsed by that programme, and reproduces none of its text. Every exercise is original, keyed to the sources cited by author in the methodology note, and on three of the five cases the warranted answer is that training is not the answer.
The scoring is a profile-match correlation, done the way the measurement literature says it has to be done. For each case, your six weights are correlated with the six the records warrant; a correlation is blind to your level and your spread, so a reader who gives everything 70 and a reader who gives everything 20 are judged on the shape of their call and nothing else. That correlation is then read against the correlation an ordinary respondent reaches, where ordinary means somebody drawing each weight from the way people typically weight that cause on that case, declared per exercise and printed on the report; on most cases that puts the mass on skill, because that is the answer the request arrives with. Zero means indistinguishable from somebody who knows nothing but weights the causes the way people typically do. A reader who gives every cause the same weight is refused a figure, in plain words, rather than scored.
The report is a margin ledger. Each case is reproduced down the page as it arrived, and your own calls sit in the margin beside the line they bear on: the weight you gave each cause, the move you chose, the response you sized, each with a mark, a word and one sentence on what that call would have cost on this case. Beside the margin, your cause profile is drawn against the warranted one as a labelled bar row, with the single largest divergence named in words, so the finding is checkable against the very text printed to its left rather than asserted. Then the six causes, each with your lean and the pair that belongs to it: the strength, and the honest downside, printed heavy for the direction your weights showed and light for the direction they did not. The downside on skill is a real one in both directions, because an analyst who never says training is the answer is as wrong as one who always does.
Around the ledger: the cause-call index with its 68 and 95 per cent bands drawn on the scale and written in words, the five cases as its contributors, the analysis-move strand and the proportionality strand reported beside it and never folded in, the written strand apart with a deferred exercise printed as deferred and never as a zero, one boxed if-then change drawn from your largest lean, the movement a re-sitting would have to clear to count as real change, the reliability table with every refusal printed, what zero means with the declared marginals, and the plain list of what the report does not say. At Rs 899 inclusive of GST in India, or US$8.99 elsewhere, that report is the product: one-day needs-analysis courses in India run at many multiples of it, and free templates carry no case, no scoring and no report.
What you walk away with
How closely the shape of your six weights matched the warranted shape across five cases, chance-corrected so that zero is the analyst who weights the causes the way people typically do and 100 is the warranted profile exactly. The 68 per cent band is drawn as a shaded block, the 95 per cent band as an outlined box, and both are written in words beneath it, with the five cases listed as contributors.
Each case reproduced as it arrived, the request, the records, what the manager said and the number, with your weight on each cause sitting in the margin beside the line it bears on, marked held, over-weighted or under-weighted, and one sentence on what that call would have cost on this case.
For every case, the six causes as a labelled bar row, your shares against the warranted shares, and the largest divergence written out: you put the weight on skill and knowledge, and took it from tools and environment; the records say a default the system fills by itself carries this case.
Whether you over-weight or under-weight each cause across the cases, evidenced only past a declared threshold, with the pair for the direction you showed printed heavy and the other printed light and labelled as not shown. On skill and knowledge the downside is written in both directions, because reaching for training by habit and refusing it by habit are both costly.
Eight exercises on what to find out first, which source answers which question, what a manager's remark points to and what the stated solution hides; four on whether the response is sized to the diagnosis. Chance-corrected against declared per-option marginals and reported on their own, because a correlation cannot carry level.
A single boxed if-then sentence naming a specific trigger and a specific replacement behaviour, drawn from your largest evidenced lean or from the commonest divergence in your ledgers. Beside it, the number of points a re-sitting after four to six weeks would have to move to count as real change rather than measurement error.
Inside your report
Illustrative sample — your report is generated from your own responses.
Down the left, the case: the sponsor’s request, the records, what the manager said, the number. In the margin, the calls you made that bear on each line, each with a mark, a word and what that call would have cost on this case. held · over-weighted · under-weighted · partly sized.
The operations head writes: "We need a two-day course on the new claims system for the back-office team. Error rates are up".
Addresses the default by asking people to correct a machine, and leaves the workload alone. Cheaper than a course and half the fix.
Seven in ten errors sit in one field, the policy start date. The new system fills it with today's date; the old one left it blank.
Seven in ten errors sit in a field the system now fills wrongly by default; ignoring that trains sixty people to correct a machine.
The team manager says: "They know how to do it. They did it right on the old system for years".
A two-day course for people who did the job right for years, while the field keeps filling itself with today's date.
Claims per person per day rose from 48 to 61 after two staff left and were not replaced.
This call matched the records: 35 in 100 warranted, and your profile put it about there.
| Line | Call | You | Warranted |
|---|---|---|---|
| The request | Partly sized · Response size | A fifteen-minute desk note on checking the start date | Blank the default and refill the two posts |
| Record 2 | Under-weighted · Tools and environment | 20 of 100 | 35 in 100 |
| The manager said | Over-weighted · Skill and knowledge | 55 of 100 | 10 in 100 |
| The number | Held · Selection and capacity | 35 of 100 | 35 in 100 |
How to read it: the finding is never asserted on its own. Every call sits beside the line of record it bears on, so you can check it against the text, and every call carries the sentence on what it would have cost.
Largest divergence: you put the weight on skill and knowledge (34 against 10 in 100) and took it from tools and environment (12 against 35 in 100); the records say a default the system fills by itself and a workload up by a quarter carry this case; the knowledge was there for years.
About two chances in three that a re-sitting today would land between 14 and 48; nineteen in twenty between -3 and 65. The band is the wider of the model’s standard error and the spread of your five case figures, because a correlation over six causes is noisy and the page says so.
| Cause | You | Warranted | Call |
|---|---|---|---|
| skill and knowledge | 34 | 10 | Over-weighted |
| expectation and feedback | 12 | 10 | Held |
| tools and environment | 12 | 35 | Under-weighted |
| incentives and consequences | 9 | 5 | Held |
| process and permission | 12 | 5 | Held |
| selection and capacity | 21 | 35 | Held |
How to read it: the profile is drawn for every case on a full report. A correlation judges the shape of your six weights and nothing else, so the bars show your weights scaled to add to 100 against the warranted shares; the divergence is named in words so it can be checked against the case text above it.
Built for
- Learning and development partners and training managers who receive requests for courses and want evidence of how they read the cause before they build anything
- HR business partners asked to turn a manager's complaint into a plan, who want to know which cause they reach for by habit and what it costs
- Performance consultants and needs-analysis practitioners who want a work sample scored against the published literature rather than a quiz on the names of models
- Heads of L&D choosing or developing analysts, who need one specific change per person with the case text beside the finding
Find out which cause you reach for before the course runs and the number does not move
44 exercises across eight formats · about 50 minutes · five cases, six causes, one cause-call index with both bands, a margin ledger per case with every call costed, your cause profile drawn against the warranted one, and two written exercises graded apart. Rs 899 in India inclusive of GST, or US$8.99 elsewhere. Sit it again after four to six weeks of the boxed change.
₹899 (incl. GST) · assessment and full report, nothing further to pay
Frequently asked questions
It measures the cause call: whether the weight you give to skill and knowledge, expectation and feedback, tools and environment, incentives and consequences, process and permission, and selection and capacity on a real performance problem matches the weight its records warrant. Knowing the names of the models is not the construct. A sponsor asks for a course and attaches the records; the records show a system default, a rota with no overlap or a bonus that pays for skipping a step; the analyst who still weights skill has answered the sponsor's question with the sponsor's answer. That is what this measures, through five cases and thirty weights, and the report puts your calls in the margin beside the very lines of record they bear on.
By a chance-corrected profile match. For each case your six weights are correlated with the six the records warrant, so your level and your spread drop out and only the shape of your call is judged. That correlation is read against the correlation an ordinary respondent reaches, where ordinary means somebody drawing each weight from the way people typically weight that cause on that case, declared per exercise and printed on the report. Zero means indistinguishable from somebody who knows nothing but weights the causes the way people typically do; 100 is the warranted profile exactly on every case; a negative figure is further off than ordinary. A reader who gives every cause the same weight is refused a figure rather than scored.
The Instructional Design Work Sample measures alignment across the design chain, which is what happens after somebody has decided training is the answer. The Training Evaluation and Programme Effectiveness Assessment measures which level of evidence a programme's claim needs after it has run. The Technical Trainer Readiness diagnostic measures a subject expert's readiness to train their own team. This is the decision before all three: whether there is a training need at all, and what the cause actually is. On three of its five cases the warranted answer is that training is not the answer, and the report says so in its refusal block.
Each case is reproduced down the page as it arrived: the request in the sponsor's words, the records, what the manager said and the number. Beside each line sit the calls you made that bear on it, each with a mark, a word such as held, over-weighted or under-weighted, and one sentence on what that call would have cost on this case. Beneath the text your six weights are drawn as a bar row against the warranted shares, with the largest divergence named in words, so you can check the finding against the record printed above it instead of taking the report's word for it. The six causes then each carry your lean, with the strength and the honest downside printed heavy for the direction your weights showed.
Rs 899 in India inclusive of GST, or US$8.99 elsewhere, one time; 30 credits on a business plan. Forty-four exercises across eight formats take about fifty minutes. The report names a re-sitting after four to six weeks of the boxed change and prints the number of points the cause-call index would have to move to count as real change rather than measurement error. No total is printed across the six causes, no percentile appears anywhere, and you are ranked against nobody. One-day needs-analysis courses in India run at many multiples of this price, and free templates carry no case, no scoring and no report.
Each one takes a single capability, puts you inside the situations where it is actually tested, and scores your choices against published evidence — with a report designed for that capability alone, not a template. They span hiring, compliance, education, operations and personal skill.
Browse the catalogue →Methodology: Forty-four original exercises across eight formats: thirty cause-weight exercises, eighteen on sliders and twelve on six-point scales, across five performance cases, each asking how much of the weight each of six named causes carries; two first-move scenarios with graded options; four single-choice exercises; three select-every-line exercises; two match-the-following exercises; one ordering exercise; and two written exercises graded by rubric and reported apart. DECLARED RESPONSE INSTRUCTION: one instruction for the whole instrument, what do the records of this case WARRANT, judged against the published performance-analysis literature; a knowledge-style instruction about the case, not a behavioural tendency, mixed with nothing. CONSTRUCT STATEMENT: This measures the cause call in a training needs analysis: whether the weight a reader gives to skill and knowledge, expectation and feedback, tools and environment, incentives and consequences, process and permission, and selection and capacity, on a performance problem as it arrives, matches the weight the records of that problem warrant. It is measured as the profile-match correlation between the reader's six weights and the warranted six, per case, chance-corrected against what people typically give each cause on that case. It does not measure instructional design, facilitation, training evaluation, writing style, or knowledge of any named model. It is the decision BEFORE those: whether there is a training need at all, and what the cause actually is. FRAMEWORK: the competency vocabulary is anchored on the generalised and detailed work activities for training and development specialists and human resources specialists in O*NET, the occupational database the US Department of Labor's programme publishes under a public licence; this instrument is not affiliated with or endorsed by that department or programme, and none of the database's text is reproduced. SCORING DESIGN, C3, profile-match correlation: for each case, match = Pearson correlation between the reader's answered weights and the warranted weights across the six causes; a correlation is invariant to the reader's level and spread, so no ipsatisation is needed and elevation drops out. E[match_null] is the seeded Monte-Carlo mean (400 draws) of that correlation for a respondent drawing each cause's weight from THAT ITEM'S DECLARED MARGINAL (value_base_rate), never uniform random and never the scale midpoint; index per case = 100 (match - E[match_null]) / (1 - E[match_null]); the cause-call index is the mean over scored cases, with SEM and 68 and 95 per cent bands. A case on which the reader gave every cause the same weight has no shape; it is refused and flagged on the RAW presses before any mapping, never correlated and never divided by zero, and a reader flat on three or more cases is refused the index. Because a correlation discards level, two keyed strands carry what level carries and are reported APART and never folded in: the analysis MOVE (what to find out first, which source answers which question, what the stated solution hides) and PROPORTIONALITY (whether the response is sized to the diagnosis), each chance-corrected against declared per-option marginals. The six cause competencies carry each cause's share of the squared standardised departure from the warranted profile, chance-corrected the same way, plus the keyed exercises tagged to that cause; a competency carries a number only with at least eight answered items and omega at or above .70 unrounded, and otherwise a three-way placement word. The written strand is reported apart as criteria met or DEFERRED. REPORT DESIGN, D17, the Margin Ledger: each case is reproduced as the spine of the page, the request in the sponsor's words, the records, what the manager said and the number, and the reader's own calls sit in the margin beside the line they bear on, each with a geometric mark, a word and one sentence on what that call would have cost on this case; beside the margin the cause profile is drawn as a labelled bar row, the reader's shares against the warranted, with the largest divergence named in words. SOURCES: Mager and Pipe, Analyzing Performance Problems (CEP Press, 3rd ed., 1997), for the skill-deficiency question keyed on the flagship exercise and the ordering exercise; Gilbert, Human Competence: Engineering Worthy Performance (McGraw-Hill, 1978), for the behaviour engineering model's six cells and the priority of environment over person that orders the cause set; Rummler and Brache, Improving Performance: How to Manage the White Space on the Organization Chart (Jossey-Bass, 2nd ed., 1995), for the performance-system view keyed on the process and tools causes; Robinson and Robinson, Performance Consulting (Berrett-Koehler, 1995), for the should-is gap and the move from stated solution to business need; Rossett, Training Needs Assessment (Educational Technology Publications, 1987) and First Things Fast (Pfeiffer, 1999), for extant data, optimals and actuals, and the ordering of sources; Kaufman, Strategic Planning Plus (Sage, 1992), for the organisational elements model that separates means from ends; Harless, An Ounce of Analysis Is Worth a Pound of Objectives (Harless Performance Guild, 1975), for front-end analysis; Stolovitch and Keeps (eds.), Handbook of Human Performance Technology (Jossey-Bass, 2nd ed., 1999), for the convergent cause taxonomy; Kirkpatrick, Evaluating Training Programs (Berrett-Koehler, 1994), and Alliger, Tannenbaum, Bennett, Traver and Shotland, A meta-analysis of the relations among training criteria (Personnel Psychology, 1997), for the evidence levels the proportionality strand stops short of; Holton, The flawed four-level evaluation model (Human Resource Development Quarterly, 1996), for the critique that evaluation without a cause model cannot say why training did not work; Baldwin and Ford, Transfer of training (Personnel Psychology, 1988), for the work-environment explanation of a trained behaviour that does not appear; Cronbach and Gleser, Assessing similarity between profiles (Psychological Bulletin, 1953), for the elevation, scatter and shape decomposition that justifies a correlation match; Baumgartner and Steenkamp, Response styles in marketing research (Journal of Marketing Research, 2001), for the response-style contamination the correlation removes; Gollwitzer and Sheeran, Implementation intentions and goal achievement (Advances in Experimental Social Psychology, 2006), for the if-then form of the one change; Hattie and Timperley, The power of feedback (Review of Educational Research, 2007), for the structure of the report; Haladyna, Downing and Rodriguez, A review of multiple-choice item-writing guidelines (Applied Measurement in Education, 2002), for cue control. All forty-four items are original works written for this instrument; no item from any published questionnaire, certification programme or commercial needs-analysis course is reproduced or adapted, no branded instrument or vendor product is named or used, no competitor is named anywhere, and the instrument is not affiliated with or endorsed by any author, publisher, professional body, government programme or certification scheme named above.