HR Analytics and Innovations in Workforce Planning

1. 1Define competency as a key observable behavior: skills, knowledge, and experience you can see, measure, and improve. If you cannot observe it, it is not a competency for measurement purposes.
2. 2Allocate only the key competencies to each job, the ones critical to it. Do not try to cover every item of work. Too many competencies make measurement unattainable and kill training-needs analysis.
3. 3Map each competency to one of the three outcomes it must deliver: quality assurance, conformance to standards, or doing things in a safe and legal way. Drop any competency that serves none of them.
4. 4Define the core skills for the role in concrete terms. The passages' loan-service example lists Point software, Desktop Originator software, Loan Prospector software, telephone usage and predictive language, accuracy of documentation, and product knowledge.
5. 5Set up Terms of Reference (the measurement requirements) before you score anyone. The example set includes total calls vs. total sales, letters of praise, a phone matrix, a competency matrix, and referrals.
6. 6Score every employee as an exact number on a linear 0-100 scale, not a tick-box band. Collect it annually, linked to the performance appraisal cycle.
7. 7Build a training program with specific objectives and phases, so that loan counselors, processors, and their leaders and managers reach a basic designated performance level against the terms of reference and core skills. Back it with lesson plans and a testing program.
8. 8Test the training in a remedial mode on existing staff, then revise it from their feedback. Report competency trends by individual and by department, and project them forward alongside performance and reliability.

The passages describe a loan service-center competency build. The core skills are defined as Point, Desktop Originator and Loan Prospector software, telephone usage and predictive language, accuracy of documentation, and product knowledge. The Terms of Reference are set as total calls vs. total sales, letters of praise, a phone matrix, a competency matrix, and referrals. A total training program is then developed with specific objectives and phases, so loan counselors, loan processors, and their leaders and managers reach a basic designated performance level against those terms and skills. Lesson plans and a testing program follow. The program is tested in a remedial mode on existing staff and revised from their feedback. The sequence runs core skills, then measures, then training, then test, then revise. Scores are recorded on the 0-100 scale so gaps by person and department can be modeled.

• If your appraisal form uses Unsatisfactory / Below average / Average / Outstanding boxes, then replace it with a 0-100 exact score. Each band spans 25 points and is too coarse for modeling and forecasting.
• If a job carries a long competency list, then cut it to the key competencies only. If you cannot measure and train against the list, it is too long.
• If a proposed competency is not an observable behavior (skills, knowledge, experience), then rewrite it until a manager can see and score it, or drop it.
• If a competency does not support quality assurance, conformance to standards, or safe and legal working, then it does not belong in the framework.
• If competency scores are low in a department, then feed that into training-needs analysis and the department-level competency strengths and weaknesses report.

• Tick-box scoring: managers find it easy, but scores cluster in four bands. The tell is that you cannot trend or project competency from the data.
• Competency inflation: every task gets its own competency. The tell is that no one completes the measurement and training-needs analysis is unattainable.
• Treating competency as a trait or attitude rather than observable behavior. The tell is that two managers score the same person very differently with no evidence.
• Collecting competency without performance and reliability. The tell is that productivity is explained by one component, when sustainable productivity needs all three.
• Rolling out training without a test and a remedial pilot. The tell is that the program reaches staff before feedback has been used to revise it.

• ▸Job / role: name the job being assessed.
• ▸Key competency (observable behavior): state it as something a manager can see and measure.
• ▸Skills / knowledge / experience: list the components that make up the behavior.
• ▸Outcome served: mark quality assurance, conformance to standards, or safe and legal working.
• ▸Terms of Reference measure: name the evidence used, such as a phone matrix, competency matrix, or referrals.
• ▸Score (0-100): record an exact number, not a band.
• ▸Date and appraisal cycle: record the annual collection point.
• ▸Training need: describe the gap and the training action it triggers.
• ▸Department roll-up: record the department average to show strengths and weaknesses.

1. 1Define performance as raw output only: how much a person does. Keep it separate from competence (what they can do). Record the two in different columns so a skilled but slow employee is not scored as a strong performer.
2. 2Pick the measures from the source list: speed, time, efficiency, unit cost, volume. Assign one or two to each role (for example, volume of files processed, or unit cost per file). Do not use all five on every role.
3. 3Write the expected performance into the contract of employment as a clear, countable level. Take legal counsel first, because employment law statutes on this vary geographically.
4. 4Express the target as a dated objective with exact numbers, not percentages. Follow the source pattern: a plan outline date, an approval date, a first batch (50 files by 1 October 2015) and a total (600 extra files by 1 October 2016). Break big objectives into smaller chunks to prevent slippage.
5. 5Measure against true hours worked, not hours paid. Take actual output per actual hour and compare it with the standard to get each person's performance level.
6. 6Categorize poor performers and cost the lost performance. The source relies on a 'work done' survey for this. Convert the shortfall into unit cost and hours so it can feed rightsizing decisions.
7. 7Lock performance above the required level into a bonus or reward system. If the original criteria are right, employees are already near their optimal level, so extra output should be hard to produce in the same time.
8. 8Choose bonus or overtime for a given role, never both. Then report the results on a productivity dashboard, tracked over time, so trends can be projected.

The source's own objective example gives the pattern. A department must show 600 extra files processed, and the objective is split into dated chunks: outline plan approval by 2 April 2015, final agreement by 1 May 2015, a first batch of 50 extra files by 1 October 2015, and all 600 by 1 October 2016. Performance here is pure volume, counted in files and tied to dates, with no percentages to misread. Each month you compare files actually completed with the schedule. The Triple W method does not say how to do the work, so the team works that out itself, which the source says raises commitment and accountability. If someone clears more than the required files in the same hours, the excess goes to the bonus scheme rather than overtime. At the organization level, the source's benchmark is that most companies are overstaffed by 15 to 20 percent. Output per hour data tells you whether your own department is in that range.

• If a person's output is low but their skills assessment is high, then treat it as a performance problem (volume, speed, hours), not a competence problem, and do not send them to training first.
• If output exceeds the contracted level in the same hours, then pay a bonus or reward, and do not also pay overtime for the same work.
• If a target is written as a percentage, then restate it as an exact number with a day, month and year to avoid misunderstanding.
• If the objective runs a long time, then split it into dated batches, because long undivided objectives invite slippage.
• If measured output per true hour shows a persistent gap across a unit, then cost the lost performance and feed it into the rightsizing analysis. The source suggests overstaffing of 15 to 20 percent is common.

• Confusing performance with competence: a strong assessment score is used to excuse low output. The tell is that high-competence people appear on the poor-performer list with no volume or time data behind the rating.
• Measuring paid hours instead of true hours worked. The tell is that output per head looks fine while actual productive time is never recorded.
• Vague contract performance expectations. The tell is that managers cannot say what 'enough' output is, and disputes arise with no legal footing.
• Paying both bonus and overtime for the same output. The tell is that labor cost per unit rises even though volume does not.
• Objectives written with percentages and no dates. The tell is that nobody can say on a given day whether the team is on schedule.

• ▸Role / person: name the role and incumbent being measured.
• ▸Primary measure(s): choose from speed, time, efficiency, unit cost, volume.
• ▸Contracted performance level: the exact number expected, as written in the employment contract.
• ▸True hours worked (period): actual hours, not hours paid.
• ▸Actual output (period): units completed in those hours.
• ▸Output per true hour vs standard: compute and note the gap as an exact number.
• ▸Dated milestones: batch targets with day, month and year (e.g., first batch, final total).
• ▸Above-standard output rule: record bonus OR overtime, never both.
• ▸Cost of lost performance: convert the gap to hours and unit cost for the rightsizing review.

1. 1Define the metric before you collect it. Reliability is whether a person works the hours they are paid for. Track only the two areas the source says matter most: sickness and unsubstantiated days off (uncertified sickness or other reasons). Log late starts and early finishes as secondary signals.
2. 2Set the cost base first. Pull the cost per employee per day and per hour (the source points to its Chapter 7 for this). Reliability cost is meaningless without these two numbers.
3. 3Collect the data on an exact 0–100 linear scale, not tick-boxes. The source warns that bands such as Unsatisfactory 0–25, Below average 26–50, Average 51–75 and Outstanding 76–100 are too coarse for modeling and forecasting. Convert reliability into a precise score, not a category.
4. 4Gather it at least annually, tied to the performance appraisal cycle, alongside the competence and performance scores. Reliability is the third component of sustainable productivity, and the one most often neglected.
5. 5Pull the absence fields the old manpower records already hold: number of employees absent, days lost, occupation and level, grade, age and sex. Re-cut them by individual, section and department, because the cost can be calculated at each of those levels.
6. 6Convert lost time into money: late minutes and early departures converted to hours at the hourly cost, plus uncertified sick days at the daily cost. The source says it has a formula for this but does not reproduce it, so build and document your own explicitly and apply it identically everywhere.
7. 7Project the cost forward. The source's predictive tools show it over 5, 10 and 15 years. Run the same projection with your own method, and watch the sickness and age trends and the time-off/smoker correlation the source lists as critical forecasts.
8. 8Wire reliability into pay and reporting. Condition individual bonuses on average competence and reliability scores being achieved. For team bonuses, use the Hopper scheme, where every team member must meet the competence and reliability score before any bonus is earned. Report the figure on the productivity dashboard.

The source's own case is one employee who arrives 30 minutes late every day and takes 14 uncertified sick days in a year. The source asks for the one-year reliability cost, but its figure is cut off in the supplied text, so no pound value is stated here. The method is as follows. Step 1: take the employee's hourly cost and daily cost from your cost model. Step 2: lateness cost = 0.5 hours × hourly cost × the number of days worked in the year. Step 3: sickness cost = 14 × daily cost. Step 4: add the two for the one-year reliability cost. Step 5: project it over 5, 10 and 15 years, as the source says its tools do. Step 6: repeat for every person in the section, then sum to the section and department. The source's hours-worked study makes the stakes visible. In a 40-hour week, talented workers (17 percent) worked 32 hours, average workers (61 percent) worked 22, and poor performers (22 percent) worked 5. That weights to roughly 20 productive hours of 40 (5.44 + 13.42 + 1.1), so paid hours and worked hours diverge sharply before sickness is even counted.

• If a manager's appraisal uses tick-boxes for reliability, then re-collect it as an exact 0–100 score, because banded data cannot be modeled or projected.
• If an employee is paid a bonus, then require their average competence and reliability scores to meet the threshold first. For teams, use the Hopper rule: nobody earns until everyone meets the score.
• If you cost reliability, then calculate it by individual, section and department and project it over 5, 10 and 15 years, because the single-year figure understates the exposure.
• If the cause is unsubstantiated absence rather than certified sickness, then treat it as reliability (not generic absenteeism) and target the uncertified sick days and unexplained days off first.
• If overall performance looks weak and managers ask for more staff, then check reliability and available time before approving headcount. The source notes that poor performance is compensated for by demanding more staff, which produces overstaffed organizations.

• Treating reliability as just absenteeism and counting only certified sickness. The tell: uncertified sick days and unexplained days off do not appear in your reports.
• Measuring days lost without a cost. The tell: the report says 'X days absent' with no hourly or daily cost attached, so nobody acts on it.
• Collecting reliability on a banded or tick-box scale. The tell: scores cluster in four ranges and trend lines cannot be fitted.
• Reporting only the current year. The tell: no 5-, 10- or 15-year projection exists, so the figure reads as minor.
• Paying individual or team bonuses without a reliability gate, or letting forced ranking override it. The tell: high-target achievers with poor attendance still collect full bonuses, or one absent team member does not affect the team payout.

• ▸Employee / section / department: identify the unit so the cost can roll up to section and department.
• ▸Hourly cost and daily cost: take both from your employee cost model; do not estimate them here.
• ▸Late-start minutes per day and working days per year: record the average lateness and the days it applies to.
• ▸Early-departure minutes per day: record leaving-early time as a separate line.
• ▸Uncertified sick days: count the days of uncertified sickness in the year.
• ▸Other unsubstantiated days off: count days off without a substantiated reason that are not sickness.
• ▸One-year reliability cost: (late and early hours × hourly cost) + ((uncertified sick + other unsubstantiated days) × daily cost).
• ▸Projected cost at 5 / 10 / 15 years: carry the figure forward using your documented projection method.
• ▸Reliability score (0–100) and bonus gate: record the exact appraisal score and whether it meets the threshold for bonus eligibility.

1. 1Define the score before you collect it. Use one 0-100 staff satisfaction index, built from the same questions every cycle, so periodic results compare. The source gives no question set, so fix yours and freeze it.
2. 2Tie the survey to a purpose. The source places employee surveys under the culture strategy: they link the current state of corporate culture to the desired future state. Write the target state first, then choose questions that show the gap.
3. 3Run a parallel manager survey. The source pairs employee surveys with manager surveys on cultural alignment and satisfaction with the working environment. Compare the two scores. A gap between managers and staff is the first finding.
4. 4Cut results by the fields the HR data already holds: department, occupation and level, grade, age, sex, length of service, and temporary/permanent/shift status. Aggregate scores hide the group that is leaving.
5. 5Join the score to leaver records: reason for leaving, length of service at leaving, age and grade. Add absence data (employees absent, days lost). Satisfaction matters here because it is tied to turnover and time lost, so test whether low-scoring groups are the ones leaving or absent.
6. 6Price the link. Put the costs of recruitment, training, staff turnover and lost time against the low-scoring groups. This turns the score into financial information, which the source says analytics must produce.
7. 7Re-survey after every structural or people change, such as delayering, rightsizing, a new management style or a job-security measure. Report the before and after score next to the productivity result.
8. 8Act on it: pick one lever from the source, such as job security (long-term contracts, share/stock options, development programs), career development, or management style against shared values. Then forecast the effect on turnover and productivity.

The source lists "improved worker satisfaction scores" as a benefit of delayering, alongside a 28 percent productivity improvement and less management. It cites British Airways removing 450 managers with no reported operational impact, and Essex County Council removing 20 percent of its management layers. It gives no actual satisfaction figures, so you supply your own. Before delayering, run the employee and manager surveys and record the 0-100 index by department and level. Ask the source's question, "Where in the organization is the work actually done?", and remove layers accordingly. Re-survey after the change and compare each cut to its baseline. The Anglian Water case shows why you segment. Staff fell from 6500 to 4500 with higher productivity, and John Simpson was popular with hard workers and field experts, while poor performers feared and resented him. A single average could mask both reactions, so split scores by performance and role. Then check the leaver data and report whether the score, turnover and productivity moved together. The source warns that downsizing carries risk and advises using a consultant with proven experience.

• If the employee score and the manager score diverge, then treat it as a cultural-alignment problem and investigate before launching any retention program.
• If a department's score is low and its leaver or days-lost numbers are high, then cost that group's turnover and lost time first and act there.
• If a restructure is planned, then take the baseline survey beforehand. Without it, you cannot show the "improved satisfaction" the source associates with delayering.
• If your workforce is highly skilled and needs little supervision, then reduce supervision layers and expect satisfaction to rise. If it does not, check shared values and style (the Miller/Sporlein model) rather than assuming more management is the fix.
• If the overall average moves but a segment (for example high performers) drops, then act on the segment. Averages hide who is leaving.

• Changing the questions between cycles. The tell: scores jump with no matching change in turnover, absence or events.
• Reporting one company-wide number. The tell: the average is stable while specific departments or grades show rising leavers.
• Surveying employees only. The tell: managers say the culture is aligned while staff scores say otherwise, and nobody has compared them.
• Never linking the score to leaver reasons, absence or cost. The tell: the report ends at the score, with no financial implication or forecast.
• Claiming satisfaction gains from a restructure with no baseline. The tell: the "before" figure is missing or remembered rather than measured.

• ▸Survey cycle and date: record when the survey ran and keep the same question set as the previous cycle.
• ▸Target culture state: write the desired future state the survey measures against.
• ▸Employee index (0-100): record the overall score and the score per department, level and grade.
• ▸Manager index (0-100): record the manager survey score on cultural alignment and working environment, and the gap versus employees.
• ▸Lowest-scoring segment: name the group, its headcount and length of service profile.
• ▸Leavers and days lost in that segment: record the number of leavers, reasons for leaving, and days lost to absence.
• ▸Cost of turnover and lost time: put the costs of recruitment, training, turnover and lost time against the segment.
• ▸Change event and baseline: note any delayering, rightsizing or style change, the pre-change score and the post-change score.
• ▸Chosen action and owner: select one lever (job security, career development, management style) and name who delivers it and by when.

1. 1Fix the scoring scale before anything else. Collect every competence and performance score on a linear 0–100 scale with exact numbers. Tick-box bands (Unsatisfactory 0–25, Below average 26–50, Average 51–75, Outstanding 76–100) are too coarse for modeling or forecasting, because a 51 and a 75 both read as 'Average'.
2. 2Pull the scores from the performance appraisal process, which is the source for both competence and performance figures. Check that appraisers enter a number rather than a box. Also collect reliability, which the source calls the most neglected of the three measures.
3. 3Keep the competency list short so the scores mean something. Rate only key competencies: no more than eight per employee, six on average. Over-allocation makes measurement unattainable and degrades the data you are about to split into groups.
4. 4Set cut-points on the 0–100 scale that divide people into poor, average, and high performers. Write the cut-points down and apply them identically to every department. The source gives no cut-point values, so choose them deliberately rather than inheriting the old tick-box bands.
5. 5Count headcount in each group, then express each group as a proportion of the total workforce. This is the composition figure.
6. 6Measure actual hours worked per group instead of assuming contracted hours. The source's 'work done' survey categorized poor performers so their lost performance could be measured and costed. Run an equivalent study, and record hours and output for each group separately.
7. 7Carry composition and hours into the forecast and rightsizing calculation as separate inputs. Do not feed in a single average-employee figure. The source calls this differentiation 'the foundation' and 'essential when rightsizing'.
8. 8Re-run the composition each appraisal cycle and track the shift as a trend. Composition is a forecast input, so a change in the poor-performer share changes the headcount answer.

The source's call-center case shows why one average hides the composition. Shellie Musker, a salesperson in a large call center, trained out of office hours and became the best salesperson in the center. She stayed the best for just over a year, was beaten for one month, then produced the highest sales ever, at twice the level of average performers. Take an average performer as 1.0 unit of sales: she is 2.0, so counting her as one head undercounts her output by half. Now compare a workforce that is 'all average' with one that has the same headcount but includes people like her. A forecast built on average output understates capacity in the second case, and a rightsizing cut made by headcount could remove the output of high performers. The reverse holds for poor performers: if their actual hours and output fall below average, a headcount-based figure overstates capacity. The source says the 'work done' survey existed so that poor performers could be categorized and their lost performance measured and costed. The survey results themselves are not in the supplied passages, so you must gather your own hours and output data per group.

• If appraisal data arrives as tick-box bands, then re-collect it on the 0–100 linear scale before computing any composition, because band-level data cannot support forecasting.
• If a forecast or rightsizing model uses one average output per employee, then replace it with group proportions multiplied by each group's actual hours and output.
• If you have only contracted hours for a group, then run a work-done survey for that group's actual hours before trusting the capacity figure.
• If a department's poor-performer share is rising across appraisal cycles, then treat it as a forecast driver and cost the lost performance rather than hiding it in the average.
• If someone proposes cutting headcount to hit a target, then first check how the cut falls across performer groups, because the design and the composition determine the right numbers.

• Using tick-box appraisal bands as the group definition. The tell: a wide spread of people (51 to 75) all labelled 'Average', and no way to see movement within a band.
• Assuming everyone works contracted hours. The tell: capacity forecasts that match the roster exactly but miss actual output.
• Treating all employees as equivalent in rightsizing. The tell: the model has one productivity number, and cuts are made by headcount percentage.
• Overloading the competency list so scores are unreliable. The tell: more than eight competencies per role and appraisers rating by guesswork or leaving fields blank.
• Measuring composition once and never refreshing it. The tell: forecasts use last year's group proportions while turnover, training, and team changes have shifted them.

• ▸Department / population: name the group being analyzed and the appraisal cycle the data comes from.
• ▸Scale check: confirm scores are exact numbers on a 0–100 linear scale (yes/no); if no, stop and re-collect.
• ▸Cut-points: write the 0–100 score boundaries used for poor, average, and high performers.
• ▸Headcount per group: record the number of people in poor, average, and high.
• ▸Proportion per group: divide each group's headcount by total headcount.
• ▸Actual hours worked per group: record measured hours (from the work-done survey or equivalent), not contracted hours.
• ▸Output per group relative to average: set average = 1.0 and record each group's output ratio (the call-center high performer sold 2.0× average).
• ▸Cost of lost performance (poor group): record the measured and costed shortfall against average.
• ▸Forecast / rightsizing implication: note how the group mix changes the capacity or headcount result versus an all-average assumption.

1. 1Decide who owns it before anything else. The source calls BPR 'an orphan in an organization with nobody taking care of it', and the usual reaction is 'we need consultants' when it is already too late and very costly. Assign BPR to workforce planning as a standing responsibility, with the chief executive and main board as its customers.
2. 2Pick the timing from evidence, not from irritation. Plot the organization or department on the MILLER curve using the Appendix 3 questionnaire. The book lists 'critical periods for Business process Re-engineering' among the forecasts workforce planning produces. The warning sign is the decline phase: too many people, too procedure-bound, not enough output. Also check which HR actions in the strategic plan will force process change.
3. 3Choose one process and draw it on a swim-lane chart, with one lane per role or department. Mark every hand-off. Record elapsed time (clock time from trigger to finished output) separately from working time, because the aim is to cut elapsed time.
4. 4Apply Drucker's test to every step: 'Would the roof cave in if we stopped doing this altogether?' If not, abolish the step. Modify only the steps that survive. The aim is to stop processes that 'take energy away from the business'.
5. 5Score each surviving step against Hammer's FACE: Fast, Accurate, Cheap, Easy. Apply it to internal and external customers alike. Redesign any step that fails at least one letter, and aim for doing things 'completely differently', not for copying past practice.
6. 6Redraw the swim-lane chart as a radical redesign of the whole flow, not an edit of the old one. Fewer lanes and fewer hand-offs are the visible result.
7. 7Cost the change in the source's three impact areas. PEOPLE: will headcount change to fit the new process, and what is the value of the saving? PROCESS: how will the change be implemented and carried out, and what is it worth? STRUCTURE: what organizational structure supports the re-engineered process?
8. 8Report the result to the board as a financial outcome, since the book wants workforce planning to produce 'information with financial implications/benefits' and measurable added value. Then re-plot on the MILLER curve to see whether the organization has moved.

The source gives no timed case study, so this applies its own tests to the one process problem it names, recruitment. The source says line managers leave recruitment too late when workloads rise or business plans expand, so it becomes urgent and is handled by correcting mistakes. Draw the current requisition-to-start process on a swim-lane chart with lanes for line manager, HR, and approver. Record the elapsed days at each hand-off (your own figures; the source supplies none). Run Drucker's test on each step: any approval or report that could stop without the roof caving in is removed. Score what remains on FACE and flag the steps that are slow or hard to use. Redesign so the trigger comes from workload and business-plan forecasts, not from a manager's late request, which fits the book's point that workforce planning is predictive. Then cost it in the three areas the source names: PEOPLE (does the headcount needed to run the process change, and what is the saving), PROCESS (what implementation costs and delivers), STRUCTURE (which role owns it now).

• If the MILLER questionnaire puts the unit past its peak, with too many people, too many procedures, and falling output, then treat that as the critical period for BPR and start now, not after the problem is urgent.
• If a step fails Drucker's 'would the roof cave in' test, then abolish it. Do not streamline it.
• If a surviving step is not Fast, Accurate, Cheap, and Easy, then redesign it. Do not add effort to it, because hard work cannot compensate for a faulty process.
• If a redesign changes headcount, procedure, or reporting lines, then state the value of each in PEOPLE, PROCESS, and STRUCTURE terms before approval.
• If the proposal is 'copy what we or others did before', then reject it. The source says competitive advantage is not achieved by copying the past.

• BPR has no owner and is outsourced in a panic. The tell: 'we need consultants' is the first response, and the work starts after the problem is urgent and costly.
• Teams tune the existing process, with a faster form or a reminder, instead of redesigning it. The tell: the new swim-lane chart has the same lanes and hand-offs as the old one.
• Activities are kept without asking whether they matter. The tell: no one can say what would break if a step stopped.
• Activity is mistaken for process quality: managers push people harder or cover over a process failure. The tell: the problem is solved by urgent day-to-day correction and recurs every time workload rises.
• The change is never costed. The tell: no figure exists for headcount, process, or structure impact, so the board sees no financial benefit.

• ▸Process name and owner: one named person, not 'HR' in general.
• ▸Trigger and trigger date: why now? Cite the MILLER curve position or the strategic-plan action forcing change.
• ▸Current swim-lane chart: attach it, with one lane per role or department and every hand-off marked.
• ▸Elapsed time vs working time: record both for the whole flow and per step, using your own measured figures.
• ▸Drucker test per step: 'Would the roof cave in if we stopped this?' Write yes/no and keep or abolish.
• ▸FACE score per surviving step: mark Fast, Accurate, Cheap, Easy as pass or fail, and note the failing letter.
• ▸Redesigned swim-lane chart: attach it, with fewer lanes and hand-offs than the current one.
• ▸PEOPLE impact: number of staff needed under the new process, and the value of the saving.
• ▸PROCESS impact and STRUCTURE impact: how the change will be implemented and carried out, its value, and the organizational structure needed to support it.

1. 1Define productivity as three components you score separately: competency, performance, reliability. Do not collapse them into one appraisal grade. The source calls reliability the most neglected, so give it its own column from the start.
2. 2Collect all three at least annually and tie collection to the performance appraisal cycle, since competence and performance figures normally come from it. Treat the appraisal form as your data-capture instrument.
3. 3Replace tick-box ratings with exact scores on a 1-to-100 scale. A box such as 'Average' covers 51 to 75, which is too coarse for modeling and forecasting. Instruct managers to enter a number, not a band.
4. 4Score competency against defined units, such as Communicating and Influencing or Leadership. Each unit has a definition and a behavioral anchor, for example Leadership is 'getting the best from others'. Set an organizational minimum standard (the source illustrates 50 percent) and flag everyone below it.
5. 5Measure performance as raw output, meaning how much gets done. Pick from the source's list: speed, time, efficiency, unit cost, volume. Choose the measures that fit each role and record them as numbers.
6. 6Measure reliability separately from competency and performance. The source names time off, sickness, and lost time among the costed items, so start there. The passages do not give a full reliability formula, so define yours explicitly and keep it constant year to year.
7. 7Cost the gaps. Use the source's cost categories (recruitment, training, staff turnover, lost time, reliability) to put a monetary value on lost performance, as the 'work done' survey did when it categorized and costed poor performers.
8. 8Feed the scores into trend and correlation work (for example training against productivity, time off against other factors) and into rightsizing, so the data drives forecasts rather than sitting in appraisal files.

The source warns that tick-box appraisals use four bands: Unsatisfactory 0 to 25, Below average 26 to 50, Average 51 to 75, Outstanding 76 to 100. Take two illustrative employees (the scores are invented to show the effect of the bands). One is truly a 52 and the other a 74. A manager ticks 'Average' for both, so the data says they are identical although 22 points separate them. Neither shows movement year to year unless they cross a band boundary, so trends and correlations cannot be computed. Now apply the source's fix: the manager records 52 and 74 on the 1-to-100 scale. Against a 50 percent minimum competency standard, the 52 sits only two points above the line and becomes a visible watch item, while the 74 does not. Repeat each year and the movement of each score becomes a usable trend, which a repeated tick in the same box cannot provide.

• If your appraisal form uses tick-box bands, then change it to exact 1-to-100 scores before building any model, because banded data is 'useless for statistical analysis and forecasting'.
• If a person scores below the organizational minimum competency standard (50 percent in the source's illustration), then treat it as a training or placement question first, and read performance and reliability alongside it before judging.
• If competency and performance are strong but reliability is weak, then do not report productivity as healthy. Cost the lost time and treat reliability as its own lever.
• If you see signs of excess headcount or procedure-bound output, then check the unit's position on the MILLER curve (Appendix 3 questionnaire) and consider rightsizing (formula 10) rather than downsizing, since the source puts rightsizing at about 4 months against 1 to 2 years.
• If data is collected less than once a year, then treat any productivity trend as unreliable, since the source requires it at least annually.

• Tick-box scoring. Tell: scores cluster in 'Average' and year-on-year change is invisible. Fix with exact numbers on a 1-to-100 scale.
• Treating productivity as performance alone. Tell: reports cite only volume or speed and say nothing on competency or reliability. Fix by scoring all three.
• Neglecting reliability. Tell: no lost-time or absence figure sits beside the productivity score. Fix by costing it separately.
• Using the past as a fixed benchmark. Tell: targets are always last year's output. The source's design advice is to avoid using the past as a benchmark, and the MILLER model shows organizations can decline over time, so check where the unit sits on the curve.
• Collecting data without costing or acting on it. Tell: appraisal scores exist but no lost-performance cost or forecast does. Fix by attaching unit costs and financial implications, as the source requires.

• ▸Employee/team and appraisal period: identify who is scored and when, aligned to the appraisal cycle.
• ▸Competency score (1-100): enter an exact number, not a band, for each competency unit used.
• ▸Minimum competency standard: record the organizational threshold (the source's illustration uses 50) and mark below/above.
• ▸Performance measure(s) chosen: pick from speed, time, efficiency, unit cost, volume, and state the unit of measure.
• ▸Performance score (1-100): enter an exact number against the chosen measure.
• ▸Reliability measure: define it (for example time off or lost time) and record the figure for the period.
• ▸Prior-year scores for all three: record last year's numbers so you can calculate the trend.
• ▸Cost of the gap: estimate the cost of lost performance, lost time, or turnover using your organization's unit costs.
• ▸Action and owner: state the training, placement, or rightsizing follow-up and who owns it.

1. 1Write down the formula before you start: Formula 9, AV (actual business value created in 1 year) − total cost of activity = added value (or loss). A negative result is a loss, so the formula can show failure as well as success.
2. 2Fix the measurement window at one year. The source ties this to the budgetary cycle, so the result can be read directly against the annual budget.
3. 3Build the total cost of the activity bottom-up in hours and pay rates. In the appraisal example, each activity is hours × a £46 hourly rate, summed per person, then multiplied by headcount. Include every participant, such as appraiser, appraisee and HR processing time, and add known hidden costs such as misdirected training.
4. 4Name the measurable business benefits before the activity runs. In the BMW case, the value created was a 7% improvement in productivity plus a reduction in sickness, set against process costs of €50,000. Choose benefits that can be quantified in money, not reported as activity counts.
5. 5Set the break-even line first: ROI requires benefits equal to total cost. Then set the added value target above it. The source's example is 20%, and it advises starting at a manageable figure such as 5% and raising it as you gain confidence and success.
6. 6Compute target benefit as total cost × (1 + target %). Compare actual measured benefit at year end and report the difference as added value or loss.
7. 7Start with workforce planning rather than all of HR. The source calls it the best first choice for moving to added value results, because its predictive work can be given a financial value.
8. 8Link a bonus scheme to the added value percentage achieved. The source proposes bonuses based on the amount of value created each year, which supports treating the department as a profit center.

The source's appraisal example costs one appraisal cycle for 3,000 employees. Each appraisal is a sum of hours × £46: £23.00 for the lines at half an hour, plus £46.00 each for appraiser time and appraisee time, with subtotal £207.00 per employee. Multiplying by 3,000 employees gives £621,000.00 total cost. The source notes misdirected training identified from appraisal could be as high as 70 percent of the training budget and should be added to the cost. To break even on ROI, the system must deliver £621,000.00 of measurable bottom-line benefits each year. With a 20 percent added value expectation, the minimum measurable benefit is £745,200.00 (£621,000 × 1.2). Applying Formula 9, if actual benefits are £745,200, added value is £124,200, which is 20% of cost. If benefits are £600,000, the result is a loss of £21,000.

• If measurable benefits are below total cost, then report a loss, not a 'soft' benefit. Fix the activity or stop it.
• If benefits equal total cost, then you have only achieved ROI. Added value requires measurable benefit above cost.
• If you have no track record of delivering value, then set the target low (the source suggests around 5%) and raise it as you gain confidence and success.
• If a benefit cannot be measured in money within the year, then keep it out of the AV calculation. The source requires actual business value created in one year.
• If you must choose where to apply added value first, then choose workforce planning, because its predictive work lets you show financial implications and benefits.

• Counting only the obvious cost. The tell: the cost line excludes participants' time, HR processing time or misdirected training. This inflates added value.
• Using a multi-year or open-ended horizon. The tell: benefits are cited as 'long-term' with no figure inside the budget year.
• Reporting activity instead of value. The tell: the output is volume of retrospective data and management information, not money above cost. The source describes this as old manpower planning, a data producer.
• Setting an ambitious target such as 20% before the function has credibility. The tell: the department misses the bar in year one and the profit-center idea is discredited.
• Claiming benefits with no stated figure. The tell: results are described as 'very substantial' without values, as in the BMW case where the values were withheld. That example is not auditable by the reader, and your own report should show the numbers.

• ▸Activity name: the single activity or department being measured.
• ▸Measurement year: the budget year the calculation covers (one year only).
• ▸Cost lines (hours × rate × people): list each step, rate and headcount, e.g. appraiser, appraisee, HR processing.
• ▸Hidden costs: add related waste, such as misdirected training, or process costs such as the €50,000 in the BMW example.
• ▸Total cost of activity: sum of all cost lines and hidden costs.
• ▸Added value target %: set at 5% to start, up to 20%, with a note on why.
• ▸Required benefit: total cost × (1 + target %).
• ▸Measured benefits (money): each benefit, e.g. productivity gain or sickness reduction, valued in currency for the year.
• ▸Added value or loss: actual business value created − total cost, plus % of cost, to compare to target and drive any bonus.

1. 1Audit your current output first. List every recurring report and mark each one retrospective (counts of jobs, employees, recruits, promotions, absences) or predictive. Old manpower planning spent about 95% of its effort on this kind of data farming, and the target for the new function is roughly the reverse, with 95% of the work predictive. Cut or automate reports that answer only past questions.
2. 2Audit your formulas. The source traced over 55 legacy manpower formulas and found fewer than 10 with real business benefit today. Keep only a formula you can tie to a decision (cost, reliability, turnover, reorganization timing). Retire any whose only use is filling a monthly report.
3. 3Convert data collection to a linear 0–100 scale. Replace tick-box appraisal bands (Unsatisfactory 0–25, Below average 26–50, Average 51–75, Outstanding 76–100) with exact numbers from 0 to 100. Banded data collapses 26 points into one value and cannot be modeled or projected.
4. 4Limit what you measure. Score only the key, observable-behavior competencies for each job. Allocating a competency to every item of work makes measurement unattainable and breaks the training-needs analysis that feeds your forecasts.
5. 5Build the forecast set from the source's list of critical information: trends (age, sickness, organizational stability); predictions (when to reorganize using the MILLER model, skill and working-pattern changes); correlations (time off and smokers, training and productivity); and costs (recruitment, training, staff turnover, lost time, reliability).
6. 6Put a money figure on every forecast. Multiply the unit cost (hourly rate × hours lost × days) by the number of affected staff. Then extend it across 5-, 10-, and 15-year horizons so the cost of inaction is explicit.
7. 7Run the model in specialist predictive software. The source points to low-cost programs that run in Excel, and the construct calls for tools such as Monte Carlo simulation. Do not hand-estimate: 'crude estimates' and partial solutions are named as no longer acceptable.
8. 8Deliver each forecast with an explanation and professional advice, as a turnkey recommendation covering short-, medium-, and long-term plans, including emergency, growth, and contraction cases. A projection with no recommended action is still just a report.

The source costs 'reliability' for one employee. The employee is 30 minutes late (0.5 hour) on 226 days, and has 14 uncertified sick days of 8 hours each, at a £46 unit rate. Lateness costs £46 × 0.5 × 226 = £5,198. Sickness costs £46 × 8 × 14 = £5,152. Total is £10,350 per employee per year. In a 3,000-person workforce, 20% fall into this pattern, which is 600 people, so the annual cost is 600 × £10,350 = £6,210,000. Projected over 5, 10, and 15 years, the cost is £31,050,000, £62,100,000, and £93,150,000. The source calls these conservative. A retrospective report would show only 'days lost'. The predictive version converts the same data into a cost, a horizon, and a case for action. The projections are flat multiples of one year, so a real model would add turnover and trend assumptions.

• If a report only counts what already happened (headcount, absences, promotions), then convert it into a forecast with a cost attached or stop producing it.
• If an appraisal or performance field uses banded tick boxes, then replace it with a 0–100 exact score before you attempt any modeling.
• If a formula cannot be tied to a business decision, then drop it. Fewer than 10 of 55-plus legacy formulas earned their place.
• If a forecast is ready to share, then pair it with an explanation, advice, and a financial figure before it reaches management.
• If you are tempted to estimate by hand or to supply only part of the solution, then move the work into predictive software and provide the full turnkey answer.

• Reporting dressed up as forecasting. The tell is a monthly pack of counts and percentages with no future dates, costs, or recommended actions.
• Coarse data that cannot support models. The tell is appraisal results clustered in four bands, so correlations such as training and productivity cannot be computed.
• Competency overload. The tell is a job profile with dozens of competencies, no one scoring them consistently, and training-needs analysis stalled.
• Using inherited formulas without testing their relevance. The tell is that no one can name the decision a formula informs.
• Crude or partial estimates. The tell is a spreadsheet figure with no software model, no horizon, and no cost, so executives ask 'so what?' and the work is not treated as a solution.

• ▸Forecast topic: name the trend or cost driver (e.g., lateness, uncertified sickness, turnover).
• ▸Data source and scale: confirm the field is collected as an exact 0–100 or numeric value, not a banded tick box.
• ▸Unit rate: enter the hourly or per-event cost (the source example uses £46).
• ▸Hours or events per employee per year: e.g., 0.5 hr × 226 days; 8 hr × 14 days.
• ▸Annual cost per employee: sum the components (source example £10,350).
• ▸Affected headcount: total workforce × % affected (source example 20% of 3,000 = 600).
• ▸Annual organizational cost: per-employee cost × affected headcount.
• ▸Horizon projections: multiply by 5, 10, and 15 years and note which assumptions are held flat.
• ▸Correlation or trend checks: list related drivers (e.g., time off and smokers, training and productivity) to model in software.
• ▸Recommendation and plan horizon: explanation, professional advice, and a short-, medium-, or long-term action.

1. 1Baseline your current design first. Count management layers from top to front line, and record the supervisor-to-employee ratio for each unit. Flag any unit running a uniform ratio near 1:7. The source describes that ratio as Adam Smith's 1760 recommendation and estimates that 95 percent of businesses are still built on it. A uniform ratio across all units is the sign of a symmetrical (Smith-type) design.
2. 2Find where the organization sits on the MILLER curve. Use the Appendix 3 questionnaire for the organization or department. The decline phase is marked by too many people, too many procedures and too little output. A position in decline justifies restructuring now. A position still on the growth side means you can design more gradually.
3. 3Segment the work by role, capability and type of work. The asymmetrical principle sets different ratios for different levels and parts of the business. Label each segment by how much supervision the work needs and how capable the people are. Do not assume everyone needs the same oversight.
4. 4Set a target ratio and layer count per segment. Where staff are smart and educated, widen the span and flatten the structure. The source says the majority of an asymmetrical organization can be only three to four tiers deep. Keep Smith-type tighter spans only where the work and people still need them.
5. 5Test the structure against the 7S 'Structure' question: is the structure correct to carry out the strategy? Then check the other S's the redesign touches: Staff (sufficient, too many, right caliber), Style (management style that matches the new spans) and Systems (will processes and computer systems support fewer layers).
6. 6Check headcount against workload before you move boxes. Divide the actual total hours worked by hours per employee per year to get the employees needed. The source's figure is 226 prime working days × 7 hours = 1,582 hours. Compare the result with the establishment (authorized headcount).
7. 7Define the success measures before the change. Use the productivity dashboard presets from the source: Competency 70%, Performance 75%, Reliability 95%. Use Formula 9 for the business case: actual business value created in 1 year minus total cost of the activity equals added value or loss. Start with a manageable added-value target, say 5 percent.
8. 8Re-measure after the change against the same baseline. Treat the source's 28 percent productivity improvement as the benefit claimed for the symmetrical-to-asymmetrical shift. Do not assume you will get it. Confirm it in your own dashboard data.

The source's contrast is between a symmetrical design and an asymmetrical one. The symmetrical design uses Smith's 1:7 span. The asymmetrical design varies ratios by level and part of the business and is listed with a 28 percent productivity improvement. Here is a purely hypothetical illustration of the arithmetic, not source data: a 49-person unit at a uniform 1:7 needs 7 supervisors plus a manager above them, which is three tiers in total. That is 8 people whose job is oversight. Under the asymmetrical approach you first ask which of those 7 teams do work that is skilled, educated and self-directed. Those teams are the candidates for wider spans or self-managed operation. The source's reasoning is that today's employee needs less management control and less process control, because education, literacy and IT competence have risen. Then you check the workload. At 226 prime days × 7 hours, each employee supplies 1,582 hours a year. Divide your actual total hours worked by 1,582 to see how many people the work needs. You compare that with the establishment before deciding how many supervisory roles to keep. Afterwards you track the change on the dashboard: Competency 70%, Performance 75%, Reliability 95%.

• If every unit shows the same ratio near 1:7, treat the design as symmetrical (Smith-type). Test whether the narrow span reflects the work or only the manager's preference for less work and less skill.
• If the workforce is educated, smart and competent with work-related IT, widen the span and cut layers. The source says the asymmetrical design will always work with smart people.
• If the work or people still need close direction, keep a tighter ratio in that segment only. Asymmetry means you do not flatten everything.
• If the MILLER questionnaire places the organization in decline (too many people, too procedure-bound, too little output), start restructuring now. Do not wait for a routine cycle.
• If the redesign cannot show added value over total cost in a year, or the dashboard stays under Competency 70%, Performance 75% or Reliability 95%, stop and revisit the span and layer choices before extending the change.

• Defaulting to small spans because they feel safe. The tell is that nobody can answer 'what do we pay managers for?' beyond oversight of capable people.
• Applying one ratio across the whole organization. The tell is identical spans in units with very different roles, capability and work type.
• Cutting layers without checking systems, style and skills. The tell is that the 7S questions on Structure, Style, Staff and Systems were never asked and the flatter structure stalls on old processes.
• Resizing without a workload check. The tell is that spans are set by org chart shape rather than by actual hours worked divided by hours per employee per year (1,582 in the source's example).
• Declaring victory without measurement. The tell is that no pre-set added-value target, no Formula 9 calculation and no dashboard baseline exist, so the claimed 28 percent gain is quoted, not observed.

• ▸Unit/segment: name the part of the business being assessed.
• ▸Current ratio and layers: record the supervisor-to-employee ratio and the tiers from top to front line.
• ▸Nature of work and capability: describe the work type and how skilled and educated the people are.
• ▸MILLER position: note where the questionnaire (Appendix 3) places this unit, growth or decline.
• ▸Target ratio and layers: set the proposed span and tier count for this segment, and say why it differs or matches 1:7.
• ▸Workload check: actual total hours worked ÷ hours per employee per year (e.g. 1,582) = employees needed, versus establishment.
• ▸7S check: note Structure, Staff, Style and Systems issues the new design creates.
• ▸Dashboard baseline and targets: record current Competency, Performance and Reliability against 70%, 75% and 95%.
• ▸Added value: Formula 9 result (actual 1-year business value minus total cost) and the preset target percent.

1. 1Build the map in three parts before any meeting: Strategic Input, Strategic Output, Business Plans for Action. Treat HR's plan as one input into the business strategy, not a standalone document. The source says HR strategy 'does not stand on its own'.
2. 2Set the timelines first. Sort every item into retrospective (what we did), current (issues to resolve), or future. Put most of the effort into the future, because it is the only timeline you control. Future items often contain retrospective and current issues, so carry those over rather than dropping them.
3. 3Complete Input 1, the strategic models, in full. Run PEST as a long-range tool: political/legal, economic, social, technology. Run Five Forces: competition among existing firms, new entrants, substitutes, customer bargaining power, supplier bargaining power. Add the MILLER organizational maturity model, which you also use to time reorganizations. Record a finding or 'no impact' for every element so gaps are visible.
4. 4Work the rest of the 10 critical inputs as a checklist and decide for each whether it belongs in the plan. Examples: Competence/Performance/Reliability (task, individual, team); Alignment (culture, job retention, job security); Actions (training, education, development); Surveys (employee, managers); outsourcing and budgets. Mark the six inputs where workforce planning contributes directly (1, 2, 4, 5, 6, 7) and give them predictive data, not retrospective reports.
5. 5Bring the HR draft to the strategic mix. Each business partner brings PEST, Five Forces and its own function-specific requirements. In consultative meetings, adjust the HR plan until it reflects the total picture. Arrive with proposals, not an empty-handed 'what do you need from us'.
6. 6Plan on several rework cycles. The source calls repeated rework of the HR strategy 'inevitable' before it integrates into the overall business strategy. Log what changed in each cycle and why.
7. 7Submit the blended plan for strategic approval. Start execution only after approval.
8. 8Convert each approved HR strategic action into an action plan with resource and cost, using the 6 S model. Roll the plans into one HR business plan with short, medium and long-term horizons. Load them into a single project tool (the source names Microsoft Project) so HR runs inside the business plan.

The source describes a common failure, which this example works through. A business plan expands, and line managers leave recruitment too late, so workload outruns headcount and hiring becomes an urgent, poor-quality scramble. Under the map, HR does not wait for the request. Before the strategic mix, it runs PEST, Five Forces and MILLER (Input 1) and prepares the workforce inputs the source lists: forecast budgets (salary, training, allowances, bonus) and trend/prediction data on turnover and organizational stability. In the consultative meeting, the business partner states the expansion requirement. HR's draft is already built on predictive information and is adjusted to the total picture, so HR brings proposals instead of arriving with 'nothing to offer'. After approval, the recruitment action is turned into an action plan with resource and cost via the 6 S model, and loaded into the integrated HR business plan with its short, medium and long-term phases. The late-panic hire becomes a scheduled, costed plan. This is an illustrative walk-through of the source's own sequence; the passages give no company figures.

• If an input is a retrospective report with no forward implication, then drop or convert it into a prediction with an explanation and advice. The source ties strategic value to predictive data.
• If a PEST, Five Forces or MILLER element shows no HR impact, then record 'no impact' explicitly rather than leaving it blank, so reviewers can see it was checked.
• If a business partner's requirement changes HR's draft in the mix, then rework the plan and re-run the affected inputs. Expect several cycles, so do not treat a rework as a failure.
• If a strategic action has no resource and cost attached, then it is not ready for approval. Translate every action into an action plan first.
• If a change in the organization needs triggering, then HR must say so. Use the MILLER model to time reorganizations and raise it at the strategic table rather than waiting for a request.

• Treating the HR strategy as self-contained. The tell: the plan was never taken to the strategic mix, or no business partner's PEST/Five Forces/requirements appear in it.
• Arriving at the strategic table with no ideas. The tell: HR waits for the business to dictate requirements and has no pre-built proposals.
• Filling the 10 inputs with retrospective data. The tell: monthly and quarterly data-heavy reports that no one can use for future plans, the 'data farming' the source criticizes.
• Skipping the future timeline. The tell: the plan consists of past performance and current fixes, with no short/medium/long-term view or emergency, growth and contraction scenarios.
• Stopping at approval. The tell: approved strategy has no action plans with resource and cost, and is not integrated into one plan, so execution runs in separate threads.

• ▸Timeline split: list items under retrospective, current and future; flag the future items that carry over past or current issues.
• ▸Input 1 findings: one line each for PEST (P, E, S, T), the five Five Forces, and the MILLER maturity position/reorganization trigger.
• ▸Inputs 2–10 checklist: for each, write 'include in plan' or 'no impact' and the reason; mark the six where workforce planning contributes directly (1, 2, 4, 5, 6, 7).
• ▸Predictive evidence: for each included input, state the forecast or trend and the explanation/advice that goes with it.
• ▸Business partner requirements: record each partner's PEST/Five Forces and function-specific requirement heard in the mix.
• ▸Alignment changes: note what HR changed in the draft after each consultative meeting, with a rework-cycle number.
• ▸Approval status: record the approval date and approver for the blended plan.
• ▸Action plan per HR action: list the action, its resource and its cost under the 6 S model.
• ▸Horizon placement: short, medium or long term, and the project-plan entry it is loaded into.

1. 1Scope the role first, then cap the list. Write every behavior the job touches, then cut to the ones that make the difference. The hard limit is eight competencies per role, and six is the practical average. If you cannot cut below eight, you are listing tasks, not key competencies.
2. 2Write each competency as a key observable behavior, something a manager can see, measure and improve. A competency is built from skills, knowledge and experience. Reject any wording that describes a trait or intention nobody can observe.
3. 3Build each competency as a unit with smaller elements beneath it, as in the Team Leader illustration. Rate and train at the unit level. Leave the smallest elements unrated, because their training happens on the job.
4. 4Set three standards per competency: minimum, average and maximum. Each is a behavior description anchored to a number on a linear 0–100 scale. The minimum is the floor for safe, legal, conforming work, which are the quality-assurance outcomes a framework exists to deliver. Fix the numbers before the first appraisal, not after.
5. 5Collect scores as exact 0–100 numbers, not tick-boxes. Four boxes (0–25, 26–50, 51–75, 76–100) are easy for managers but cannot support modeling or forecasting. Tie collection to the annual appraisal cycle, and collect competence, performance and reliability together.
6. 6Define the measurement requirements (terms of reference) alongside the competencies. Pair each competency with the evidence that will show it: for example total calls vs. total sales, letters of praise, referrals, a competency matrix.
7. 7Turn gaps into training, not a form. Compare each person's score with the role standards, aggregate by department, and train against the unit-level gaps. The competence requirement is already scoped, approved and funded at corporate level, so a gap is an organizational training need, not only an individual one.
8. 8Pilot before rollout. Run the training and testing in a remedial mode with existing staff, then revise from their feedback. Apply the Drucker test to everything you add: would the roof cave in if you stopped doing this? Keep it only if it is fast, accurate, cheap and easy (FACE).

The source describes a service-center project for loan counselors, loan processors and their leaders. The core skills are six: Point software, Desktop Originator software, Loan Prospector software, telephone usage and predictive language, accuracy of documentation, and product knowledge. That is exactly the six-competency average, and each is observable. The terms of reference are set alongside: total calls vs. total sales, letters of praise, a phone matrix, a competency matrix and referrals. The stated goal is that at the end of training, counselors, processors and their managers reach a basic designated performance level based on the terms of reference and core skills. To apply the framework here, you would score each person 0–100 on each of the six skills, with minimum, average and maximum behavior descriptions fixed beforehand. Lesson plans and a testing program follow, the program is tested in remedial mode on existing staff, and it is revised from their feedback. Department-level score patterns then show which skill (say, documentation accuracy) needs the training budget.

• If a role's list exceeds eight competencies, then cut. Keep only behaviors that make the difference to the job and merge or drop the rest.
• If a proposed competency cannot be observed and scored by a manager, then rewrite it as a behavior or remove it.
• If an appraisal form uses tick-box bands, then replace it with an exact 0–100 score, because the bands are too coarse for modeling and forecasting.
• If a person scores below the minimum standard on a competency, then treat it as a training-needs item at unit level and check the standard has been funded and scoped at corporate level before blaming the individual.
• If a process or form added to the framework fails the FACE test or the Drucker 'roof cave in' question, then abolish or simplify it.

• Allocating every task a competency. The tell: lists of twelve-plus items per role and managers who say measurement is unattainable.
• Importing a complex, book-style model that adds confusion. The tell: managers cannot explain the framework and business users doubt its credibility.
• Tick-box scoring. The tell: everyone lands in 'Average' and trend lines or forecasts cannot be built from the data.
• Describing competencies as attitudes or traits rather than behaviors. The tell: two managers rate the same person very differently with no evidence either can point to.
• Scoring competence only. The tell: no performance or reliability data collected alongside it, so you cannot tell whether capable staff deliver consistently.

• ▸Role title: name the single role this card covers.
• ▸Competency (key observable behavior): one line, six on average, eight maximum.
• ▸Competency units: list the units beneath it that you will rate and train; leave the smallest elements to on-the-job training.
• ▸Minimum standard (0–100 score and behavior): the floor for safe, legal, conforming work.
• ▸Average standard (0–100 score and behavior): what typical competent performance looks like.
• ▸Maximum standard (0–100 score and behavior): what top performance looks like.
• ▸Evidence / terms of reference: the measure that confirms the score (e.g., calls vs. sales, letters of praise, referrals, competency matrix).
• ▸Current score (0–100): exact number from the appraisal, never a tick-box band.
• ▸Training action and owner: the unit-level training that closes the gap to standard, and who delivers it.

1. 1Write the performance target as an exact number with a calendar date before you buy any training. The source rule is to use exact numbers, not percentages, and to date by day, month and year. 'Process 600 extra files by 1 October 2016' is a performance target. 'Improve efficiency' is not.
2. 2Find where output is lost before you choose who to train. Use the 'work done' approach: categorize poor performers, then measure and cost their lost performance. Do this by department or team, because the source reads competency strengths and weaknesses by department and projects team productivity gains.
3. 3Separate competence from performance in your data. The source tracks 'speed to competency' and 'speed to performance' as different measures, and training efficiency 'relating to productivity' as different from training efficiency 'relating to competencies'. Record both for each unit. A unit can be fully competent and still under-produce.
4. 4Test whether training moves output. The source lists Training/productivity among the correlations to forecast. Put training hours or spend per unit next to that unit's output over time. Fund the units where the relationship is visible and the lost-performance cost is largest.
5. 5Set the training budget as a forward line with an ROI attached. The source asks for 'training budget required for future' and 'ROI on training'. Cost it alongside recruitment, turnover, lost time and reliability, so training is compared with the other ways of closing the same gap.
6. 6Keep the training short and aimed at the person who controls output. For the Triple W objective method, the source says managers and supervisors need only 'a short but focused piece of training'. Teach them to write the What, Who and When of an objective. Do not script the how. People who work out the method themselves are more committed and accountable.
7. 7Document the investment as a strategic action plan: what is to be achieved, what is covered, how and when each item is delivered, risks, and dependencies. Use the standard format and get approval before and after delivery.
8. 8Report results on the productivity dashboard at unit and team level. Check delivered output against the dated milestones, not against attendance or course completion.

The source's own objective example is a unit that must show 600 extra files. The plan needs outline approval by 2 April 2015 and final approval by 1 May 2015. The first batch of 50 extra files is due by 1 October 2015, and all 600 by 1 October 2016. The training investment is not a course on files. It is the short, focused training given to that unit's managers and supervisors so they can write and hold objectives in exactly this form: exact numbers, day-month-year dates, and no prescribed method. The unit then works out how to hit the numbers, which the source says raises commitment and accountability. Success is judged on the milestones: 50 files by 1 October 2015, then 600 by 1 October 2016. Course completion does not count. The source does not give a cost or ROI figure for this example, so you would add the training cost and the costed lost performance from your own 'work done' data to complete the ROI line.

• If a unit's lost performance, costed from the work-done categorization, is small, then do not fund performance training there. Put the money where the costed gap is largest.
• If staff are competent but output is flat, then the gap is performance, not competence. Fund objective-setting and management training, not more skills courses.
• If you cannot state the target as an exact number with a day-month-year date, then do not approve the spend. Rewrite the objective first.
• If the training/productivity correlation for a unit is absent after a reasonable period, then stop funding that training for that unit and re-examine the cause of the gap.
• If the training is for managers or supervisors, then keep it short and focused. Check for well-written objectives afterward, because managers saying they can write them is not evidence.

• Measuring compliance instead of output. The tell: your training report lists attendance and certificates, and no unit output figure.
• Spreading the budget evenly across the organization. The tell: no department-level lost-performance costing exists, and every unit gets the same hours.
• Setting targets in percentages or loose time frames. The tell: the objective reads 'improve by a good margin soon', and nobody can say if it was met.
• Prescribing the method in the objective. The tell: staff comply but own nothing, and accountability for the result is vague.
• Treating competence and performance as the same. The tell: speed-to-competency improves while speed-to-performance and output do not move.

• ▸Unit/department: name the single team or department this investment targets.
• ▸Output measure and target: state the exact number and the day-month-year deadline, with no percentages.
• ▸Interim milestone: give the first dated checkpoint and its exact number.
• ▸Costed lost performance: enter the cost of poor-performer output loss from the 'work done' categorization.
• ▸Competency vs performance gap: record speed to competency and speed to performance, and note which one is lagging.
• ▸Training description and audience: say who is trained (for example managers and supervisors) and keep it short and focused.
• ▸Training cost: include the direct budget line plus related costs of lost time and turnover.
• ▸Training/productivity correlation: note the observed relationship between training and output for this unit.
• ▸Plan approvals and dashboard check: record the outline approval date, final approval date, and the review dates against the productivity dashboard.