People Analytics For Dummies

1. 1Define alignment operationally before you measure it. Use the passage's three parts: the employee knows what they are expected to accomplish, under what conditions, and how they are performing against those expectations. Write these as three checks for each role so you can tell which part is broken later.
2. 2Field the two alignment items from the CAMS survey as written. Team: 'There is a clear objective around which myself and the people I work with rally.' Individual: 'I have a clear understanding of the difference between an average contribution and a great contribution for my role.' Do not reword them. The individual item tests whether the evaluation standard is visible, not just whether a goal exists.
3. 3Use the 0–10 agreement scale, which allows 11 possible responses (0 through 10). Keep both items positively worded so 0 is always worst and 10 always best. If you switch to a 1–5 or 1–7 scale, or add or remove items, recompute the index logic to match.
4. 4Embed the two items among all eight CAMS items (capability, alignment, motivation, support) but hide the category labels from respondents. Keep the categories in the background so you can report Alignment alone, the team and individual formats separately, or the overall index across all eight items.
5. 5Report team-format and individual-format alignment as separate lines, then segment by the cuts executives care about, such as division, job function, management level, or tenure. A single company average hides the pocket where expectations are unclear.
6. 6Read alignment alongside the other three CAMS components, not alone. The source's logic is that capability, alignment, motivation, and support each gate performance. Low alignment alongside high scores elsewhere points to expectations as the bottleneck, so fix that first.
7. 7Pull the three levers the source names, matched to where the gap sits: goal setting (low team item), performance appraisal (low individual item), and regular executive, manager, and employee communication (both low, or a gap between levels).
8. 8Check for conflicting objectives between teams or peers. The source flags competing objectives that punish or fail to reward people for doing the right thing for the company. This is the usual reason a team item is low even where each team has a goal. Training investments are unproductive if influences aren't aligned.
9. 9Test whether the actions you took changed anything. The source's analogous test for pay is to compare highest-performer attrition against average and lowest performers to see whether the performance-based differential does what it was meant to do. Re-survey after goal-setting or appraisal changes and compare segment scores.

The passages give no company data for alignment, so this walkthrough uses the source's own instruments and logic. You run the 8-item CAMS survey on a 0–10 scale. The two alignment items are embedded among the others with category labels hidden. When results arrive, you split Alignment into its two formats. Suppose the team item ('a clear objective around which myself and the people I work with rally') scores well but the individual item ('difference between an average contribution and a great contribution') scores poorly in one job function. That pattern says goals exist but the evaluation standard isn't visible, so the lever is performance appraisal and manager communication, not more goal-setting. If instead the team item is low, you look for the source's other cause: competing objectives between teams that punish people for doing the right thing. Before you act, you check the other three CAMS components for that function. Training spend, for example, is wasted if influences are misaligned. After the fix, you re-run the survey and compare the same segments. That mirrors the source's pay example: don't just assume the action worked, test whether the outcome moved.

• If the team alignment item is low but the individual item is high, then look for conflicting objectives between teams or leaders and use goal setting and executive-to-manager communication. People know their own bar but not the shared target.
• If the individual alignment item is low but the team item is high, then fix performance appraisal and manager-employee communication. The goal is clear but the 'average versus great' standard is not.
• If alignment is low and the other CAMS components are also low, then start with alignment and check whether it is driving the rest. The source says none of the four matters if expectations aren't understood and agreed on.
• If alignment is high but performance is still weak, then look at capability, motivation, and support instead of adding more goal-setting. Each factor can independently block performance.
• If you want to change the scale or the number of items, then do so deliberately and rebuild the index, because the proposed calculation assumes 8 positively worded items on 0–10.

• Treating alignment as 'we have goals.' The tell: every team has an objective on paper, yet the individual item ('average versus great contribution') scores low.
• Printing the category labels (Alignment, Capability, and so on) on the survey. The tell: respondents answer to the label rather than the statement, and the survey feels like an HR diagnostic. The source says keep categories in the background.
• Reporting only the company-wide average. The tell: the overall score looks acceptable while one division, function, or level is confused. Segment it.
• Reverse-wording items or mixing scales so the index can't be read consistently. The tell: a 0 means 'worst' on some items and 'best' on others, and the combined score makes no sense.
• Fixing alignment by adding support or training while team objectives still compete. The tell: spend rises but the team item doesn't move, because peers are still penalized for doing the right thing for the company.

• ▸Segment: name the division, function, level, or tenure group being read.
• ▸Team item score (0–10): average response to 'There is a clear objective around which myself and the people I work with rally.'
• ▸Individual item score (0–10): average response to 'I have a clear understanding of the difference between an average contribution and a great contribution for my role.'
• ▸Gap read: note which format is lower and what the pattern implies (goals vs evaluation standard).
• ▸Other CAMS components for this segment: record Capability, Motivation, and Support scores to see whether alignment is the binding constraint.
• ▸Competing objectives check: list any team or peer objectives that punish or fail to reward doing the right thing for the company.
• ▸Lever chosen: goal setting, performance appraisal, or executive/manager/employee communication, plus one specific action.
• ▸Owner and date: who acts and when.
• ▸Re-measure result: same segment, same two items, scores after the action and whether the gap closed.

1. 1Define motivation operationally before you measure it: use the passage's definition of engagement, an index of survey questions measuring how committed employees are to the company and how motivated they are to apply discretionary effort. Write down which items make up your index so the number means the same thing each time you report it.
2. 2Cut the engagement index by business-relevant segments: division, job function, performance rating, tenure, region, and any segment your executives care about. A company-wide average hides the pockets where people are only 'standing by and collecting a paycheck.'
3. 3Add two simple survey questions as cross-checks: likelihood to pursue a new job in the next 12 months, and likelihood to recommend the company as an employer. The source reports these correlate with actual outcomes and business performance.
4. 4Pair the motivation data with voluntary attrition rate by the same segments. Low motivation in a segment that also loses people (for example, a large share of key-job employees being poached by one competitor) is a stronger signal than either number alone.
5. 5Benchmark where you can. In the pharmaceutical company's survey, roughly 100 items across about 50 aspects of the employee experience were compared against high-performing companies. Read the categories where you trail, not just the headline motivation score.
6. 6Diagnose obstacles to motivation by asking directly: is it a bad manager, missing tools and resources, arguments among employees, or a lack of competitive pay and benefits? Match each hypothesis to a data source (manager-level survey cuts, resource items, pay data) and test one at a time.
7. 7Frame the motivation problem for executives with the ABC logic: Antecedents start the process, Behaviors are the effort or change required, and Consequences drive the buy-in (the source puts C at roughly 80 percent of motivation in the author's experience). Make the consequence large enough to pull people through the hard work of B.
8. 8Check whether the actions you take are working: segment attrition by performance rating and see whether your highest performers leave at a lower rate than average or lowest performers. If not, the pay-for-performance lever is not producing the commitment it was designed for.

The pharmaceutical company in the passages joined a confidential cross-company survey of about 100 items covering roughly 50 aspects of the employee experience, benchmarked against the highest-performing companies in all industries. Motivation and commitment were measured alongside leadership, managers, pay, benefits, and learning opportunities. Employees were on average more committed, more motivated, and happier than the benchmark, and this was statistically validated. The one weak spot was the Speaking Up category: employees felt slightly less safe expressing concerns or disagreements to superiors, which clashed with the company's reputation for consensus decisions. Because everything else scored well, no substantive action was decided. The lesson for you: a strong motivation score does not clear the weak category beneath it. Treat a below-benchmark item as a hypothesis to investigate even when the headline is green, and decide explicitly whether to act, rather than letting the good average absorb it.

• If engagement or motivation looks fine overall, then segment by division, function, performance rating, tenure, and region before you conclude anything; the problem usually sits in a segment.
• If motivation is low and attrition is high in the same segment, then prioritize that segment first; if only motivation is low, investigate obstacles (manager, tools, pay, conflict) before assuming a retention issue.
• If you must choose where to spend scarce resources, then focus on key jobs where attrition or disengagement hurts the business most, as the pet retailer did, rather than spreading budget thinly across all jobs.
• If you are persuading an executive to act on motivation data, then identify whether they respond to gain or fear, and to feelings, time and money, or the need to lead, and lead with the consequence that matters to them.
• If your highest performers leave at the same or a higher rate than others, then the performance-based pay differential is not working as intended; revisit it before adding other programs.

• Reading the average and stopping: the overall motivation score is strong, so nobody looks deeper. The tell: a weak sub-category (like Speaking Up) is explained away as 'odd' and no action is decided.
• Treating motivation as a pure attitude score and ignoring behavior and consequences. The tell: you can report the index but cannot name what changes in effort or outcomes would follow.
• Spreading the people budget evenly across all jobs. The tell: the same spend everywhere and no measurable effect in the jobs that matter most to the business.
• Pitching an abstract case to executives. The tell: they agree it is sound but do not invest, because they cannot see how their results or standing improve; the consequence (C) was never made large or specific.
• Assuming you know why people disengage. The tell: you launch a fix (pay, perks) without testing alternatives like a bad manager, missing tools, or conflict among employees.

• ▸Segment: name the division, function, performance rating, tenure band, or region being examined.
• ▸Engagement index score: record the segment's score and the survey items that compose it.
• ▸Intent-to-leave (next 12 months): record the percent likely to pursue a new job.
• ▸Likelihood to recommend as employer: record the segment's score.
• ▸Voluntary attrition rate: record the segment's rate and note whether a single competitor is poaching.
• ▸Benchmark gap: list any survey categories (such as Speaking Up) scoring below comparison companies.
• ▸Suspected obstacle: pick one of bad manager, missing tools/resources, employee conflict, or uncompetitive pay and benefits, and name the data that would test it.
• ▸Executive consequence (C): state the consequence of acting or not acting that matters to the decision-maker.
• ▸Check on action: compare attrition of highest performers against average and lowest performers to see whether the current lever is working.

1. 1Define Support as three measurable strands before you write any items: technical (resources, tools, information), managerial, and peer (cooperation from others). Write at least one survey item per strand so a low score points to a fix. The source's Appendix B (sample survey questions, standard convention, organized by topic) is the place to start sourcing item wording.
2. 2Collect Support as a scored survey measure, not as comments. Put the items in the same survey as your outcome measures (engagement, Intent to Stay, Team Effectiveness). The source's CAMS survey setup (Tables 7-1 and 10-1) and the item-level analysis in Chapter 13 both assume the survey carries these outcome items alongside the driver items.
3. 3Roll the Support items into one score per person. The source's example scatterplot compares Person A at Support = 20 with Person B at Support = 58, so build the score to produce a spread of individual values you can plot.
4. 4Sort every item from least to most favorable (the source's Figure 13-1 and 13-2 pattern: complete list, then bottom 10 and top 10). Mark where the Support items land. Low favorability tells you where people say support is weak, but it does not yet tell you whether that weakness matters.
5. 5Test whether Support matters. Plot Support (x) against an outcome such as engagement (y), one dot per person, as in the Acme Company scatterplot (Figure 13-9). Set the data up in Excel and compute the correlation (Figure 13-12). Then add a least-squares trendline with the regression equation and r-squared (Figures 13-14 to 13-16).
6. 6Rank the Support items by correlation to your outcome KPIs, such as Intent to Stay and Team Effectiveness (Figures 13-3 to 13-5). Then compute item Importance (Figures 13-6 and 13-7) and place each item on the key-driver quadrant (Figure 13-8). Items that are both unfavorable and important go to the top of the fix list.
7. 7When Support must be separated from other influences, run a multiple regression in Excel's Data Analysis Regression tool (Chapter 14). Chapter 14's example uses job quality, company quality, and manager quality alongside engagement. Support stays a candidate driver, not an assumed one.
8. 8Cut the Support score by segment (the Chapter 4 methods, such as tenure group, calculated segments, and cross-tabs) so you can see which groups have the gap. Act on the segment where low Support and high importance coincide, then re-survey to see whether your fix moved the number.

The source's own Support illustration is the Acme Company scatterplot of Support against engagement (Figure 13-9), followed by a two-person comparison: Person A with Support = 20 and Person B with Support = 58 (Figure 13-10). The passages supply only the figure titles, not the engagement scores or the correlation, so treat the following as the method rather than as reported results. Plot every Acme respondent with Support on the horizontal axis and engagement on the vertical axis. Locate Person A far to the left and Person B well to the right of center. If the cloud of dots slopes upward, higher-Support people tend to be more engaged, and the correlation and trendline quantify how strongly. Add the regression equation and r-squared (Figure 13-15) to see how much of the variation in engagement Support accounts for. The scatterplot shows association, not cause. Before you commit budget, check whether Support still ranks as a driver on the Importance calculation (Figure 13-7) and the key-driver quadrant.

• If Support items are unfavorable but show weak correlation or low Importance to your outcome KPIs, don't prioritize them. Fix the items that are both unfavorable and important.
• If the Support-engagement scatterplot shows a clear slope, then test it in a multiple regression with other drivers (such as manager quality) before you attribute the effect to Support alone.
• If the overall Support score looks fine but one segment (tenure group, team, location) scores low, act at the segment level instead of launching a company-wide program.
• If one strand (technical, managerial, or peer) drives the low score, target that strand. A single blended score hides which kind of support is missing.
• If a fix has been in place for a full survey cycle and the Support score and linked outcome have not moved, stop investing and re-examine the item importance. The source's stance is to let go of programs the data shows are not working.

• Reading correlation as cause. The tell: you announce 'raise Support to raise engagement' based on one scatterplot with no regression controls or follow-up measurement.
• Using a single vague Support question. The tell: the score is low but nobody can say whether tools, manager, or peers are the problem.
• Prioritizing by favorability alone. The tell: you fix the bottom-10 items without computing Importance, and outcomes don't change.
• Weak survey data (garbage in, garbage out, one of the source's listed pitfalls). The tell: item wording is ad hoc rather than drawn from a standard-convention library such as Appendix B, so results are inconsistent or uninterpretable.
• Analyzing in isolation. The tell: HR reports a Support gap but the managers and teams who control tools and cooperation were never involved (the source lists 'Not Involving Other People in the Right Ways' as a pitfall), so no one owns the fix.

• ▸Strand: technical, managerial, or peer. Label each Support item with one strand.
• ▸Survey item wording: paste the exact item text, ideally from a standard-convention source such as Appendix B.
• ▸Percent favorable: record favorability and the item's rank from least to most favorable.
• ▸Outcome KPI: name the outcome you test against (engagement, Intent to Stay, Team Effectiveness).
• ▸Correlation to KPI: the Excel correlation between the item (or the Support score) and the outcome.
• ▸Importance score: the item's computed Importance and its rank among all survey items.
• ▸Key-driver quadrant: place the item (e.g. unfavorable and important) to set its priority.
• ▸Segment with the largest gap: the tenure, team, or location group with the lowest Support score.
• ▸Owner and action: the named person responsible for the fix, and the one action they will take.
• ▸Re-survey result: the Support and outcome scores after the next cycle, and keep or stop.

1. 1Split the definition into three measurable parts before you collect anything: belief in the organization's goals, willingness to exert effort on its behalf, and desire to stay. A single 'engagement' number hides which part is weak, so write one or more survey statements for each part.
2. 2Measure commitment with a survey, not a guess. Include commitment next to related attitudes such as overall satisfaction, motivation, leadership, managers, pay, benefits, and learning and development, so you can later test what moves it. The source's example survey used roughly 100 items across about 50 aspects of the employee experience.
3. 3Benchmark where you can. In the source example, the company compared itself with a selection of the highest-performing companies, through a consulting firm that guaranteed confidentiality. Treat a gap in any category, even when the overall picture is strong, as a finding to explain.
4. 4Segment before you average. Cut commitment scores and exit rates by the same segments, such as tenure group, using basic employee facts. For tenure, calculate it from Hire Date in Excel, convert days to years, and add an if-then flag (1 if tenure is under 1 year, otherwise 0). Then compare the segments, as the book's 'Exit Rate % by segment' view does.
5. 5Pair attitude data with behavior. Commitment is an attitude and attrition is the outcome, so look at who stays and who goes. Identify which employee characteristics or conditions make exit more predictable, and test whether the common assumptions about why people leave hold in your data.
6. 6Rank what drives commitment. Use key driver analysis (correlation of survey items with the commitment items) to find which experiences, such as bad managers, missing tools and resources, conflict among employees, or uncompetitive pay and benefits, travel with low commitment. Use the ranking to choose where to act.
7. 7Choose a specific action using the ABC model: define the antecedent (what triggers it), the behavior you want, and the consequence that reinforces it. Act where the data points, and do not add programs to every low score.
8. 8Close the loop. Record the prediction ('if we change X, commitment or exit in segment Y will move'), re-measure, and check whether management's actions worked. Drop activity that did not move the numbers.

The source describes a well-run pharmaceutical company that joined a confidential multi-company employee survey of about 100 items covering roughly 50 aspects of the employee experience. Against other high-performing companies, its employees were on average more committed, more motivated, and happier, and this was statistically validated. One category scored slightly below peers: 'Speaking Up', which measures whether employees feel safe expressing concerns or disagreements to superiors. That looked odd, because the company was known for consensus decision-making and new scientists were told to expect it. Because every other result was strong, no substantive new action was decided, and the head of HR expressed some concern. The lesson for commitment work is that high average commitment can sit alongside a specific weakness. A practitioner would test whether the Speaking Up gap concentrates in particular segments, such as short-tenure scientists, and whether those segments show higher exit. That follow-up is my application of the method and is not stated in the source passage.

• If overall commitment beats the benchmark but one category lags, then investigate that category by segment instead of dismissing it, because strong averages do not explain an anomaly.
• If commitment is low in one tenure or organizational segment but fine company-wide, then act on that segment only and leave company-wide programs alone.
• If you cannot say which of the three components (belief in goals, effort, desire to stay) is weak, then redesign the survey items before choosing any action.
• If a proposed retention action rests on a 'best practice' from other companies, then check it against your own data on who stays and who leaves before funding it.
• If exit data and commitment scores disagree for a segment, then test your assumptions about why people leave before trusting either number.

• Treating one high average score as proof the culture is healthy. The tell: a below-benchmark category such as Speaking Up is explained away and no follow-up is assigned.
• Copying other companies' retention programs as best practices. The tell: you cannot name the data from your own workforce that justifies the program.
• Measuring only attrition and calling it commitment. The tell: your dashboard has exit rates but no attitude items on belief in goals, effort, or intent to stay.
• Reporting only company-wide figures. The tell: no tenure or other segment cut exists, so a high exit rate among under-1-year employees stays invisible.
• Adding actions without measuring them or removing old ones. The tell: the HR calendar grows each year and no one can say which activity moved commitment.

• ▸Segment: name the group (for example, tenure under 1 year) and how it is defined in your data.
• ▸Belief-in-goals score: survey result for items on the company's goals and prospects, and the benchmark if you have one.
• ▸Effort-willingness score: survey result for items on willingness to exert effort, and the benchmark if you have one.
• ▸Desire-to-stay score: survey result for items on intent to remain, and the benchmark if you have one.
• ▸Exit rate for the period: leavers in this segment divided by headcount, to compare against the other segments.
• ▸Lowest-scoring survey category: record any category below benchmark (for example Speaking Up) and its gap.
• ▸Top suspected drivers: rank the experiences most correlated with the commitment items (manager, tools and resources, pay and benefits, conflict among employees).
• ▸Planned action (antecedent, behavior, consequence): one sentence for each part of the ABC model.
• ▸Prediction and re-measure date: state the expected change in score or exit rate, and when you will check it.

1. 1Anchor attraction to the employee journey, which starts the day a person becomes aware of the job opportunity. Name the stages from awareness to start date, because outcomes are measured at the end of that chain and diagnosed along it.
2. 2Write one definition for each outcome before pulling data: volume (hires and hire rate, growth rate), speed (time-to-fill, time-to-start), cost, and quality. State the start event and stop event for every time measure. The source treats time-to-fill and time-to-start as separate calculations, so keep them separate.
3. 3Pull the data from the systems that already hold it: the applicant tracking system (ATS), HRIS, payroll, and job boards. Use market pay benchmark surveys and government Census and Department of Labor data as external context. Standardize definitions against a metric-definitions convention so the numbers compare across teams.
4. 4Build the funnel view. The source's 'tale of two funnels' and 'talent acquisition efficiency' figures imply comparing funnels, for example by recruiter, source, or job family. Record candidates at each stage and the conversion between stages. Look for the stage where one funnel loses more than the other.
5. 5Measure quality with a defined standard, not a gut call. The source points to a Behaviorally Anchored Rating Scale (BARS), with a Listening-skill example, and to the Critical Incident Technique. Use job analysis to identify the behaviors that separate good from poor performance, then rate hires against anchored descriptions.
6. 6Segment every outcome by job or job family. The pet retailer found that jobs requiring more distinct pet knowledge mattered more to results, so a blended average hides where hiring matters most.
7. 7Translate outcomes into value using Employee Lifetime Value (ELV). Compare the ELV change from hiring 50 people (the source's software-engineer figure) and test more than one lever: faster, cheaper, or better hires. Prioritize the lever that moves ELV most for the segment that matters.
8. 8Peel the onion before acting on any gap. Ask: how do you know this is a problem, what measurements show it, what is the situation right now? Record a baseline so you can later tell whether your fix worked.

The passages give the structure of the attraction analysis but not Chapter 9's numbers, so this example uses the pet retailer's reasoning, which is the only complete real case supplied. The retailer found that service employees (groomers, for example) affected customer loyalty and sales more than cashiers or stockers did. It also found that the more pet-specific knowledge a job required, the more attrition in that job hurt. It therefore stopped spreading its people budget evenly across all jobs and concentrated it on the key jobs. Apply the same logic to attraction outcomes. First, rank jobs by business impact. Second, report volume, time-to-fill, time-to-start, cost, and quality separately for the top-ranked jobs, not as one company-wide average. Third, put the extra recruiting effort where a slow or poor hire costs most. The retailer's margins were thin and so is most hiring budgets, which is why prioritizing beats spreading money evenly. Use your own ATS figures to fill in each metric.

• If a time metric changes, then first check that its start and stop events are unchanged. Time-to-fill and time-to-start measure different intervals and can move in opposite directions.
• If one funnel converts worse than another, then compare stage by stage and fix the stage where candidates drop, not the final hire count.
• If hiring volume is hitting target but quality ratings are low, then do not celebrate. Re-check the BARS or critical-incident standard and the sources the poor hires came from.
• If a job has high business impact, as the pet-knowledge jobs did, then give it more recruiting investment and a separate dashboard instead of a blended average.
• If you are asked to improve an outcome, then estimate the ELV effect of the proposed change (speed, cost, or quality) before committing resources, since there is more than one way to raise ELV.

• Reporting only volume and speed. The tell: hire counts look strong, yet nobody can say how good the hires are because no quality scale exists.
• Using inconsistent metric definitions across recruiters or systems. The tell: two teams report different time-to-fill for the same requisition.
• Using company-wide averages. The tell: the overall number looks stable while a critical job family is stalled or understaffed.
• Collecting process metrics without a business question. The tell: lots of activity and dashboards but no clear chain of logic to a business outcome, and no baseline to measure progress against.
• Treating attraction as separate from the rest of the journey. The tell: hires arrive quickly but activation or attrition problems follow, so hiring gains do not show up in lifetime value.

• ▸Job family / segment: name the segment and why it matters to business results.
• ▸Volume: hires in the period, hire rate, and growth rate, each with its written definition.
• ▸Speed: time-to-fill and time-to-start, each with its start event and stop event.
• ▸Cost: total acquisition cost for the segment, with the cost components included.
• ▸Quality measure and standard: the BARS or critical-incident criteria used, and the average rating of new hires.
• ▸Funnel conversion by stage: candidates entering and leaving each stage; mark the weakest stage.
• ▸Data sources: ATS, HRIS, payroll, job boards, or external benchmarks used for each number.
• ▸ELV impact estimate: expected ELV change if the weakest outcome improves, and which lever you would pull.
• ▸Baseline and next review date: today's values and when you will re-measure to test whether the fix worked.

1. 1Segment by job before you measure anything. The source says a good chef is not a good host or waiter, and a good accountant is not a good manager. Mixing jobs produces no clear answer, so each job family gets its own definition of behavioral performance.
2. 2Run or refresh job analysis first (the passages point to Chapter 13 and Appendix C). It shows how jobs differ and gives you the segmentation breaks. If you have no job analysis, stop here and do it.
3. 3For each job segment, write down what good and bad look like in observable terms. Cover task proficiency, productivity, work quality, and citizenship behavior. The source asks you to be specific: 'How do you define better? How do you measure it?'
4. 4Anchor the behaviors. The book's Chapter 9 figure list includes a Behaviorally Anchored Rating Scale (the example is the Listening skill) and a Critical Incident Technique workflow. Use these to turn a vague trait into described behaviors at defined levels. The supplied passages show only the titles, so take the method from the chapter itself.
5. 5Pull the data you already hold. The source lists performance management systems, HRIS, payroll, LMS, employee surveys, and email and calendar data. Map each behavior to a source and note where no record exists.
6. 6Connect people to performance within each segment. Ask how people with different characteristics perform in that job, then encourage the characteristics that support success. This is the source's own logic: understand the jobs, understand the people, and link the two.
7. 7Test changes on a subset. Pick a small set of units for the change and another set as comparison, then compare them with simple math on the metrics you already track. Expand, modify, or abandon based on the result.
8. 8Report results as numbers tied to business outcomes. The source says numbers keep the conversation anchored to what is real and drive accountability for real results.

The source's pet retailer wanted to know whether adding pet services such as grooming would raise performance. The services needed new training and more skilled, higher-paid employees, so a full rollout risked bankruptcy. The company introduced services in a small set of stores and kept another set without them as a comparison. Using the same data and metrics it already had, it compared locations on customer store visits, sales, and loyalty. The services stores showed more visits, higher overall customer spending, and greater loyalty. Customers who got Fido groomed also spent more on Fido over his whole lifetime, which was the less obvious finding. The new higher-paying service roles also created a job ladder, so the jobs read as a career and not a short-term fix. The lesson for behavioral performance is to define the new job's behaviors, pilot them in some units, and compare against units that did not change. The passages do not describe how individual employee behavior was rated in this case.

• If you cannot say what good looks like for a specific job, then do job analysis before building any performance metric.
• If a metric blends several job types, then split it by job segment, because averages across different jobs hide the answer.
• If a change would be costly or risky to roll out company-wide, then pilot it in a few units with a comparison group first.
• If a behavior is only described by a trait word such as 'good listener', then anchor it with observable behavioral descriptions (BARS or critical incidents) before scoring.
• If you already hold the data in an existing system, then compare groups using those metrics before you start new data collection.

• One performance definition for every job. The tell: the same scorecard is used for chefs, hosts, and managers, and the results do not separate good from bad.
• Trying to be best at everything. The tell: your leaders say they hire 'better than average' people but cannot say how they define or measure better.
• Deciding by gut, copying competitors, or tradition. The tell: the justification for a performance practice is 'that's how it's always been done' or 'others do it'.
• Full rollout with no comparison group. The tell: results improve or worsen after a change, but nothing shows whether the change caused it.
• Ignoring data you already hold. The tell: a new data collection effort starts while performance management, LMS, and HRIS data go unused.

• ▸Job segment: name the job family this sheet covers, based on your job analysis.
• ▸Source of job analysis: cite the job analysis (date, owner) that justifies this segment.
• ▸Task proficiency behaviors: list observable actions showing the core tasks are done well.
• ▸Productivity and quality indicators: list the measures you can count, and where each lives (HRIS, performance system, etc.).
• ▸Citizenship behaviors: list observable actions that help the team or company beyond the formal job.
• ▸Behavioral anchors: for each behavior, write what poor, acceptable, and strong look like (BARS) or log critical incidents.
• ▸Person characteristics to test: list the traits or backgrounds you will compare against performance within this segment.
• ▸Pilot and comparison groups: name the units receiving a change and the units kept as comparison.
• ▸Decision after comparison: state in advance what result means expand, modify, or abandon.

1. 1Define the rate before you pull any data. Attrition is exits in a period divided by headcount, expressed as a percentage (the passages use 'exit rate' and 'attrition rate' for the same idea, as in 'a given year'). Fix the period (12 months), fix the headcount basis, and keep both identical across every cut you make.
2. 2Split every exit by type: voluntary or involuntary. Then split by value: regrettable or non-regrettable. The definition is built on these two axes, and a single blended rate hides which kind of talent loss you are measuring.
3. 3Segment the rate using basic employee facts. Chapter 4 shows 'Exit Rate % by segment' (Figure 4-2) as the pattern. Cut by the segments the company already uses, then keep slicing until a segment stops being informative.
4. 4Create a calculated segment for tenure. Per Figures 4-5 to 4-8, calculate tenure from Hire Date in Excel, convert days to years, then add an if-then statement that flags tenure under 1 year as 1 and everyone else as 0. You can then compute first-year attrition separately.
5. 5Benchmark each segment against the company-wide rate by dividing the segment rate by the overall rate. The hospital's 25 percent first-year nurse attrition against 10 percent overall gives 2.5 times worse than average. A ratio makes the problem size comparable across segments.
6. 6Cross-tab to find the driver. Compare the high-attrition segment against applicant and employee history data (for example, prior experience at hire) to see which hiring characteristic separates stayers from leavers.
7. 7Pair attrition with commitment and exit-reason evidence. Use a KDA-style survey with attrition (or Intent to Stay or a Commitment Index) as the KPI (y variable) and the other survey items as x variables. Treat exit surveys cautiously, since Table 11-1 exists to address their problems.
8. 8Act on the finding, then re-measure the same segment on the same definition. For the hospital, the action was hiring more experienced nurses instead of new graduates. Check the early-tenure rate afterward to see if it moved.

A children's hospital knew that nurses in their first year left at a 25 percent rate, so 1 in 4 hires exited within a year. The average attrition rate for all employees was 10 percent, or 1 in 10. Dividing 25 by 10 shows early-tenure nurse attrition was 2.5 times worse than average. The problem fed itself: each exit needed a replacement who had to be identified, hired, onboarded, and trained, and that replacement had a fairly high chance of leaving just as quickly. The cost went beyond money, because new nurses are less familiar with complex situations and more likely to make mistakes than experienced nurses. The analysts looked at applicant and employee history data and found that hiring more experienced nurses, rather than nurses straight out of nursing school, produced nurses more likely to stay. The conclusion looked obvious in hindsight, but the hospital did not know how much it would matter operationally until the data showed it.

• If the company-wide rate looks acceptable, segment it before you conclude anything. The hospital's 10 percent average hid a 25 percent first-year nurse rate.
• If a segment's rate is a multiple of the average (such as 2.5 times), prioritize it. Check whether replacements keep leaving, because early-tenure attrition reinforces itself.
• If you report a rate, state the type (voluntary or involuntary) and the value (regrettable or non-regrettable). Otherwise the number cannot tell you whether talent is being lost.
• If you need to know why people leave, do not rely on exit surveys alone. Use correlation-based key driver analysis with attrition or intent to stay as the KPI, and review the exit-survey problems in Table 11-1.
• If a survey item shows little value in explaining the KPI, remove it from the next KDA survey and test a new item in its place. Keep the KPI measure itself consistent over time.

• Reporting one blended attrition rate. The tell: leadership says 'attrition is 10 percent' and nobody can name which roles, tenures, or hire sources drive it.
• Counting all exits as talent loss. The tell: involuntary and non-regrettable exits inflate the figure, and the response targets people you were glad to see go.
• Changing the period or headcount basis between segments or years. The tell: segment rates cannot be compared and the trend moves without any change in behavior.
• Assuming the cause is obvious and skipping the data. The tell: the fix is chosen by intuition, and you cannot say how much difference it will make, as the hospital could not until it looked.
• Replacing departed employees without changing the hiring or onboarding inputs. The tell: the same segment's rate stays high because replacements leave at the same pace.

• ▸Period and headcount basis: state the 12-month window and how headcount is counted, and hold both constant.
• ▸Overall attrition rate: exits in the period divided by headcount, as a percentage.
• ▸Exit split by type: count and rate for voluntary versus involuntary.
• ▸Exit split by value: count and rate for regrettable versus non-regrettable.
• ▸Segment and segment rate: the segment (for example tenure under 1 year, by role) and its exit rate.
• ▸Ratio to overall: segment rate divided by overall rate, to show how many times worse or better it is.
• ▸Hypothesized driver: the hiring or employment characteristic you suspect (for example, prior experience at hire).
• ▸Evidence source: the applicant and employee history data, KDA survey, or exit survey used to test it, with known limitations noted.
• ▸Action and re-measure date: the intervention chosen and the date you will recompute the same segment rate.

1. 1Fix the unit of analysis first. ELV covers an average employee in one segment, not the whole company. Choose the segment (for example, software engineers) before you calculate anything. The source stresses that cost of acquisition and value both vary by segment, so a company-wide average hides the differences you need.
2. 2Define the employee lifetime as the period from the day the person first joins to their last day. Compute tenure from hire date and exit date (the source's Excel method: tenure from Hire Date, converted from days to years). Use this same window for both the cost side and the value side.
3. 3Build the cost side for the segment. Include what the company spends to acquire the employee and to keep them over that lifetime. The source's Figure 6-1 for an average software engineer separates cost, value, lifetime value and ELV, so keep these as separate lines.
4. 4Build the value side as the estimated financial value (profit) the average employee in the segment brings over that same lifetime. Subtract cost from value to get ELV. Note that the passages describe the components but give no formula or figures, so document every assumption you make.
5. 5Choose the variant that fits your question: the average employee, the average of a segment, all members of a segment, a past individual (only if you have historical data and have waited long enough), or a forward-looking individual estimate (only if you can classify them into the right segment).
6. 6Calculate ELV for several segments and compare them on a relative basis. The test is the CLV logic: why spend the finite budget chasing segments with less long-term value when the same dollars return more on higher-value segments?
7. 7Trace each segment's ELV to attraction, activation and attrition. Ask which lever (cheaper acquisition, higher activated productivity, longer tenure) moves ELV most. The source shows 'more than one way to increase ELV' and models the gain from 50 additional hires.
8. 8Use the result for time-and-resource decisions: where recruiting effort, retention programs and hiring volume go. Re-run it when segment assumptions change. Treat ELV as a way to shift management from a short-term transactional view to a long-term relationship view.

The passages give the structure of ELV but no dollar figures, so this example uses their real cases and invents no numbers. Figure 6-1 lays out the average software engineer as four items: cost, value, lifetime value and ELV. Figure 6-2 then shows the increase in ELV from 50 hires. To reproduce this, start with the 'software engineer' segment and set the lifetime from hire date to last day. Put acquisition and retention cost against the profit value of the average engineer, and the difference is the segment's ELV. Multiply by 50 hires to see the total ELV added. The pet retailer case supports the segment logic. Its store-level data showed that attrition mattered more in jobs requiring distinct pet-related knowledge than in cashier or stocker jobs. That is evidence that ELV, and the cost of cutting it short through attrition, differs by segment. A recruiter filling rare niche jobs faces a different ELV profile from one filling common roles, which is why the average across segments misleads.

• If you are comparing where to spend recruiting or retention budget, then calculate ELV per segment and rank the segments. Do not rely on one company-wide average.
• If you want a past individual's ELV, then require complete historical data and a finished tenure. If either is missing, use the segment average and label it an estimate.
• If a segment's ELV is low, then test all three levers (attraction cost, activation, attrition) before cutting the segment. The source says there is more than one way to raise ELV.
• If you want a forward-looking ELV for an individual, then first confirm you can classify that person into the right segment. Otherwise, report at segment level only.
• If attrition in a segment hits a role with scarce, specialized knowledge, then weight retention spending there more heavily, as the pet retailer's findings suggest.

• Using one company-wide average. The tell: every segment shows the same ELV, or budget decisions never differ by segment.
• Mixing time windows, such as costs measured to date against value projected over the full tenure. The tell: the lifetime start and end dates differ between the cost and value lines.
• Reading a past individual's ELV from incomplete history or from someone still employed. The tell: the 'lifetime' is shorter than the person's actual tenure.
• Treating ELV as only a recruiting number. The tell: attrition and activation are absent from the analysis, although the source ties ELV to attraction, activation and attrition.
• Presenting a precise ELV with undocumented assumptions. The tell: no one can say how value and cost were defined for the segment.

• ▸Segment: name the employee segment (for example, software engineers) and how it is defined.
• ▸ELV variant: average employee, segment average, all members of the segment, past individual, or forward-looking individual.
• ▸Lifetime window: start (hire date) and end (last day, or projected last day) used for the segment.
• ▸Average tenure: computed from hire and exit dates, in years.
• ▸Cost: acquisition and retention costs over the lifetime, with the source of each.
• ▸Value (profit): estimated financial value the average employee brings over the same lifetime, with the assumptions behind it.
• ▸ELV: value minus cost for the segment, labeled actual or estimated.
• ▸Segment comparison: rank of this segment's ELV against other segments.
• ▸Lever to move: attraction, activation or attrition, with the action chosen and the expected ELV change.

1. 1Name the business indicators before you touch people data. The passages anchor on the numbers a board or investor call tracks: customer store visits, overall customer spending, customer loyalty, sales, and bottom-line performance. Pick two or three as your outcome measures and write down the definition and source of each.
2. 2Reuse the metrics the business already trusts. The pet retailer measured its experiment 'using the same data and metrics they had already been using.' Do not invent new performance measures; pull the existing ones for each location, team, or segment.
3. 3State the people decision as a testable change. Examples from the passages are adding services roles, hiring more skilled and better-paid staff, and building a ladder of jobs. Write it as 'we change X in these units and expect Y to move.'
4. 4Pilot before you roll out. Introduce the change in a small set of units and choose a second set without it as a comparison group. Rolling it out everywhere removes your ability to tell whether it worked and, in the passage's words, 'could bankrupt the company.'
5. 5Compare the two groups on each indicator with simple math: pilot versus comparison on visits, spending, and loyalty. Report the gap per indicator rather than a blended score.
6. 6Extend the time horizon past the immediate transaction. The retailer's less obvious finding was that grooming customers spent more over the whole lifetime of the pet, not just on grooming visits. Check whether your effect shows up beyond the first measured period.
7. 7Decide from the result: expand to more units, modify the program, or abandon it. Then record what the numbers showed so the next decision is accountable to the same indicators.
8. 8Account for the people-side cost and effect in the same view. The services roles required new sourcing, hiring, training, and higher pay, and the job ladder made the company more attractive to employees it wanted long term. Put those costs and benefits beside the business indicators.

A pet retailer wanted to know whether adding pet services such as grooming would draw more people into its stores, just as increased pet knowledge among staff had raised sales. The services required new training, more skilled employees, and higher pay, so the company had to learn to source, hire, train, and pay entirely new types of people. Rolling the idea out to all stores without a period of observation could bankrupt the company, and a full rollout would have left no way to judge whether it worked. Instead it launched services in a small set of stores and kept another set without services as a comparison. Using the metrics it already tracked, it compared the two groups on customer store visits, sales, and loyalty. The services stores showed higher store visits, overall customer spending, and customer loyalty. A less obvious result was that customers who got Fido groomed also spent more on Fido over the pet's whole lifetime, not only on grooming visits. The higher-paying services roles also created a ladder of jobs of increasing skill and pay, which made the company more attractive to the employees it wanted long term.

• If you cannot name the business indicator the people decision should move, then do not start the analysis; define the indicator first.
• If the change is costly or unproven, then pilot it in a small set of units with a comparison set rather than rolling it out everywhere.
• If the pilot beats the comparison group on the agreed indicators, then expand to more units; if it does not, then modify or abandon the program.
• If you have an existing trusted metric for the outcome, then use it instead of building a new one, so the result is credible to executives.
• If the effect may accrue over time, then measure beyond the immediate transaction, as the lifetime pet spending finding shows.

• Launching in every location at once. The tell: no comparison group exists, so you cannot say whether results came from the change.
• Measuring only the immediate effect. The tell: you credit the program only for the grooming-visit sales and miss the lifetime spending effect.
• Ignoring people costs in the performance case. The tell: the proposal shows revenue lift but omits the new hiring, training, and higher pay the roles require.
• Anchoring on instinct, imitation, or tradition rather than the numbers. The tell: the justification is 'competitors do this' or 'we have always done it this way,' with no indicator attached.
• Tracking a single indicator in isolation. The tell: visits rise but nobody checked spending or loyalty, so the result cannot be called better performance.

• ▸People decision tested: describe the change in one sentence (e.g., add higher-skilled, higher-paid services roles).
• ▸Business indicators: list two or three existing metrics (store visits, customer spending, loyalty, sales) and where each comes from.
• ▸Pilot units: name the small set of locations or teams receiving the change.
• ▸Comparison units: name the set without the change that serves as the control.
• ▸Observation period: state how long you will measure and whether it includes effects beyond the immediate transaction.
• ▸People costs: record sourcing, hiring, training, and pay differences the change requires.
• ▸Result by indicator: enter pilot versus comparison for each indicator and the gap.
• ▸Decision: expand, modify, or abandon, with the one-line reason tied to the numbers.
• ▸Follow-up: set the date and indicator you will recheck after the decision.

1. 1Anchor attraction in the employee journey. The journey starts the day a person becomes aware of the job opportunity, so your first map stage is awareness, not application. List every stage from awareness to hire to start, then add the company touchpoints at each stage (the book's Figures 8-2 and 8-3 build the map this way: stages first, touchpoints second).
2. 2Tie attraction to the business problem. Under the triple-A framework attraction is one of three problems, alongside activation and attrition. Write one sentence naming the business outcome a hire is supposed to produce, so you can later judge whether recruiting activity serves it.
3. 3Instrument the process with process metrics before changing anything. The source's Chapter 9 names hire rate, growth rate, time-to-fill, time-to-start, and talent acquisition efficiency. Define each in writing using one standard convention (the HR Metric Definitions Guide, Appendix A, exists for this). Pull each by segment, such as role family, so nothing hides inside an average.
4. 4Compare funnels, not totals. The source's "A tale of two funnels" figure implies setting two funnels side by side, for example two roles, channels, or recruiters. Look for the stage where one funnel loses candidates the other keeps, and fix that stage first.
5. 5Build interviews from job analysis and competency analysis (Appendices C and D). Score with a Behaviorally Anchored Rating Scale (BARS) per competency, as in the source's Listening-skill example. Derive the anchors from real examples gathered with the Critical Incident Technique. Interviewers then rate observed behavior against shared anchors instead of impressions.
6. 6Value hires, not just count them. Use Employee Lifetime Value (ELV) by segment, because the source shows ELV varies by segment and that there is more than one way to raise it. Spend recruiting time and budget where an extra hire or a better-fit hire moves ELV most.
7. 7Close the loop with downstream data. Attraction can produce problems that appear later in activation and attrition (the source's Figure 7-1 covers pitfalls across attraction, activation, and attrition). Survey new hires on their journey and test which attraction-stage variables correlate with later outcomes, using key driver analysis or regression.
8. 8Retire practices you cannot defend. Do not adopt a recruiting practice because a leading company uses it. Keep a practice only if your own data shows it works, and drop the ones that show no effect.

The passages give method and figure titles, not numerical case data, so this example shows the logic rather than invented figures. Take software engineers as one segment. Following the book's ELV figures, start from the average engineer's cost, value, and lifetime value, which together give ELV. Then model the ELV gain from 50 additional hires. You also see that ELV can be raised in more than one way, and that it differs by segment, so you do not treat all roles alike. Next, set two engineering funnels side by side (the 'tale of two funnels' idea) and compute hire rate, time-to-fill, and time-to-start for each. Suppose one funnel stalls at interviewing. You replace unstructured interviews with a BARS for each competency, with anchors drawn from critical incidents. You re-measure the same metrics after the change. Last, you check whether hires from the better funnel show different activation or attrition later. The decision then rests on your own data, not on what a famous employer does.

• If a proposed recruiting practice is justified only by 'other successful companies do it,' then treat it as an untested assumption and run it against your own funnel and outcome data before funding it.
• If you are choosing where to invest recruiting effort, then rank segments by ELV impact rather than by hiring volume or loudest hiring manager.
• If two funnels differ in results, then diagnose the stage where they diverge before changing channels, branding, or offers wholesale.
• If interviewers disagree or rate on gut feel, then rebuild the rating around a BARS tied to a job-analysis competency, with anchors from critical incidents.
• If a recruiting activity shows no link to downstream outcomes and no funnel improvement, then stop it; adding activities without removing any is the 'old HR' overload pattern.

• Copying competitors' attraction tactics (perks, branding, channels) as 'best practices.' The tell: nobody can say what problem the tactic solves in your funnel or whether it is profitable for you.
• Reporting averages across all roles. The tell: one blended time-to-fill or hire rate, with no segment cut, and no one can say which role is the bottleneck.
• Using undefined metrics. The tell: time-to-fill or time-to-start is calculated differently by different recruiters or reports, so trends cannot be compared.
• Optimizing attraction in isolation. The tell: hiring speed and volume improve while activation or attrition worsens, and no one links hire source or process to later outcomes.
• Unanchored interviewing. The tell: interviewers use the same vague terms ('good communicator') with no shared behavioral definition, and ratings do not predict anything afterward.

• ▸Segment: name the role family or job (e.g., software engineer) this sheet covers.
• ▸Business outcome: one sentence on what a successful hire in this segment must produce.
• ▸Journey stages and touchpoints: list stages from job awareness to start date, and the company touchpoint at each.
• ▸Metric definitions: write the exact definition used for hire rate, growth rate, time-to-fill, time-to-start, and efficiency.
• ▸Funnel A vs. Funnel B values: record each metric for two funnels (channel, recruiter, or period) and mark the stage where they diverge.
• ▸Interview design: list the competencies assessed, the source (job or competency analysis), and whether a BARS with critical-incident anchors exists.
• ▸Segment ELV estimate: note cost, value, and lifetime value inputs, and the ELV change expected from the planned hiring.
• ▸Downstream check: record activation and attrition outcomes for hires from each funnel, and any survey findings from new hires.
• ▸Decision and review date: state which practice you keep, change, or stop, based on your data, and when you will re-measure.

1. 1Scope activation as one of the three people problems (attraction, activation, attrition). Treat it as the question 'are the people we hired reaching productive potential?', not as a list of HR programs. Pick one segment first (by role, tenure, or location, per the segmenting chapter) so the analysis stays narrow.
2. 2Build a journey map for that segment. Follow the book's sequence: (1) identify the stages, (2) add the company touchpoints, (3) decide how you will measure each touchpoint, (4) add data you already hold in systems. Onboarding, goal-setting, feedback, training, and tools and resources are your activation touchpoints.
3. 3Inventory every current activation practice and write down what you believe it does. Mark each one you cannot show worked in the past. This is the 'old HR' trap: activities get added every year and none get removed.
4. 4Measure the experience at each touchpoint with a survey, then run key driver analysis to see which items correlate most strongly with business outcomes. Use this to choose where to focus. Do not spread effort evenly across the whole survey.
5. 5Use the ABC lens (antecedents, behaviors, consequences) on each prioritized practice. Name what the practice sets up (the antecedent, such as a goal or a tool), the behavior you expect from it, and the consequence the employee experiences. A practice with no identifiable behavior or consequence is a candidate to cut.
6. 6Score activation and read it alongside employee lifetime value (the book's net activated value idea). This shows where high-value segments are under-activated, so time and resources go there first.
7. 7Write your prediction before you change anything: which metric moves, for which segment, and by roughly how much. After the change, check the prediction against the data.
8. 8Drop or reduce practices that the data does not support. Reallocate the freed time to the drivers that matter most.

The source describes the 'high quality (follow the leader)' company. If a leading company changes its 401k match, goes open-plan, or buys beanbags, ping-pong tables, and free lunches, this company copies it, perhaps with one free lunch a week. The passage names the flaw: you have limited time and money, so you must choose, but you do not know why the other company does what it does. Here is how the activation method replaces that copying. Say new hires in one segment seem slow to get productive. You map their journey, list the touchpoints (onboarding, goals, feedback, training, tools), and survey them on each. Key driver analysis then shows which touchpoint ratings track most closely with the business outcome you care about. You invest in the top driver. You also decide in advance what should improve, and you stop a perk that shows no link to outcomes. The decision now rests on your own data, where before it rested on a competitor's choices. (This is an illustration of the method. The source does not report numbers for it.)

• If a proposed activation program is justified only by 'other successful companies do it,' then require evidence from your own data or a prediction you can test before funding it.
• If survey items are numerous and the budget is limited, then rank them with key driver analysis and act on the top drivers only.
• If you add a new activation practice, then name an existing activity to reduce or remove. Otherwise you repeat the 'old HR' pattern of overcommitment.
• If activation scores are weak in a segment with high lifetime value, then prioritize it over segments with equal scores but lower value.
• If a practice has run for years and nobody can show it works, then treat it as a removal candidate, because the data lets you drop it with confidence.

• Copying other companies' perks and calling it activation. The tell: your justification is 'competitors do this,' not a measured link to your outcomes.
• Adding programs every year without retiring any. The tell: HR is overworked and no single area shows consistent results.
• Running activation work in functional silos (onboarding, training, and feedback each owned separately). The tell: no single measurement connects them to employee and business needs.
• Measuring satisfaction with practices but never tying it to business outcomes. The tell: lots of survey output and no ranked priorities.
• Changing a practice without a stated prediction. The tell: after the change, no one can say whether it worked.

• ▸Segment: the employee group under review (role, tenure, location).
• ▸Journey stage and touchpoint: where the practice occurs (onboarding, goal-setting, feedback, training, tools and resources).
• ▸Practice description: what the company actually does today.
• ▸Antecedent / expected behavior / consequence: the ABC chain the practice is meant to create.
• ▸Measure and data source: the survey item or system data that tracks it.
• ▸Link to business outcome: key driver result (strong, weak, or untested) and the outcome it was tested against.
• ▸Segment lifetime value / activation score: your estimate, used to set priority.
• ▸Prediction: which metric should move, in which direction, by when.
• ▸Decision: keep, change, or stop, with the time and resources freed or reallocated.

1. 1Name the key jobs first. Ask which jobs produce the business outcomes your strategy depends on, such as the pet retailer's in-store service roles. Retention spending that covers every job equally is spread thin and 'might produce inferior results — or no results'.
2. 2Cut attrition by key job and performance segment before you read any overall number. An aggregate attrition figure is 'a good sign' but gives little insight. The differences between important segments are where the insight is.
3. 3Measure three things per segment: exit rate, retention rate, and commitment. These are the Chapter 11 measures. Keep the segment definitions fixed so you can compare them over time.
4. 4Link attrition in each segment to a business outcome. The pet retailer analyzed store-level employee data over time and found that service employees' satisfaction and retention moved customer loyalty and sales more than cashiers' or stockers' did. Run the same check in your own data.
5. 5Find why people leave in the priority segments. Use exit and survey data, and run key driver analysis to see which survey items correlate with commitment and attrition. Treat the drivers as correlations to test, not as proven causes.
6. 6Build the career path and internal mobility route into the key jobs. The pet retailer let any entry-level store worker apply for the more lucrative service jobs, and many did. Publish the route, the criteria, and the pay logic. Job responsibility is a legitimate basis for pay differences.
7. 7Test interventions as experiments. Apply a manager intervention or a path program to some units and compare against others. State your prediction beforehand, then check whether it came true.
8. 8Review annually and drop what fails. Retention programs that no longer serve the company, or never clearly worked, get cut. Don't keep adding programs without removing old ones, because that is how HR teams end up overworked and inconsistent.

A pet retailer added in-store services and wanted to know whether the investment in service employees paid off. Stores with services produced more of the desired business outcomes than stores without them. Analysis of store-level employee data over time then showed that the satisfaction and retention of the new service employees affected customer loyalty and sales more than that of cashiers or stockers. Across all jobs, the more distinct pet-related knowledge a job required, the more its attrition hurt the company. Margins were thin, so the retailer concentrated its people budget on reducing attrition in those key jobs instead of spreading the same money across every role. It also kept an internal route open: entry-level store workers could apply for the more lucrative service jobs, and frequently did. The pattern to copy is to segment first, tie each segment to outcomes, put the money where attrition costs most, and give internal candidates a path to the key roles.

• If a job's attrition correlates with customer or sales outcomes (as service roles did at the pet retailer), then fund retention there first. If it does not, then hold spending flat.
• If overall attrition looks acceptable, then still split it by key job and performance segment before concluding. A healthy average can hide losses in the segments that matter.
• If a proposed retention practice is justified only by 'other successful companies do it', then treat it as an untested assumption. Predict its effect, pilot it, and compare against the prediction.
• If a key job requires distinct, hard-to-replace knowledge, then build an internal path from adjacent entry-level roles. Don't rely only on external hiring.
• If a program shows no effect in your own data, then retire it and reassign the time and budget to a driver that does.

• Spreading the retention budget evenly across all jobs. The tell is that spending per job is similar while attrition costs differ widely.
• Reporting only a company-wide attrition rate. The tell is a single number in the dashboard with no key-job or performance split.
• Copying 'best practices' from successful companies without checking they caused those companies' success. The tell is a justification that names competitors but cites no internal data.
• Adding retention programs every year without removing old ones. The tell is a long program list, an overworked HR team, and no evidence that any one program moved attrition.
• Treating survey drivers as proven causes and rolling out company-wide. The tell is skipping the experiment or comparison group and having no written prediction to check.

• ▸Key job and why it is key: state the business outcome it drives and how much distinct knowledge it requires.
• ▸Segment cut: list the job and performance segments you will track attrition for.
• ▸Exit rate / retention rate / commitment: record the current value of each measure for this segment.
• ▸Outcome link: record the evidence that attrition in this job moves sales, loyalty, or another business result.
• ▸Top reasons for leaving: list the drivers from exit data and key driver analysis, marked as correlation or tested.
• ▸Internal mobility route: name the feeder jobs, the application process, and the pay logic for the move.
• ▸Manager intervention: describe the specific action, who owns it, and when.
• ▸Prediction and test design: state the expected change and the comparison group before launch.
• ▸Review result and decision: record what happened and whether to scale, change, or stop.

1. 1Define the candidate segments first. The customer journey map starts by defining the path that key customer types take to the product, and the same applies here. Pick the one or two candidate groups you most need (for example, the KSAO profile of your best performers) and map them separately. A single blended 'candidate' hides the differences in what each group sees.
2. 2Lay out the candidate stages of the employee journey map (Figure 8-2 logic). List each stage a candidate passes through before and during the hire, from first awareness through offer. Perceived attraction is scored stage by stage, not as one number.
3. 3List the company touch points at each stage (Figure 8-3 logic): every major interaction the candidate has with you. For each touch point, record the candidate's changing feelings, motivations and questions. Those three items are what 'desirable place to work' is made of at that moment.
4. 4Assign a feedback tool to each stage (Figure 8-4 logic). Choose how you will quantify the candidate experience there, for example a short survey item at that touch point. Add data from systems where you have it (Figure 8-5). Write each survey item so it measures one of the three components: brand, culture or the specific job.
5. 5Compare perceived to intended. The journey map is used to compare customers' perceived interactions with the company's vision of the experience. Write down the experience you intend a candidate to have at each touch point, then place the measured perception beside it. Gaps are your work list.
6. 6Test the specific offer against the market. Compare what you offer with what qualified people need in order to leave their current employer. If offers keep failing to attract the talent you want, the job-opportunity component is the problem, whatever the brand scores say.
7. 7Test whether your differentiators are visible outside the company. If you pay top performers well, ask whether the best employees at other companies know it. The source calls this an important targeted attraction tool if they do, and not one if they don't. A real advantage that no candidate perceives contributes nothing to perceived attraction.
8. 8Close the loop with the summary view. Put stage, touch point, feedback tool and what the data shows on one page, as the journey map does. Review it against talent acquisition process metrics (hire rate, time-to-fill, time-to-start) to see whether weak perception at a stage shows up as lost candidates or slow hires.

The source gives a pay scenario that shows perceived attraction failing at the job-opportunity component. Five years ago you hired someone at $30 per hour. Annual raises took that employee to $35. The economy has since taken off, and you can't find anyone qualified who will leave their current employer for less than $55 per hour. Your $35 offers are failing to attract the level of talent you want. Map it: at the offer touch point, qualified candidates see a number $20 below what they need, so their perception of the opportunity is poor however strong your brand or culture is. A survey item at the offer stage would show the drop-off there and not at earlier stages. The source lists market pay at time of hire as a normal reason people in the same job are paid different amounts. The source also warns that pay is sensitive and that people who don't work with pay data can't see the problem. So the analysis has two sides: quantify the gap between your offer and the market-clearing rate, and quantify how pay differences are perceived inside the company before you decide. The data tells you which problem you actually have.

• If candidate perception scores fall at one specific stage or touch point, fix that touch point before touching brand messaging.
• If offers are failing while earlier stages score well, treat it as a job-opportunity (pay and offer) problem, and compare your offers with the rate qualified people need to leave their current employer.
• If you pay top performers distinctly well but the best employees at other companies don't know it, treat it as a visibility problem. Aim the message at them. If they do know it, use it as a targeted attraction tool.
• If the intended experience and the perceived experience diverge at a touch point, trust the candidate-side measurement and redesign the touch point, not the vision document.
• If you cannot name the feedback tool for a stage, you have no measurement there. Add one before drawing conclusions about that stage.

• Scoring one blended 'attractiveness' number. The tell: you cannot say which stage, touch point or component (brand, culture, job) is weak.
• Mapping only what the company does, not what the candidate feels. The tell: the touch-point list is complete but has no feelings, motivations or questions attached.
• Assuming a real advantage is a perceived one, such as strong pay for top performers. The tell: top outside talent has never heard of it, so it does no attracting.
• Ignoring the market gap in offers. The tell: repeated offer rejections while earlier-stage scores look healthy, and offers still sit far below what qualified people need.
• Treating pay differences as self-explanatory. The tell: stakeholders react to a new offer rate versus incumbent pay without understanding the reasons (market at hire, tenure, performance and so on) that make such differences normal.

• ▸Candidate segment: name the key candidate group this map covers (e.g., a target KSAO profile).
• ▸Stage: the candidate stage being scored (e.g., awareness, application, interview, offer).
• ▸Touch point: the specific interaction the candidate has with your company at this stage.
• ▸Candidate feelings, motivations, questions: what they feel, want and ask at this touch point.
• ▸Intended experience: what you want the candidate to perceive here.
• ▸Feedback tool: the survey item or system data used to quantify the perception at this stage.
• ▸Perception score by component: brand, culture, job opportunity, each rated from the feedback tool.
• ▸Gap to intent / market: the difference between intended and perceived experience, and for offers, between your rate and what qualified people require.
• ▸Action and owner: the single change you will make at this touch point, and who owns it.

1. 1Define the performance you are testing capability against. Capability only means something relative to a job. Start from strategic planning and job analysis, because the source names them as the basis for selection criteria related to job performance. Write the actions the job requires before you list any traits.
2. 2List the KSAOs those actions require, in the source's categories: knowledge, skills, abilities, and other characteristics. Use the source's examples as prompts: technical knowledge, learning agility, social skills / EQ, grit. Keep the list to what the job analysis supports. Do not add traits that merely sound good.
3. 3Sort each KSAO by how it can be acquired: select for it, train for it, or neither. The source gives the two levers (recruiting and training) and two caveats: some characteristics cannot be created through training, and some can but cost too much time and money. Mark each item on that basis.
4. 4Make selection your primary channel. The source calls optimal selection of people for jobs, using criteria tied to job performance, the company's primary channel for raising capability. Put the non-trainable and too-costly-to-train KSAOs into your selection criteria first. Use training for the remainder.
5. 5Assess capability separately from the other three activation conditions. Performance requires capability, alignment, motivation, and support (CAMS). If any one is missing, reliable performance is difficult or impossible. Measure each condition on its own so a support gap is not recorded as a capability gap.
6. 6Check whether you have controlled for the other conditions before concluding a KSAO matters. The source warns that when the other factors are handled well, some characteristics thought critical turn out not to be. Treat each KSAO as a hypothesis until alignment, motivation, and support are accounted for.
7. 7Test the KSAOs against outcomes with the analysis tools the source describes. Correlate each capability item with a performance KPI, as in key driver analysis. Remember that correlation shows only the possibility of a causal relationship. A larger coefficient raises that possibility and a smaller one lowers it.
8. 8Revise the capability list as the evidence comes in. The source frames people analytics as a loop of theory, test, and learn, then improve the model. Drop KSAOs that never matter in practice, and move stubborn gaps from the training column to the selection column.

This is an illustrative walk-through. The passages give no company case for capability, so the job and every value below are assumed. Suppose a team lead role shows weak performance and leadership suspects the people lack capability. Job analysis lists the actions the role needs, and from those you list KSAOs: technical knowledge, learning agility, social skills / EQ, and grit. You classify them. Technical knowledge is trainable at reasonable cost. Learning agility and grit are hard or too costly to create through training, so they move into selection criteria. Before blaming capability, you check the other three CAMS conditions. If people are unclear on what a good result looks like (alignment), unmotivated, or missing tools, then capability is not the binding constraint. The source's warning applies here: even an extraordinarily capable person placed in a situation without appropriate supports will fail. Only when the other three conditions are in place do you read the remaining performance gap as a capability gap. At that point you correlate each KSAO measure with the performance KPI. You keep the KSAOs that move with performance and treat the rest as unproven.

• If performance is weak and you have not checked alignment, motivation, and support, then do not call it a capability problem yet. Those three conditions come first, because a missing condition makes reliable performance difficult or impossible.
• If a KSAO cannot be built through training, or training would cost too much time and money, then move it into selection criteria rather than a training plan.
• If a selection criterion cannot be traced to job analysis or strategic planning, then drop it or test it. The source ties selection criteria to job performance determined that way.
• If a trait looks critical but shows no relationship to performance once the other conditions are handled well, then treat it as not actually critical and remove it from the profile.
• If a capability measure correlates strongly with a performance KPI, then treat it as a candidate cause worth testing further. Do not treat it as proven, because correlation does not establish causation.

• Treating every performance gap as a skills gap and defaulting to training. The tell: training spend rises while performance does not move, and nobody has checked alignment, motivation, or support.
• Writing selection criteria from intuition or copying other companies instead of from job analysis. The source describes this habit as traditional HR decision-making. The tell: the criteria cannot be linked to any specific job action or outcome.
• Hiring a highly capable person into a role without support. The tell: strong profiles fail early, and the explanations point to missing tools or resources.
• Keeping a long KSAO list of 'critical' traits that never get tested. The tell: criteria persist even though no analysis shows they relate to performance.
• Doing a lot of analytics work with no strategic grounding. The source says this produces work nobody uses. The tell: capability dashboards exist but no one uses them to decide on hiring or training.

• ▸Job / team: name the role or team you are assessing.
• ▸Performance actions required: list the actions the job must perform, taken from job analysis and strategic planning.
• ▸KSAO list: for each action, list the knowledge, skills, abilities, and other characteristics needed (e.g., technical knowledge, learning agility, EQ, grit).
• ▸Acquisition route per KSAO: mark each as Select, Train, or Too costly / not trainable.
• ▸Selection criteria: write the criteria that screen for the Select and Not-trainable KSAOs, with the job-performance link for each.
• ▸Other CAMS conditions check: record whether alignment, motivation, and support are in place (yes / no / unknown) before reading capability results.
• ▸Performance KPI: name the outcome measure you will test each KSAO against.
• ▸Evidence: record the correlation of each KSAO measure with the KPI, and note that it is correlational only.
• ▸Decision: for each KSAO, mark keep, retest, or drop, then note the next change to selection or training.