Excellence in People Analytics

1. 1Reject the 'fix all people data first' trap. Write down that perfect data is not a precondition for starting. Then pick the specific topics where better availability, quality and accessibility would pay off.
2. 2Build a data inventory for the project in front of you. List what people data exists, where it sits, and what is missing, outdated or undefined. The aim is to know what is available, not to cleanse everything.
3. 3Run the checks the source names as the practical data issues: is it good enough, missing, outdated, absent altogether, non-normal in distribution, full of outliers, or defined inconsistently across systems. Record each finding against the question you are answering.
4. 4Score data availability, accessibility and quality as one input to project complexity on the Complexity-Impact Matrix, next to technology, resources, politics, sponsorship and ease of productizing. Projects with accessible, good-enough data and a sponsor become Quick Wins. Data-poor, high-value projects go in the Big Bet portfolio.
5. 5Set up governance explicitly. Decide who structures, manages, governs and secures people data. Assign stewardship, which in some leading companies sits with the people analytics function (HSBC). Name a privacy and security owner, as Microsoft HR Business Insights did with a privacy manager who also advises other areas of Microsoft on their use of employee data.
6. 6Build the skills to capture, clean, process, analyse and present the data. Where needed, recruit governance roles covering ethics, data privacy, data management and risk frameworks.
7. 7Make quality visible to stakeholders. Apply a consistent 'stamp of approval' to validated work, as Trimble does with its logo. Treat unstamped analyses as requiring validation before anyone relies on them.
8. 8Keep a standing list of new data sources to seek, including data from outside HR (the source cites Uber's call to integrate all types of data, and Jeremy's team adding external data). Review it as requests and challenges grow.

At Trimble, Suku's people analytics team built its brand on credible analysis, one small step at a time. There was no single moment of having 'arrived'. The brand became synonymous with quality and reliability, so he created a logo for Trimble People Analytics. CHRO Mike Scarpa describes it as a 'stamp of approval': where the logo appears, the recipient knows the work is credible, of high quality and trustworthy. When stakeholders see people analyses or insights without the logo, they ask the team to validate them. Even the finance team brings data and insights to the people analytics team for validation. The team also began with Quick Wins such as why people were leaving, which kept the work on problems that mattered without waiting on perfect data. The lesson is that data quality works as a managed, visible commitment, and it earns trust before the dataset is perfect.

• If someone says the data must be fixed before analytics can start, then reject that and pick the specific data topics that matter for the current question.
• If a project's data is hard to get at, low in quality or missing, then score it as higher complexity on the matrix. Do not treat it as a Quick Win, and plan it as a Big Bet only if the impact justifies the sponsorship.
• If your function handles sensitive employee data, then assign clear stewardship and a named privacy and security owner before you scale the solutions.
• If the analysis needs context beyond HR records, then pursue other data sources, including external and non-HR data, rather than limiting the work to HR data.
• If stakeholders keep asking you to re-validate outputs, then make quality visible through a consistent validation mark and credible analysis, so trust transfers to the work.

• Waiting for perfect data: projects stall and the team spends its time cleaning. The tell is that no analysis has been delivered while 'data readiness' is still the stated blocker.
• Treating data as only a technical task and ignoring governance and security: no named steward, no privacy owner. The tell is that nobody can say who owns a dataset or who approves its use.
• Inconsistent data definitions across sources: the same metric means different things in different reports. The tell is stakeholders disputing figures instead of discussing decisions.
• Ignoring accessibility when scoping: you choose high-impact projects without checking whether the data can be reached. The tell is projects that were expected to be Quick Wins turning into long, complex efforts.
• Staying within HR data only: insights are narrow and miss business context. The tell is that the people analytics team cannot link its findings to workforce planning, location or budget questions.

• ▸Business question: state the decision the analysis will inform.
• ▸Data needed and source system: list each dataset and where it lives.
• ▸Availability: available, partial or absent; note any workaround if absent.
• ▸Accessibility: who can access it, how, and what blocks access.
• ▸Quality issues: missing, outdated, outliers, non-normal distribution, inconsistent definitions.
• ▸Good-enough call: is the data adequate for this question, and what caveat will you state?
• ▸Steward and privacy/security owner: name the person responsible for each.
• ▸Complexity score input: data availability, accessibility and quality rating for the Complexity-Impact Matrix.
• ▸New or non-HR data sources to seek: external or cross-functional data that would strengthen the answer.
• ▸Validation stamp: has the output been checked and marked as credible before release?

1. 1Start from an outcome executives already care about. Thomas Rasmussen (National Australia Bank) says: 'make sure there is an outcome that business executives really care about.' Write the business challenge in one sentence before you scope any analysis. If no executive owns that sentence, don't start.
2. 2Secure a committed business sponsor before work begins. The book's methodology treats prioritization, defined processes and committed sponsors as the three pillars. Name the sponsor, state what they will do with the result, and confirm they will drive the change. In Rasmussen's words, 'The business leader is the most important part of the change management activity.'
3. 3Prioritize against a shared framework, not by request order. NAB's central analytics hub supplies a prioritization framework so analytics teams focus on the most important topics group-wide. Rank each candidate project by executive relevance and sponsor commitment, and decline or defer the rest.
4. 4Staff the project cross-functionally and keep the functional expertise. In NAB's hub-and-spoke model, central governance sits with the hub and each spoke is aligned to a function, because each function's expertise guides the analysis. Put people from the business function alongside the analysts on the single business challenge.
5. 5Run the analysis in a defined process, and treat unexpected findings as a trigger for further work. Do not report them as trivia. NAB found that branches led by female leaders had higher average customer satisfaction. The team's response was to ask what causes it, not just to report it.
6. 6Convert insight into an operating change, not a slide. At NAB the findings changed the reward approach, linking branch performance with employee engagement. The three key people factors received more prominence and weighting and became forward-looking metrics. Specify which decision, incentive, metric or process changes, and who owns it.
7. 7Scale only after the first solution is embedded. Rasmussen pairs the business-led approach with a digital component: 'The digital component enables you to scale it.' Productize the solution once leaders use it, and don't build scale ahead of adoption.
8. 8Choose the analytics type to match the question. The book's range runs from reporting and KPIs through predictive to advanced analytics and AI. Use the simplest type that answers the executive's question, and don't default to advanced methods.

At National Australia Bank, Thomas Rasmussen's team worked on a cross-functional analysis aligned to a common business challenge: branch performance. The analysis identified three key people factors that predicted branch success. It also produced an unexpected insight: branches led by female leaders had a higher average customer satisfaction score than those led by male leaders. The team followed up by asking what it is about those leaders that produces the result, rather than stopping at the headline. The results changed the bank's reward approach, linking branch performance with employee engagement. The three people factors were given more prominence, weighting and focus, and became forward-looking early indicators of branch performance. Branch leaders valued the indicators because they were evidence-based predictors of success. The sequence is the pattern to copy: a business outcome, a cross-functional team, further analysis of the surprise, then a change in how the business operates.

• If no business executive will sponsor and act on the result, then do not start the project, however interesting the question.
• If the analysis surfaces an unexpected finding, then run a further analysis into the underlying cause before reporting it.
• If competing requests exceed capacity, then rank them with a shared prioritization framework tied to the most important business topics, not to who asked first.
• If an insight does not change a decision, metric, incentive or process, then treat the project as unfinished and plan the change management with the business leader.
• If a solution works for one business area and leaders use it, then add the digital layer to scale it; if not, fix adoption first.

• Analysis without an executive-owned outcome. The tell: the readout is met with interest but no one names a decision it will change.
• Treating change management as an afterthought. The tell: insights are delivered and then stall, with no business leader driving action.
• Analytics isolated from functional expertise. The tell: findings that people in the function say miss how the work really operates, because the team lacks input from the function that guides the analysis.
• Stopping at the first answer. The tell: a surprising result is reported as a footnote and never investigated for causes.
• Scaling before embedding, or ad hoc prioritization. The tell: dashboards nobody uses, and a project list set by the loudest requester instead of a framework.

• ▸Business challenge: one sentence stating the outcome an executive cares about.
• ▸Executive sponsor: name, and what they commit to do with the result.
• ▸Priority rationale: why this ranks above other requests under your prioritization framework.
• ▸Cross-functional team: the analysts plus the functional experts who guide the analysis.
• ▸Analysis type: reporting, KPI, predictive or advanced, with a reason that matches the question.
• ▸Unexpected-finding follow-up: the question you will ask if a surprising result appears.
• ▸Change commitment: the decision, metric, incentive or process that will change, and its owner.
• ▸Forward-looking indicators: early metrics leaders can track, and who uses them.
• ▸Scale path: the digital step to take only after leaders adopt the solution.

1. 1Define the construct as three things you can observe: skill (HR can do the analysis), mindset (HR asks for evidence before deciding), and confidence (HR is willing to present data to executives). Rate each across the HR function, not just the analytics team. The source treats culture as a separate dimension from skills, so measure it separately.
2. 2Map who is doing analytics in HR today. The source says growth comes from two streams: HR professionals acquiring analytics skills, and data science and analytics professionals entering HR from finance, marketing and operations. Count both streams so you know whether you are building skill in HR generalists, importing it, or doing both.
3. 3Run a visible, narrow capability-building effort across HR. The source's Culture chapter covers 'Building capability in analytics across HR' and the Rabobank case 'Engage and enable HR to kick-start a data-driven culture'. Engage HR first, then enable it. Pick one small cohort and one real HR question per person, not a generic training programme.
4. 4Teach every HR practitioner to tie a people insight to a financial or business metric. Nielsen's finding that each 1 per cent of attrition avoided saves $5 million is the model. Business executives engaged because the insight connected directly to a financial metric.
5. 5Pair HR with finance on early projects. Use the Focus-Impact-Value model: Focus (scope, sponsor, budget, timeframe), Impact (effect on stakeholder groups), Value (anticipated financial value, using ROI, NPV or IRR). Use the language of finance, as MetLife did. HR practitioners learn how to build a defensible case by doing it.
6. 6Teach storytelling alongside statistics. Nielsen combined statistics with strong communication and clear storytelling about business value and employee benefit. Require each HR practitioner to present one finding in plain language to a business leader.
7. 7Create structures that spread confidence. The source's chapter pairs 'Create structures and build confidence' with the PepsiCo case 'Build globally, enable locally'. Set up a central standard (methods, tools, ethics) and a local route for HR partners to apply it in their own markets. Do not name specifics beyond this, because the supplied passages give only the headings.
8. 8Show leaders their own data and watch the reaction. At NAB, giving leaders the actual data made it tangible, and they 'loved it'. They then saw how people analytics could improve branch performance. Use the same move with HR business partners, and count how many ask follow-up questions or request a new analysis.

Nielsen found that for every 1 per cent of employee attrition it could avoid $5 million in business costs. The analysis also showed internal mobility as the key driver of retention. The source says it was the direct link between a people insight and a financial metric that made business executives connect with and get excited about people analytics. Nielsen paired the statistics with strong communication to stakeholders and clear storytelling about business value and employee benefit. For HR culture, the lesson is that analytic confidence grows when a practitioner can state a finding as a number the CFO recognises and then tell the story behind it. NAB shows the same effect from another angle. Its team found customer satisfaction was twice as high in branches with the highest engagement. Tenure in role was a better barometer of capability than competencies. Safety also mattered. Leaders were given the actual data and 'loved it'. The NAB team said the beauty of the analysis was evidence-based proof of what had only been executive intuition. To build culture, have every HR practitioner produce one Nielsen-style finding (insight, financial metric, story) and one NAB-style finding (intuition tested against data).

• If HR practitioners can describe a people problem but not attach a financial or business metric to it, then train on the insight-to-metric link first (the Nielsen pattern) before teaching any further technique.
• If an analysis needs credibility with executives, then co-build it with finance and use finance's language and valuation methods (ROI, NPV, IRR), as in the MetLife principle.
• If HR leaders rely on intuition, then put the evidence in front of them in tangible form, using their own team's or branch's data, as NAB did, instead of explaining methods in the abstract.
• If capability sits only in a central analytics team, then build structures that let local HR apply central standards, following the 'build globally, enable locally' approach in the PepsiCo case.
• If a project lacks a sponsor, budget or realistic timeframe, then do not start it as a culture-building exercise. Run the Focus step first, because a failed early project teaches HR that data work does not pay off.

• Treating culture as training attendance. The tell: courses completed rise but no HR practitioner brings a data-based recommendation to a decision meeting.
• Isolating analytics in a specialist team while the broader HR function stays passive. The tell: every data request is routed to the same few people and HR business partners cannot explain the findings.
• Reporting insights with no business metric. The tell: executives nod politely, but nothing is funded or changed, unlike Nielsen's $5 million-per-1-per-cent attrition link.
• Presenting statistics without a story. The tell: accurate analysis is met with silence. The source says Nielsen paired stats with storytelling about business value and benefit to employees.
• Running analytics without finance. The tell: investment cases are challenged on assumptions and valuation method, and the project stalls at the Value step.

• ▸Skill baseline: list which HR roles can run analysis today and which cannot, split between HR staff upskilled and analysts brought in from other functions.
• ▸Mindset signal: name one recent HR decision and state whether data was consulted before it was made.
• ▸Confidence signal: record how many HR practitioners have presented a data finding to a business leader in the last quarter.
• ▸Starter question per practitioner: write one real people question each practitioner will analyse.
• ▸Business metric link: state the financial or business metric each question connects to, as with Nielsen's attrition cost.
• ▸Finance partner: name the finance contact and the valuation method (ROI, NPV or IRR) agreed for the case.
• ▸Focus-Impact-Value check: record scope, sponsor, budget, timeframe, affected stakeholder groups and expected value.
• ▸Story plan: write the two-sentence story that explains the finding to a non-analyst, including benefit to employees.
• ▸Structure for scale: define what is set centrally and what local HR teams may adapt (build globally, enable locally).

1. 1Name the audience before you build anything. The source splits Workforce Experiences into four: the employee as an 'employee of one' (personalization), the workforce as a consumer of people processes, the manager as a consumer of people data, and the executive as a consumer of people insights. Write one line per audience on what they should experience differently.
2. 2Pick one process or decision to improve for that audience, such as performance management. Define it in business terms first. FIS's first step is to define the process under review.
3. 3Gather every kind of data on that process, not only the quantitative. FIS pulled quantitative, survey, text and sentiment data together before analysing anything.
4. 4Analyse with a stated methodology, then turn the findings into insights and recommendations. Keep the findings and the recommendations as separate outputs, so people can challenge the evidence apart from the proposed change.
5. 5Co-design the recommendations with the employees the process affects. Interview and talk with them as part of the work. The FIS lead said changes held for decades became possible once 'the data shows it and it was co-created with employees'.
6. 6Implement the recommendations and evaluate them against pre-agreed measures. FIS measured participation (95 per cent in year one of Quarterly Connects) and perceived helpfulness (93 per cent found it helpful).
7. 7Keep the people you consulted involved as a champion network. FIS champions got creative with each new deployment and ran internal talk shows and radio shows for local offices. Then iterate: step six of the FIS method is to continue co-designing with employees.
8. 8Adapt the delivery to the audience. Managers need democratized data on their own teams. Executives need interactive analytics they can explore, which is the Santander Brasil theme of bringing analytics to life in front of executives. Employees need personalized, consumer-grade experiences, which is the ABN AMRO theme of measuring employee experience.

FIS (Fidelity National Information Services) used people analytics to change performance management across the company. Isabel's team followed a six-step method: define the management performance question; gather quantitative, survey, text and sentiment data; analyse it with a data methodology; draw insights and recommendations; implement the key recommendations and evaluate success; then iterate and keep co-designing with employees. The result was Performance365, built around frequent touchpoints called Quarterly Connects, where employees align on goals and focus their development. In the first year 95 per cent of colleagues took part and 93 per cent said the process was helpful. Isabel noted that changes the organisation had wanted 'for decades' became possible because the data showed the need and employees co-created the answer. The people interviewed became a network of Performance365 champions who hosted talk shows and radio shows in their local offices to explain the work. The team also kept the six-step method as a template for future large-scale enterprise projects.

• If a people process has resisted change for years, then build the case with multi-source data (quantitative, survey, text, sentiment) and co-design the fix with the affected employees. Data plus co-creation is what moved FIS.
• If you cannot name which of the four audiences the work serves (employee of one, workforce, manager, executive), then do not start building. The format and data depth differ for each.
• If you are about to roll out a new process or recommendation, then recruit the employees you consulted as champions first. They speed up acceptance and adapt the message locally.
• If you are reporting success, then report both participation and perceived value, as FIS did with 95 per cent participation and 93 per cent finding it helpful. Participation alone can reflect compliance.
• If the audience is executives, then favour interactive analytics over static reports, because the Santander Brasil case is about bringing analytics to life in front of them.

• Treating workforce experience as a reporting job: dashboards go out, but no HR process changes. The tell is that the outputs are scorecards and no employee behaviour or process differs afterwards.
• Analysing only numbers. Ignoring text and sentiment data leaves the 'why' missing. The tell is recommendations that employees say do not match their reality.
• Designing the process for employees instead of with them. The tell is slow acceptance and no local advocates when you deploy.
• Giving every audience the same output. Managers, executives and individual employees consume insights differently. The tell is low usage by one group, such as executives who never open the pack.
• Declaring success at launch with no evaluation or iteration. The tell is that no participation or helpfulness measure was set beforehand and no step exists for revisiting the design.

• ▸Audience: employee of one, workforce, manager or executive. Pick the primary one.
• ▸Process or decision to improve: name it and state what is wrong today in business terms.
• ▸Data sources: list quantitative, survey, text and sentiment sources, and note any gap.
• ▸Analysis method: state the methodology and how findings become insights, then recommendations.
• ▸Co-design group: name the employees you will interview and involve, and when.
• ▸Recommendation and delivery format: describe the change and how the audience will experience it, such as personalized, democratized to managers, or interactive for executives.
• ▸Success measures: set participation and perceived-helpfulness targets, with baseline and date.
• ▸Champion plan: say how consulted employees will carry the change into their local areas.
• ▸Iteration trigger: state when you review results and what would prompt redesign.

1. 1Start from the executive's outcome, not your data. Thomas at NAB says to 'make sure there is an outcome that business executives really care about.' Before any project is scoped, write down which executive owns the outcome and which of their metrics moves.
2. 2Define stakeholder results first, as the dependent variables. The Foreword's balanced-scorecard categories give you a checklist: employee, strategy, customer, financial and community results. Pick the one or two your project must move and name the metric for each.
3. 3Classify the outcome type using the four categories in Figure 8.1: workforce experiences, analytics culture, business performance, societal benefits. Within business performance, name the specific type: financial impact, revenue and market share growth, or managing risk and compliance. Not every project is a financial-return project, but every project needs a named outcome.
4. 4Map the pathway from initiative to outcome. List the people initiatives (talent, leadership, organization, HR department) you believe drive the metric. In the NAB figure, skilled people, high motivation and opportunity to contribute sit on the pathway to financial and operational outcomes. Write your own version as 'driver → driver → metric.'
5. 5Apply the 'insight without outcome is simply overhead' test (Piyush Mathur, J&J). For each insight, write the decision it informs, who makes it, and by when. If you cannot name a decision-maker, the insight is not ready to present.
6. 6Assign a business leader to own the action and plan the change management. Thomas calls the business leader 'the most important part of the change management activity.' Agree the owner and the action before delivering the analysis.
7. 7Design for scale. Ask whether the solution can reach the whole workforce through technology rather than a few hundred people at a time. NAB targets its 40,000-strong workforce this way, and the scaled usage generates data that feeds back into the process.
8. 8Report the outcome in business terms (revenue, cost, productivity, risk) and use the result to make the case for further investment. Realized outcomes counter 'analytics for the sake of analytics' and justify more funding.

National Australia Bank (40,000 employees) shows the outcome logic in the passages. Thomas's rule is to anchor on an outcome executives care about, and NAB's Figure 0.7 links skilled people, high motivation and opportunity to contribute to financial outcomes (labelled 26%) and operational outcomes (labelled 22%). The passage does not explain how those percentages were derived, so treat them as labels, not a method to copy. The operational lessons are explicit. First, the business leader owns the change management that turns insight into action, which is where measurable value is created. Second, technology moves impact from initiatives touching a few hundred people at a time to the full workforce. Third, scaled usage generates new data that feeds back into the process. Replicate the structure: name the executive outcome, draw the driver pathway to it, assign the business owner, and build the delivery to scale and to feed data back.

• If you cannot name an executive who cares about the outcome metric, then do not start the project; reframe it until you can.
• If a finding has no decision-maker and no action date, then treat it as overhead and do not present it as a result.
• If the project's value is not financial (for example experience, inclusion or wellness), then still name the outcome category from Figure 8.1 and its metric, so it is not accused of being analytics for its own sake.
• If the insight only reaches a small group, then ask whether technology can scale it to the whole workforce before you report the outcome.
• If you want more investment, then lead with the business outcome achieved, because clear return on investment is what excites senior executives and elevates the function.

• Reporting HR activity measures (scorecards, dashboards, interesting tidbits) as outcomes. The tell: the headline metric is an HR measure, not a business or stakeholder result.
• Delivering insight with no owner. The tell: the readout ends with 'findings' and no business leader committed to an action.
• Starting from available data instead of the executive's problem (an inside-out approach). The tell: the project brief opens with the dataset, not a business challenge.
• Assuming every project must show financial return and dropping the others. The tell: workforce experience or societal projects are left with no stated outcome at all, instead of a defined non-financial one.
• Neglecting change management after the analysis. The tell: a good model exists but nothing changes in the business, so no measurable value appears.

• ▸Executive sponsor and the outcome they care about: name the person and state the outcome in their words.
• ▸Outcome category: pick one from Figure 8.1 (workforce experiences, analytics culture, business performance, societal benefits) and, if business performance, the sub-type (financial, revenue/market share, risk/compliance).
• ▸Stakeholder result (dependent variable): choose employee, strategy, customer, financial or community, and name the metric.
• ▸Driver pathway: write the chain of initiatives or drivers you believe moves the metric (talent, leadership, organization, HR department).
• ▸Decision this insight informs: state the decision, who makes it, and the date.
• ▸Business owner of the action: name the leader who owns change management.
• ▸Scale route: say how technology extends this beyond a small group, and what feedback data it will generate.
• ▸Reported result: state the outcome in business terms once achieved.
• ▸Next-investment case: say what further funding or scope the result supports.

1. 1Anchor the function to corporate strategy first. Before choosing projects, identify the company's strategic priorities and tie each people analytics project to one of them. The Novartis case in the Governance chapter is titled 'Start with strategy'. Use it as the model: strategy comes before the project list.
2. 2Write a vision statement with stakeholders, not alone. It should be memorable and state the purpose and ambition of people analytics in your organization. Drafting it with stakeholders makes it theirs.
3. 3Write a mission statement for the team. It defines the function's objectives, guides the team's choices and states which outcomes it will deliver. Test it by asking whether the team can use it to turn down a request.
4. 4Build a brand as a marketeer would. Make the function visible across the business so stakeholders know who you are and what you do. Trimble's case is titled 'Credibility begins with a clear brand'. Treat brand as the way credibility is established.
5. 5Form a 'board of people analytics' to give the function direction and enablement. This creates stewardship and accountability. Give it a standing role, not an occasional review.
6. 6Make ethics a standing process by writing an ethics charter. Involve stakeholders in drafting it. Consider an academic third party for review, as Lloyds did, and expect about six months to launch. Use the charter to tell colleagues openly how their data is used and what the benefits are.
7. 7Test the governance under pressure. Keep the whole charter ready to send within minutes if a regulator, executive or CHRO asks about compliance and people data. Treat a fast 'no further questions' as the pass mark.
8. 8Run governance alongside the other dimensions, not as a gate. The Nine Dimensions model is not sequential and requires no maturity threshold before moving on. Organizations with the most impact work on foundation, resources and value at the same time.

Lloyds Banking Group's people analytics team (Andy and Justine) built an ethics charter as the core of its governance. The bank already worked with an academic at Birmingham University on ethical practice, so the team asked this third party to advise, making the work more robust. After about six months the charter was launched. Justine says it let them engage HR professionals, managers and all colleagues, and be open about how data is used and what the benefits are for everyone. A year later the CHRO asked about compliance and people data, because the Financial Conduct Authority had referenced the topic. Within minutes the team sent the entire charter and got the response 'no further questions'. That reply validated the governance mechanism, and the team now builds the next stage of its work with ethics at the core.

• If you cannot name the corporate strategy priority a project serves, then do not start it; align with strategy first.
• If stakeholders do not know what your function is for or who you are, then fix the vision, mission and brand before adding projects.
• If you are drafting an ethics charter, then involve stakeholders in writing it and consider an independent academic review for robustness.
• If direction and accountability for the function sit with one person or no one, then form a board of people analytics.
• If someone asks you to delay other dimensions until governance is mature, then decline: the model is non-sequential and the strongest organizations advance foundation, resources and value together.

• Governance written in isolation: the team drafts the vision and charter alone. The tell is that stakeholders cannot restate the vision and treat the charter as HR's document.
• Strategy drift: projects are chosen by who asks, not by corporate strategy. The tell is a project list you cannot map to strategic priorities.
• Invisible function: good work with no brand. The tell is that business leaders do not know the team exists or what it delivers.
• Ethics as an afterthought: no charter exists until a question arrives. The tell is that a CHRO or regulator query triggers a scramble instead of a same-day send.
• Treating governance as a prerequisite stage: you wait for 'maturity' before progressing elsewhere. The tell is that data, skills and stakeholder work stall while policies are drafted.

• ▸Strategy link: list each active project and the corporate strategic priority it serves; flag any with none.
• ▸Vision statement: write the memorable one-sentence purpose and ambition, and name the stakeholders who helped draft it.
• ▸Mission statement: state objectives, the outcomes the team will deliver and what it will not do.
• ▸Brand visibility: list the stakeholder groups and note whether each can say what the function does (yes/no).
• ▸Board of people analytics: name members, meeting cadence and decision rights for direction and enablement.
• ▸Ethics charter status: record draft/launched, date, stakeholders involved and any independent academic reviewer.
• ▸Open data use communication: note how colleagues are told how their data is used and the benefits to them.
• ▸Compliance response test: record how long it takes to produce the full charter and governance evidence on request.
• ▸Next fix: name the weakest item above, an owner and a due date.

1. 1Write the hypothesis before anything else. For every request, state the hypothesis in one sentence and describe what the business gains at the end of the project if it is proven and if it is disproven. If you cannot describe the value of both outcomes, the work is not ready to be scoped.
2. 2Size the value at a high level in one of three currencies: financial value, risk mitigation, or engagement for customers or employees and workers. Do not build a full business case. You need enough to decide whether to proceed.
3. 3Define the scope, then check impact against all Seven Types of Stakeholders: business executives, HR leaders, managers, employees and workers, functional stakeholders, technology and data owners, and unions, works councils and employee groups. Mark each as affected or not affected. An empty cell means you have not checked it.
4. 4Run the four-question go/no-go screen from the hypothesis and value statement: (a) is the work important to undertake, (b) is a sponsor required, (c) will stakeholders release budget for skills, data or technology, (d) is there likely to be significant impact on any stakeholder group. Record the answers and the decision.
5. 5Where you can, rank candidate work by its relative impact on the five outcome areas: employee, strategy, customer, financial and community results. The most valuable initiatives rank first, and the answer to 'why this one?' is that it delivers value to results that matter.
6. 6For work that proceeds, secure a sponsor for the most important projects before you start. Sponsorship follows prioritization. It does not replace it.
7. 7Make the process defined and repeatable for a small team. The American Eagle Outfitters case names three ingredients: processes, stakeholders and a manifesto. Write the intake screen, the stakeholder check and the sponsor step down so any team member can run them the same way.
8. 8Budget stakeholder engagement as project work. Viessmann spent 50 per cent of its time on people analytics development and 50 per cent on stakeholder management and communications, and involved executives iteratively and incrementally with prototypes. Treat the process as dynamic: re-run the screen when scope, stakeholders or sponsors change.

Viessmann's ViGrow platform shows the method in practice. The purpose was clear: HR needed a data-driven platform covering development, cost, skills gap and strategic direction for business goals. Steffen did not treat the build as the whole job. His team split time 50/50 between people analytics development and stakeholder management and communications. That split is the stakeholder-impact check made operational. Employee input came through presentations, town halls and global roadshows across countries as far apart as China, Turkey and the US, and reached almost 25 per cent of employees in three months. On sponsorship, Steffen worked to secure senior executive buy-in from across the business. He involved those executives iteratively, with prototypes and modules that showed HR was solving strategic problems. Feedback sparked further use cases, which made the platform stronger. This is what a dynamic process looks like: sponsor engagement is built into delivery and is not a one-off sign-off. The source does not show Viessmann's intake form, so the screen in the tool below is the book's method, not Viessmann's documented practice.

• If you cannot state the value of the project when the hypothesis is proven and when it is disproven, then do not scope it. Send it back to the requester.
• If the work has no high-level financial value, risk mitigation or engagement value, then do not proceed. It is not important enough to take capacity.
• If stakeholders will not release budget for the skills, data or technology the work needs, then pause it. Work that cannot be resourced produces substandard value.
• If a stakeholder group among the seven is significantly affected, then bring that group in before delivery. Add a sponsor if the work needs one.
• If two candidate projects compete for capacity, then rank them by relative impact on the five outcome areas and sponsor the higher-ranked one first.

• Taking every request. The tell is a backlog with no recorded go/no-go decisions and a team that is busy but cannot connect its output to outcomes. The source warns that without governance, work loses focus and unnecessary, low-value work increases.
• Scoping before stating the hypothesis. The tell is a project brief that describes data and methods but not the value if the hypothesis is proven or disproven.
• Checking only the requester. The tell is a late surprise from works councils, data owners or managers affecting a project that looked approved. The cause is skipping the Seven Types of Stakeholders check.
• Treating sponsorship as a signature. The tell is a named sponsor who is absent after kickoff. Viessmann kept executives involved iteratively with prototypes.
• Underbudgeting stakeholder and communications time. The tell is a technically finished product with low adoption, because the 50 per cent of effort Viessmann gave to engagement was never planned.
• Staying a reporting function. The tell is a team that only answers requests. AEO moved from a 'reporting' team to an 'outside-in' business-driven function.

• ▸Request and requester: name the work and who asked for it.
• ▸Hypothesis: state it in one sentence.
• ▸Value if proven / value if disproven: describe the business gain in each case.
• ▸Value type and high-level size: choose financial value, risk mitigation, or customer/employee engagement, and give a rough size.
• ▸Outcome area ranking: note which of the five (employee, strategy, customer, financial, community) it moves most, and how it ranks against other candidates.
• ▸Scope: say what is in and what is out.
• ▸Seven-stakeholder impact: mark each group as affected or not, and note the likely impact.
• ▸Resourcing: record whether stakeholders will release budget for skills, data and technology, and who must approve it.
• ▸Sponsor: record whether one is required, the named sponsor, and how you will engage them (for example, prototype reviews).
• ▸Decision and re-check date: record proceed, pause or decline with the reason, and set when you will re-run the screen.

1. 1Start from results, not from the analytics team. Before you list anyone, write down the outcomes the business wants. The foreword's frame is five areas: employee, strategy, customer, financial and community results. Every stakeholder conversation then opens with the outcome they own, not with your data or dashboards.
2. 2Map all seven stakeholder groups the book names: (1) business executives, (2) HR leaders, (3) managers, (4) employees and workers, (5) functional stakeholders, (6) technology and data owners, (7) unions, works councils and employee groups. Put a name against each group. A blank group is a gap in your plan.
3. 3For each group, write one line on what they need from the work and one line on what you need from them. Executives need results and sponsorship. HR leaders need a partner who can use the data. Technology and data owners hold access to the data. Unions and works councils need to see that employee data is handled legitimately. Treat these as hypotheses and check them in conversation.
4. 4Secure a committed sponsor for each priority project. The passages treat 'committed sponsors' as part of methodology, with Swarovski's 'The right sponsor' as the case. They also say senior executive sponsorship let early teams at Google, Microsoft and IBM scale solutions. Choose the sponsor who owns the business problem, not simply the most senior person available.
5. 5Pick and prioritize work by its likely impact on the desired outcomes, and say so to stakeholders. The answer to 'why this initiative?' is that it delivers value to results the stakeholder already cares about. Do not justify work by interest or by the volume of data available.
6. 6Write the stakeholder plan: for each group, state the objective, the owner on your team, the cadence of contact and the evidence of buy-in. The book dedicates a section to 'Building a stakeholder plan'. Syngenta's case is titled 'Business stakeholders are critical for success'. Review the plan on a regular rhythm and change it when priorities move.
7. 7Return insight to managers and employees, not only to executives. Microsoft's CHRO describes putting data in the hands of employees and managers so they can self-correct ineffective behaviors and amplify positive ones. This turns two groups usually treated as data subjects into participants.
8. 8Handle ethics and governance openly with the groups most exposed to the data: employees, unions and works councils, and data owners. The Lloyds case is titled 'Ethics, ethics, ethics'. Bring them in before the analysis, not after it.

Microsoft's Chief People Officer describes a partnership between the HR Business Insights team and the sales organization. The team identified behaviors that make sales professionals more productive, which influenced sales outcomes. Read through the stakeholder plan, the pattern is clear. The outcome (sales results) belonged to a business stakeholder, not to HR. The work therefore started from a business result and not from an HR metric, in line with the foreword's principle that 'HR is not about HR, but about the business'. The same source says Microsoft gives managers and employees their own data so they can correct their behaviors. That addresses the manager and employee groups directly. It also shows a stakeholder group being given something useful, not just being consulted. Note the limit of the passages: they give the J&J, Viessmann and Syngenta case titles and contributors, but not the content of those cases. Take the mechanics here from Microsoft and the foreword, and read the three chapter cases for detail.

• If you cannot state the stakeholder's desired outcome in their own terms (employee, strategy, customer, financial or community results), then do not present analysis yet. Ask the question first.
• If a project has no sponsor who owns the business problem, then do not start it. The sponsor makes it possible to scale the solution.
• If a stakeholder group is on your list of seven but has no named owner or contact, then treat it as an open risk. Fill it before the work reaches that group.
• If the work uses employee data in a way employees, unions or works councils would notice, then involve them and the data owners before the analysis begins.
• If two requests compete, then rank them by relative impact on the stakeholder outcomes, and explain that ranking to the people whose request is lower.

• Serving only the executives. This echoes the early 'white-glove' function that worked for a handful of senior leaders. The tell: managers, employees and functional leads first hear about the work when results are announced.
• Presenting HR activity metrics, such as scorecards or dashboards, as if they were value. The tell: stakeholders ask 'so what?' or cannot connect the output to a business result.
• A sponsor in name only. The tell: projects stall at the point of action, and nobody with authority is asking for the findings.
• Leaving out technology and data owners until late. The tell: the analysis is designed but cannot get the data, or the data arrives late or in the wrong form.
• Treating employees and employee representatives as an afterthought on ethics. The tell: concern about data use first appears after the work is visible, and the work is paused or questioned.

• ▸Stakeholder group and named person: choose one of the seven groups (executives, HR leaders, managers, employees/workers, functional, technology/data owners, unions/works councils/employee groups) and name the individual or body.
• ▸Desired outcome they own: state it as a result in the employee, strategy, customer, financial or community area.
• ▸What they need from us: one line, written as a hypothesis to test.
• ▸What we need from them: for example sponsorship, data access, a decision, or consent.
• ▸Sponsor status: committed sponsor, potential, or none. Name the person.
• ▸Our impact-based case: one sentence on why this initiative moves their outcome.
• ▸Contact owner and cadence: who on your team holds the relationship and how often you meet.
• ▸Ethics or governance concern: what they may worry about regarding data use, and how you will address it before the work starts.
• ▸Evidence of buy-in and next review date: the observable sign that alignment exists (for example a decision taken or data released) and when you recheck.

1. 1Audit the team against the four domains in the definition: business acumen, consulting, data science, communications. Score each person 1-5 per domain, with one piece of evidence per score, such as a stakeholder meeting they led or a model they built. Names without evidence make the audit opinion, not data.
2. 2Compare the audit to the work actually being done. List every active request and tag it with the Complexity-Impact Matrix: Quick Win, Big Bet, Trivial Endeavour, Pet Project. If most effort sits in Trivial Endeavours (data requests, reporting, dashboards), the team is in the reporting role the source says most teams remain stuck in. The skill gap is then business acumen and consulting, not statistics.
3. 3Build business acumen by listening before analysing. Copy the Syngenta approach: in the first 100 days, hold structured conversations with business executives and HR leaders, ask open questions, absorb the stories, and only then extract themes. Do not react in the room.
4. 4Put consulting discipline in the intake process. Every request must start with the business problem (Rasmussen and Ulrich: 'Start with the business problem'). Require a named sponsor and the decision the work will inform before any analyst time is committed.
5. 5Staff and run a small prioritization forum: an HR executive, a senior business representative with an enterprise view, a finance executive, and the people analytics leader guiding it. The forum agrees prioritization criteria, picks Quick Wins and Big Bets, reviews progress, and re-confirms routinely that selected projects continue. This puts consulting and communication skill into a structure instead of leaving it to individual heroics.
6. 6Give the leader the data-governor role as well as the analytics role. The leader sets a people data strategy covering all data, business value, protection and innovation. Expect the leader to act as a 'doer' early (vision, foundations) and as a 'facilitator' once capabilities exist.
7. 7Train HR professionals outside the team to have an analytical mindset. This is the second Rasmussen and Ulrich recommendation. Capability that sits only inside the analytics team limits what the business can use.
8. 8Make the operating model both structured and flexible. The Merck & Co. lesson is that people analytics requires structure and flexibility at the same time. A defined prioritization approach lets the team pivot to new business requirements, as in the pandemic, without dropping its commitments. Note: the Skills chapter's detailed operating models (Standard Chartered leader, Capital One scaling, Royal Caribbean translators) are not in the excerpt, so read them before finalizing roles.

Madhura at Syngenta joined as the people analytics leader and treated her first 100 days as a capability exercise. She met approximately 70 professionals and executives, 15 of them business executives who gave her the deepest insight into how Syngenta operated. Her stated key skill was listening: ask insightful open questions, absorb the stories, extract the themes. That produced four business-first topics: identify the skills and capabilities needed to win in the market; focus on productivity to fund growth amid industry consolidation; enable collaborative, team-centric ways of working through technology and analytics; and foster a culture of diversity and inclusion. Near the end of the 100 days, in February 2019, she met the HR leadership team and ran real-time polling to prioritize the topics. This secured buy-in and also showed the power of real-time analytics. By March she had a 'business-first' strategy with full HR leadership buy-in. The skills on display are business acumen (executive conversations), consulting (theme extraction, stakeholder alignment) and communication (the live poll), all before any advanced modelling.

• If more than half of active work is data requests, reporting and dashboards, then treat consulting and business acumen as the priority hires or development, not more data science.
• If a request has no sponsor beyond a single influential individual, then route it through the prioritization forum before committing analysts.
• If the team has been operating for a while and the foundations exist, then shift the leader's data-governance posture from doer to facilitator.
• If a new business priority arrives mid-year, then re-rank against the Complexity-Impact Matrix in the forum instead of absorbing it informally.
• If HR leaders outside the team cannot read or question an analysis, then invest in analytical-mindset training for them, not only in analyst skills.

• Hiring for technical depth only. Tell: the team produces rigorous analyses nobody asked for, and the backlog is mostly Trivial Endeavours and Pet Projects.
• One powerful stakeholder commissions work alone. Tell: a project like the six-month compensation analysis that consumed analysts entirely and was cancelled when the leader disliked the findings, with no cross-functional discussion.
• Skipping the listening step and arriving with a pre-built agenda. Tell: the strategy topics do not match what business executives say they need, and the HR leadership team does not buy in.
• Rigid structure with no flexibility. Tell: the team cannot pivot when the business changes sharply, as in the pandemic, because prioritization is fixed.
• Capability confined to the analytics team. Tell: HR professionals treat analytics as a service desk, and the team stays in a reporting role.

• ▸Team member / role: name each person or planned role.
• ▸Business acumen score (1-5) and evidence: cite a specific executive conversation or business-first outcome.
• ▸Consulting score (1-5) and evidence: cite a request reframed around a business problem, or a sponsor secured.
• ▸Data science score (1-5) and evidence: cite a method or model delivered.
• ▸Communications score (1-5) and evidence: cite a presentation, poll or narrative that drove action.
• ▸Current portfolio mix: count of projects in Quick Win, Big Bet, Trivial Endeavour and Pet Project.
• ▸Forum membership: names for the HR, business and finance seats and the leader's guiding role.
• ▸Biggest capability gap vs. portfolio: the domain whose weakness most limits Quick Wins and Big Bets.
• ▸Action and owner: the hire, training or structure change that closes the gap, with a review date.

1. 1Start with stakeholders, not tools. The source is explicit that decisions on skills, operating model, data and technology come only after you meet stakeholders and understand the mission and current business challenges. Write down the business problems the stack must serve before opening any vendor conversation.
2. 2Make the build-versus-buy call per capability, not once for the whole stack. The chapter treats 'build' (Bosch: a technical architecture for workforce planning), 'buy' (Vertex: more specialized analytics technology) and 'scale and productize' (Microsoft: analytics across the enterprise) as separate decisions. List each capability you need (data gathering, analysis, visualization, productized solutions) and assign each a build or buy path with a stated reason.
3. 3For any buy, score the vendor on the same four factors Vertex uses: purpose, price, safety and post-sale customer service. Add the source's screening questions: failures the product has had and what was learnt, the customer and technical support provided, integration with your core systems, and the vendor's financial stability if it is a small firm.
4. 4Test for 'real' partnership before signing. Ask whether you can influence the product road map, what support you get during implementation and after go-live, and whether raw data is available at no extra cost if the vendor gathers data. Treat the data as the personal data of employees and the collective data of the company, available for analysis and integration once access, security and disclaimers are in place.
5. 5Run a pilot first. Implement the technology in only a portion of the company, analysing a fraction of the data you will eventually need. Use the pilot to judge the vendor and the solution, and use any benefits it shows to build the larger investment case.
6. 6Settle privacy before scale. Understand the vendor's approach to data security, privacy, storage and ownership. Bring in the company's data privacy team and ethics council (from the Governance dimension) where needed. Prefer vendors willing to co-design policies and processes on ethics and use of their technology.
7. 7Design for scale and productization from the start. Ask how the solution moves from a one-off analysis to something delivered across the enterprise, and how emerging technologies speed data gathering, analysis, insight and the democratization of data. Microsoft's case, 'scaling analytics across the enterprise', is the reference point.
8. 8Check the stack against the three outcomes the chapter ties to technology: business value, employee benefit and HR's credibility. Drop or rework any component that cannot name one.

Vertex Pharmaceuticals' People Strategy and Analytics team treats buying people analytics technology as 'an art and science'. Its rule is to consider purpose, price, safety and post-sale customer service, and its key message is 'Don't buy technology in a vacuum'. The source gives no figures for Vertex, so the sequence below is illustrative, assembled from the chapter's own criteria. First, state the purpose in business terms, drawing on stakeholder conversations of the kind Johnson & Johnson's Piyush Mathur describes (60-plus business and HR leaders in his first 90 days). Second, put the shortlisted vendor through the screening questions: integration with core systems, failures and lessons learnt, support, and financial stability. Third, check the three partnership areas: road-map influence and go-live support, raw-data access at no extra cost, and privacy and ownership. Fourth, bring in the privacy team and ethics council on data security and storage. Fifth, pilot in a portion of the company on a fraction of the data. Sixth, use the pilot's realized benefits to make the case for wider investment, or walk away if the vendor fails the pilot.

• If you cannot name the stakeholder problem the tool solves, then do not shortlist vendors yet. Go back to stakeholders first.
• If a vendor will not run a limited pilot, then treat that as weak partnership evidence and raise the bar on every other criterion before committing.
• If the vendor gathers data but will not release raw data (or charges extra for it), then do not buy. Employee data is personal and company data and must be available for analysis and integration.
• If the vendor is a small firm, then ask for evidence of financial stability, because the source lists it as an explicit screen.
• If a privacy, storage or ethics question is unresolved, then involve the data privacy team and ethics council before go-live, not after.
• If the need is a specialized analytic capability, then lean to 'buy' as Vertex does. If the need is a bespoke architecture such as workforce planning, then consider 'build' as Bosch did. Either way, state the reason in writing.

• Buying in a vacuum: the tool is selected without stakeholder input. The tell is a purchase with no named business problem and low adoption after go-live.
• Judging only on price or demo. The tell is that support after go-live, failure history and integration with core systems were never asked about.
• Vendor lock-out of your own data. The tell is that raw data is unavailable, or costs extra, when you try to integrate it with other sources.
• Skipping the pilot and committing enterprise-wide. The tell is that the investment case rests on vendor claims rather than benefits realized in a portion of the company.
• Treating privacy as a late legal check. The tell is that the privacy team and ethics council first see the vendor's data security, storage and ownership terms after the contract is signed.

• ▸Business problem and stakeholder: name the problem and who asked for it, from stakeholder meetings.
• ▸Capability and path: data management, analysis, visualization or productization, and whether it is build, buy or scale. Give the reason.
• ▸Purpose fit: how the tool addresses the stated problem.
• ▸Price: total cost to buy and run, as quoted.
• ▸Safety and privacy: vendor approach to data security, privacy, storage and ownership. Record privacy team and ethics council sign-off.
• ▸Post-sale service: support during implementation and after go-live, plus customer and technical support provided.
• ▸Integration and track record: fit with core systems, known failures and what the vendor learnt, and financial stability if a small firm.
• ▸Partnership terms: road-map influence, and raw-data access at no extra cost.
• ▸Pilot plan and decision: the portion of the company and fraction of data in scope, the benefits expected, and a go or no-go note for the wider investment case.