What one general manager actually wants from HR

The short version

The People function has customers, meaning employees and candidates, and a product, meaning their experience across the whole lifecycle, and it should ladder to one business number: human capital return on investment (HCROI), a required disclosure for organizations reporting under ISO 30414, which compares what the workforce produces against what it costs. Two branches move that number. Talent Growth covers attract, hire, and ramp, with quality of hire as the centerpiece. Talent Compounding covers retain, engage, and develop, anchored by team-level engagement, a rating curve that really moves, and a behavior gate built on humble, hungry, and people-smart, where strong output can't buy back a real miss. A foundation of rewards, service, efficiency, and risk carries the cost side. Three rules keep it working: the dashboard stays diagnostic and off comp, every metric is reported alongside its counter-metric, and leading metrics run monthly while lagging ones run quarterly. Start with Finance on the HCROI definition, then the plumbing, then quality of hire. The rest of the article is the reasoning, the evidence, and the objections.

I run a business unit, with revenue, margin, and a profit and loss statement I answer for every quarter. For years the hardest function for me to evaluate wasn't Engineering or Marketing; it was People. The work clearly mattered. But the scoreboard never connected to anything I ran the business on: engagement was up, attrition was fine, the surveys were green, and still the results weren't there. My best people were quietly burning out carrying everyone else, and nobody could tell me whether the function was making the company better or just busier.

I don't blame HR for that. The measurement was broken, and business leaders like me helped break it by never saying clearly what we wanted. This article is my attempt to say it clearly, drawing on a career at firms from Google and Electronic Arts to Groupon and Coursera, and in the U.S. Navy before that. I think of the People function as a strategic partner to the business rather than a cost center, and what follows is the operating model that makes it easier to understand how it works that way: a key performance indicator (KPI) tree, grounded in the best evidence I could find. If you lead a People function, consider this one general manager's answer to a question you should be asking us more often: what does the business actually want from me?

The People function has customers and a product

Every function I run ladders to the customer. Product turns user experience into recurring revenue. Marketing turns attention into pipeline. The People function fits the same shape. Its customers are current employees, plus the people you're trying to hire. Their experience across the whole lifecycle, from the first job posting they read to the day they leave, is the product. And the logic of the function fits in one sentence: great experience produces great people, and great people make the business better.

It means the function ladders to a business number the same way Product ladders to annual recurring revenue (ARR).

The tree ladders to one number: Human Capital ROI

Let’s get real nerdy for a moment. ISO 30414:2025, the international standard for human capital reporting, defines 69 metrics across 11 areas of workforce measurement, and human capital return on investment (HCROI) is one of its required metrics. The formula for it is:

HCROI = [Total revenue − (Total expenses − Total workforce cost)] ÷ Total workforce cost − 1

The formula expresses how much value comes back for every dollar invested in people, beyond the dollar itself. Workforce costs typically run between 25% and 60% of operating expenses. That makes it the largest controllable investment most companies have, and I'd argue the least instrumented one.

I like this apex for a reason beyond the fact that a standards body did the arguing for me: it's self-balancing. Every feel-good People metric I'd been handed over the years could be improved by overpaying and asking less. Pay top of market while lowering the bar, and retention climbs, engagement climbs, and the economics quietly erode underneath. HCROI resists that. Overpaying swells the denominator, cutting too deep collapses the numerator, and what's left is the exact tension a business leader wants managed: talent quality against talent cost, held in one number.

The claim is easier to trust after working the arithmetic once. Consider a small software firm with $1 million in revenue, $750,000 in total expenses, and $500,000 of that sitting in workforce cost: salaries, benefits, contractors, and yes, the People function's own budget. The bracketed part of the formula nets out every expense except people, so the calculation reads as revenue minus non-people cost, divided by people cost: ($1M − $250K) ÷ $500K − 1 = 0.50. Every dollar invested in people returns $1.50, or 50 cents beyond the dollar.

Now run the same firm forward two ways. If it invests $100,000 in senior hiring and development, and the stronger team lifts revenue to $1.3 million on flat non-people cost, HCROI rises to ($1.3M − $250K) ÷ $600K − 1 = 0.75. Workforce cost went up; the return went up faster. That's what a sound talent investment looks like in this arithmetic. If the firm instead cuts $100,000 of people cost while revenue holds for a year, HCROI jumps to ($1M − $250K) ÷ $400K − 1 = 0.875, briefly the best number of the three. The following year the lost capability surfaces, revenue slips to $800,000, and HCROI lands at ($800K − $250K) ÷ $400K − 1 = 0.375, below where the firm started. In other words, a cut flatters the ratio only until the numerator catches up. This is why the metric is never used alone.

One caveat, the same one I'd give a product leader about ARR: the People function influences HCROI without solely owning it. Pricing moves it, product moves it, markets move it. That's fine. Partial ownership of the right number beats full ownership of a number that connects to nothing.

Underneath the apex, value grows through two mechanisms: bringing great people in, and compounding the ones you have. Both sit on a foundation. Figure 1 shows the whole tree at a glance, and the next three sections walk through it.

Talent Growth: how great people come in

The front half of the lifecycle covers attract, hire, and ramp. Future employees are customers you haven't closed yet, and a strategic partner grows the customer base as deliberately as it serves the current one. To map the signals I borrow the happiness, engagement, adoption, retention, and task success (HEART) framework, which I've written about previously and which Google researchers introduced in 2010 to measure user experience at scale (Rodden, Hutchinson, & Fu, 2010). What I use most is its goal-signal-metric chain, a discipline that forces one question for every number: what goal is this evidence of? Plenty of familiar HR numbers can't answer it. The three below can, and Figure 2 shows the branch in detail.

Quality of hire is the centerpiece. LinkedIn's recruiting research has ranked it the single most valuable hiring KPI for years, and fewer than 40% of organizations track it consistently. Both facts are telling: it matters, and it's hard. The workable construction is a composite of new-hire performance at six to twelve months, first-year retention, and hiring manager satisfaction around day 90. The composite is imperfect, but it serves. The math isn't the hard part; the plumbing is, since it means joining recruiting data to performance data. I recommend building it anyway. It's the only number that tells you whether hiring is raising your talent density or diluting it, and every efficiency metric you already track is silent on that question.

Time-to-productivity is the adoption signal. I measure days from start to full contribution, with the bar defined per role family before the hire starts. Defined afterward, the bar has a way of drifting to wherever the new hire happens to be. Once you measure this, onboarding stops being a checklist someone completes and becomes an outcome someone owns.

Offer-accept rate and candidate experience are the happiness signals. Offer-accept is the cleanest conversion metric available on customers you haven't won yet. Candidate Net Promoter Score (NPS), the familiar would-you-recommend measure applied to your hiring funnel, tells you whether the process builds the employer brand or burns it. I make a point of surveying the people we rejected; they're the larger group, and in my experience the more candid one.

Two metrics are missing from this branch by design: time to fill and cost per hire. They measure efficiency rather than quality, so they live down in the foundation. A recruiting function optimized on speed and cost alone fills seats without filling capability. I've watched it happen.

Talent Compounding: how the people you have get more valuable

The back half covers retain, engage, and develop. Most culture measurement already lives here, so the work is less about adding metrics and more about fixing the ones you have. There are four signals, laid out in Figure 3.

Retention, split into regretted and healthy. I report the two together, always as a pair. Losing someone you fought to keep is a loss; exiting someone below the bar is the system working. One blended attrition number hides both, and it sets a trap for later. The quarter you finally get serious about accountability, healthy attrition rises, and the blended number makes a functioning system look like a retention crisis that someone will then panic and reverse.

Engagement, measured at the team level and weighted toward your best people. I came into this skeptical of engagement scores. They seemed too easy to campaign for and too loosely tied to output, and the evidence changed my mind with one large condition attached. Gallup's Q12 is a twelve-question engagement survey, one of the most studied instruments in the field. The eleventh edition of their meta-analysis, combining results across 347 organizations and more than 180,000 teams, finds a true-score correlation of 0.49 between team engagement and composite performance, with top-half teams roughly doubling their odds of success against bottom-half teams (Gallup, 2024). Evidence in organizational research rarely gets stronger than that. I don't intend to wade into the politics of survey vendors here; any validated instrument that gives you a team-level read works, and Gallup's simply has the deepest evidence base. The signal lives at the team level, rolled up by manager, with a reporting floor so results from very small teams stay anonymous. A healthy company average can hide two struggling high-performing teams, and for the same reason I treat a company-wide employee Net Promoter Score (eNPS) as one input rather than a headline number.

Performance and manager quality, tracked together. On the performance side, I watch whether the rating distribution actually moves right over time, with an inflation check alongside it. A company where 85% of people exceed expectations has a broken system, and everything downstream of it is contaminated. On the manager side, I roll engagement and regretted attrition up per manager, because people leave managers more often than they leave companies, and the Gallup evidence above is largely evidence about what managers do. Google reached the same conclusion from the opposite direction. Project Oxygen began in 2008 as an attempt by a company skeptical of management to test whether managers mattered at all, and the data ended the debate: teams with the most effective managers performed better, stayed longer, and were happier. The behaviors that made those managers effective proved learnable, and technical expertise ranked last among them (Garvin, 2013). The broader lesson sits under every branch of this tree. The People function designs and instruments the system; management makes it happen. Every metric here moves, or doesn't, through a manager's daily decisions. That's why manager quality gets its own signal instead of hiding inside an average.

This branch is also where behavioral standards get teeth, and I'll share how we did it. In his 2016 book The Ideal Team Player, Patrick Lencioni distills what makes someone worth building a team around into three virtues: humble, hungry, and people-smart (HHPS; Lencioni, 2016). Well, Lencioni uses ‘Smart,’ but I prefer the more specific people-smart to differentiate from intelligence. We assess everyone against those three. For senior individual contributors and managers, we layer on the leadership practices Jim Kouzes and Barry Posner documented in The Leadership Challenge, applied at the role level. The key design decision is that behavior operates as a gate rather than a tiebreaker: a real miss on any virtue caps the rating, no matter how strong the output. A tiebreaker leaves your highest-output, lowest-behavior people exactly where they are, since their numbers still carry them. A gate is the only version your best people will believe. A mistake I have made along the way is focusing on results without first taking a team through the stages of team development, and the behavior standard exists partly so the results conversation and the team conversation happen together. The metric that proves the gate is real is values-based exits and non-promotions. The day that number moves off zero is the day the standard moves from aspiration to practice.

Mobility and development, the compounding itself. Here I track internal promotion and mobility rates, development plans that measurably progress rather than merely exist, and skills coverage against where the roadmap is heading. This is what keeps a high-performing culture high-performing next year, not just this quarter.

The foundation: what both branches run on

The foundation holds four deliberately unglamorous cells, shown in Figure 4 with the apex formula underneath. Rewards covers comp competitiveness by level and pay equity, checked annually; the whole tree sits on paying fairly against market. Service delivery is the People team's own product quality: employee satisfaction with HR services and time-to-resolve, measured the way any support function measures itself. Efficiency holds people cost as a percent of revenue, cost per hire, and revenue per full-time equivalent (FTE), the productivity companion to HCROI. Risk means compliance, plus a standard I state to my own team in two words: no surprises.

The foundation carries the cost denominator for the whole tree. If your company ever reports human capital externally, this is also where the ISO-required disclosures mostly live. Its lack of glamour is the point.

Three operating rules: keep it diagnostic, pair every metric, split the cadence

Metrics rarely fail on selection; they fail on operation. I run this tree under three rules, each one closing a failure mode I've watched happen.

The first keeps the dashboard diagnostic: off comp, off individual incentives. The moment any of these numbers pays someone, it gets gamed. A manager comped on retention will hoard an underperformer to protect the number. A team comped on engagement will campaign for survey scores every quarter. We pay people on business results and diagnose with this.

The second pairs every metric with its counter-metric: retention with regretted attrition, engagement with performance movement, hiring speed with quality of hire, cost with capability. Paired up, no single number can be improved without its partner exposing the trade. Any review that shows one without the other is incomplete by definition.

The third runs leading metrics monthly and lagging ones quarterly. Offer-accept, time-to-productivity, low-performer actioning, and development progress all move inside a quarter; they belong on the operating rhythm. HCROI, quality of hire, regretted attrition, and promotion rate confirm whether the leading work compounded. A dashboard built entirely on lagging indicators is driving by the rear-view mirror.

Five objections, answered

An argument this confident deserves pushback, so here is the pushback I would give it myself, and where each objection lands.

"HCROI just hands the CFO a cost-cutting argument." It can, if it's used alone. Cutting workforce cost improves the ratio this quarter, and the value destruction shows up in next year's numerator, after the people who carried it are gone. The arithmetic from the worked example applies here: 0.875 for a year, then 0.375. That lag is why the counter-metric rule exists. HCROI never appears without quality of hire, regretted attrition, and top-performer engagement beside it, and a cost cut that improves the ratio while those three degrade is a loan against next year's numerator, not efficiency. The tree makes the loan visible.

"You're reducing people to numbers." The alternative to measurement is politics rather than humanity. In an unmeasured People function the loudest manager wins calibration, the burnout of your most conscientious people stays invisible, and the toxic high performer keeps getting promoted because nobody can prove the damage. Every metric on this tree exists to protect someone. Regretted attrition protects the people worth fighting for; the behavior gate protects teams from toxic producers; team-level engagement finds the struggling groups a healthy average hides. Measurement is how care scales past the number of people one leader can personally watch.

"Correlation isn't causation. Maybe winning teams just feel more engaged." Partly true. It has also been tested directly: a longitudinal study of 2,178 business units across 10 organizations found employee perceptions drive bottom-line outcomes, with the reverse effect present but weaker (Harter et al., 2010). Even so, I don't need to win that argument. A diagnostic requires signal, not proven causation. When a high-performing team's engagement drops, something changed, and I want to know before the performance follows. That is the entire job of a leading indicator.

"The behavior gate is subjective. You've built a bias vector with a virtuous name." Of the five, this is the objection I take most seriously; "culture fit" has an ugly history of meaning "people like us." Three safeguards keep the gate from becoming that. The virtues carry written, per-level behavioral anchors, so a miss on humble means observable actions rather than a manager's impression. Behavior ratings get calibrated across managers in the same session where output ratings get argued. And the gate's own metric, values-based exits and non-promotions, gets reviewed for patterns: a gate that only ever fires on one demographic is broken and gets treated that way. A standard with anchors, calibration, and pattern review is far safer than the actual alternative, which is the same judgments happening anyway, undocumented, in promotion rooms.

"If nobody's paid on these numbers, why would anyone move them?" Managers are paid on the business results these numbers predict. The tree is the instrument panel; consequence lives in performance management, where it already belongs. Keeping the layers separate preserves the integrity of the numbers. Collapsing them buys you survey campaigning and hoarded underperformers, and both cost more than any incentive gap.

Where to start: Finance first, then the plumbing, then quality of hire

The sequence that's worked for me starts with Finance: settle the HCROI definition first, because what counts as workforce cost determines what the whole tree ladders to. Then stand up the cheap, high-signal splits, such as regretted versus healthy attrition and engagement rolled up by team; these are mostly plumbing in systems you already own. The quality-of-hire pipeline comes third, and I'd start it knowing it won't report cleanly for two quarters. It's the hardest build and the highest-value box on the tree. Targets come last, after baselines, because a target set before a baseline is fiction.

Within a quarter, you and your business leadership can look at this tree together and know, not feel, whether the People function is making the workforce more valuable than it costs.

This is what I'm looking for because it's how I understand great People teams. They measure themselves like a strategic business partner. The tree is what that looks like in practice.

References

Gallup. (2024). The relationship between engagement at work and organizational outcomes: Q12 meta-analysis (11th ed.). Gallup, Inc.

Garvin, D. A. (2013). How Google sold its engineers on management. Harvard Business Review, 91(12).

Harter, J. K., Schmidt, F. L., Asplund, J. W., Killham, E. A., & Agrawal, S. (2010). Causal impact of employee work perceptions on the bottom line of organizations. Perspectives on Psychological Science, 5(4), 378–389.

International Organization for Standardization. (2025). Human resource management — Requirements and recommendations for human capital reporting and disclosure (ISO Standard No. 30414:2025).

Kouzes, J. M., & Posner, B. Z. (2023). The leadership challenge: How to make extraordinary things happen in organizations (7th ed.). Jossey-Bass.

Lencioni, P. (2016). The ideal team player: How to recognize and cultivate the three essential virtues. Jossey-Bass.

Rodden, K., Hutchinson, H., & Fu, X. (2010). Measuring the user experience on a large scale: User-centered metrics for web applications. Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, 2395–2398.