Measuring HR Effectiveness: Beyond Administrative Efficiency and Activity-Based Metrics

Any HR activity can be judged by two different questions. One asks how well the activity used resources: how many people were trained, at what cost per head, over how many days. The other asks whether the activity changed what it was meant to change: whether trainees acquired a skill, whether they use it, and whether their work improved as a result. The first is a question of efficiency, the second of effectiveness. Both are legitimate. The problem this article addresses is that organizations often answer the first and report the answer as though it settled the second.

The HR version of the distinction has a well-known vocabulary. John Boudreau and Peter Ramstad (2002, p. 10) organized HR measurement around three "anchor points." Efficiency asks, "What is the level and quality of HR practices we produce from the resources that we spend?" Effectiveness asks, "What is the relationship between our HR practices and the quality of our talent pools?", which includes whether practices change employees' capability, opportunity and motivation to contribute. Impact asks how changes in the quality of particular talent pools affect the organization's competitive success. The third question exists because the same gain in skill is worth far more in some roles than in others. Their full framework, which links the anchor points through a chain of intermediate elements, is set out in Beyond HR (Boudreau and Ramstad, 2007) and is not reproduced here.

The examples illustrate Boudreau and Ramstad's three anchor points; the table is this article's, not theirs.

Reading across any row, the measurement gets harder. Efficiency data usually sit in HR's own records. Effectiveness data have to be gathered from managers, assessments or performance systems, often months after the activity. Impact requires a view of how the business creates value and which roles are pivotal to it. Each column to the right is more informative and less certain.

The claim that HR work could be measured at all was contested within living memory. Jac Fitz-enz, who had held senior corporate HR positions, argued in a 1978 Personnel Journal article, "The Measurement Imperative," that HR activities and their effect on business results could be quantified, against the conventional view that HR could not be measured (Caudron, 2004). His 1980 article in Personnel, "Quantifying the human resources function," developed the case in operational terms. He founded the Saratoga Institute in California in the late 1970s (the institute's own materials give 1977, while a 2004 profile gives 1980) and built what was described as the first international HR benchmarking service: standard definitions of HR measures, collected from many organizations, so that a company could compare its cost per hire or HR staffing ratio with others. PricewaterhouseCoopers bought the institute from Spherion in March 2003, and Saratoga benchmarking continued under PwC.

Reducing this contribution to "HR metrics" undersells it in two ways. First, standardization was the hard part. A cost-per-hire figure is useless for comparison unless two organizations count the same costs, and that definitional work anticipated formal standards such as the ANSI/SHRM cost-per-hire standard approved in 2012 and ISO 30414, the international guideline for human capital reporting first published in 2018. Second, Fitz-enz did not stop at the HR department's own costs. In The ROI of Human Capital (2000) and The E-Aligned Enterprise (2001) he proposed financial ratios that link workforce costs to business results, of which human capital ROI is the best known.

Those limitations matter for the argument here. Human capital ROI is an outcome ratio, but it is not an effectiveness measure in Boudreau and Ramstad's sense, because it links no particular HR practice to any change in people. It can tell a board what financial return the workforce is associated with. It cannot tell an HR director whether last year's leadership programme helped. Fitz-enz's lasting legacy is the discipline of definition and comparison. Moving from a ratio to a cause is a separate problem, taken up in Section 6.

A quarter of a century after Fitz-enz's first articles, measures had multiplied, but their character had changed less. Edward Lawler, Alec Levenson and John Boudreau (2004) surveyed senior HR executives in large, mostly Fortune 500 companies connected with the Center for Effective Organizations at the University of Southern California. Thirty-seven usable surveys came back, a response rate of 38 percent.

The authors also found that firms with measures of business impact, or of the effect of HR programmes on the workforce, were significantly more likely to describe HR as a strategic partner. Efficiency measures showed no significant difference, "despite the fact that such efficiency measures are the most prevalent metrics used by HR today." Two caveats apply. The authors raised the first themselves: the firms were not a random sample and were probably well above average in their use of metrics. The second is that the headline finding is a correlation between two answers given by the same HR executives. It fits the idea that effectiveness data make HR more strategic. It fits equally well the idea that strategically positioned HR functions are the ones given the access and budget to collect such data. Research on HR measurement is exposed to the same inferential problem it asks HR departments to confront.

The simplest explanation for the dominance of efficiency measures is availability. The philosopher Abraham Kaplan (1964, p. 11) called it "the principle of the drunkard's search," after the story of a man looking for his key under a street lamp, far from where he dropped it, because "It's lighter here!" The joke is older than Kaplan's use of it (the earliest known newspaper version dates from 1924) and now usually goes by the name of the streetlight effect. Boudreau and Ramstad (2002, p. 16) applied it directly to HR outsourcing contracts, which "typically emphasize efficiency measures because they are the most measurable," and stated the point in one sentence: "Measurement availability does not equal measurement usefulness." Their later LAMP framework (logic, analytics, measures and process) takes its name from the same story (Boudreau and Ramstad, 2006; Cascio and Boudreau, 2011).

Availability is only part of the story, and the other parts show why change is slow. The reasons below are this article's interpretation of the sources cited, not a tested model.

• System design. HR information systems record transactions: requisitions opened and closed, courses attended, grievances logged. Performance and financial data usually belong to other functions.
• Accounting inheritance. Wayne Cascio and John Boudreau (2011) observe that HR measures exist mostly where accounting systems need information "to control labor costs or to monitor functional activity." Measures designed for cost control answer cost questions.
• Control. An HR department controls its own processing time. It does not control whether a trained supervisor's line becomes more productive, so measuring the second exposes it to results it cannot deliver alone.
• Timing. Efficiency results arrive this month. Effectiveness results may take a year, by which time the next budget has been set.

None of these reasons is irrational. Together they produce reporting that is accurate about HR's workload and silent about its value.

Activity measures would be a contained problem if they stayed passive. They seldom do. Writing about British monetary policy in 1975, Charles Goodhart observed that "any observed statistical regularity will tend to collapse once pressure is placed upon it for control purposes" (Goodhart, 1984). The anthropologist Marilyn Strathern (1997, p. 308), writing about audit in British universities, gave the idea its familiar form, "When a measure becomes a target, it ceases to be a good measure," and credited the label "Goodhart's law" to the educationalist Keith Hoskin. The methodologist Donald Campbell had reached a similar conclusion about public programmes: the more a quantitative indicator is used for social decision-making, "the more subject it will be to corruption pressures and the more apt it will be to distort and corrupt the social processes it is intended to monitor" (Campbell, 1979).

HR measures show the pattern clearly because most efficiency targets have a cheap route to them. Boudreau and Ramstad (2002, p. 12) gave the recruitment example: "cost-per-hire can be reduced by cutting the number of selection activities, but this may well reduce validity and subsequent workforce quality." In the same way, time to fill can be shortened by lowering the hiring bar, and a training-hours target met with long sessions of little use. Article 10 examined the wider form of this problem through Steven Kerr's "folly of rewarding A, while hoping for B." The HR form deserves separate attention, because the people who set HR targets are often the people best placed to meet them cheaply.

A dated Bangladesh example: training hours in the Annual Performance Agreement. From FY2014-15 to FY2024-25, Bangladesh's ministries and divisions, and the offices beneath them, signed an Annual Performance Agreement (APA) each year. The Cabinet Division's guideline for 2024-25 allowed agreements to include learning sessions for officers of grade 9 and above and 60 hours of training a year for employees of grade 10 and below (p. 7). To claim achievement on a training indicator, an office submitted a signed summary sheet giving the dates, content, person-hours spent, number of participants and office order number (p. 43). Achievement below 60 percent of a target scored zero (p. 38).

Every item of that evidence records activity. Nothing asks whether trainees learned anything or whether an office served citizens better afterwards; in Donald Kirkpatrick's four levels of training evaluation (reaction, learning, behaviour and results), the record stops before the first. The guideline did guard against one form of gaming, forbidding the same training from counting under more than one indicator (p. 43). But a rule that awards nothing below 60 percent of an hours target pushes offices toward sessions that are easy to schedule and count, which is the pressure Campbell described. This is an analysis of the guideline's design, not evidence about how particular offices behaved.

In October 2025 the Cabinet Division replaced the APA with a Government Performance Monitoring System (GPMS) built on three-year strategic plans and annual performance plans (The Business Standard, 2025; Prothom Alo, 2025). Press reports did not say how training will be evidenced under GPMS, and its 2025-26 guideline was not reviewed here. Whether it moves from person-hours toward effects is a fair question to put to it.

Suppose an organization does measure effectiveness. Lines run by trained supervisors turn out to be more productive, or units with high engagement scores report higher profits. Has the HR intervention caused the result? Research on HR and firm performance shows why the honest answer is usually "not yet shown."

One of the most cited studies in the field, by Mark Huselid (1995), surveyed 968 US firms (a 28 percent response rate) on their use of high performance work practices: comprehensive recruitment and selection, incentive compensation, performance management, employee involvement and training. A one-standard-deviation increase in these practices was associated with a 7.05 percent relative fall in turnover and, per employee, $27,044 more in sales, $18,641 more in market value and $3,814 more in profits. Huselid corrected for selectivity and simultaneity biases and summarized the figures as associations, yet they were widely repeated as returns on HR investment.

Patrick Wright, Timothy Gardner, Lisa Moynihan and Mathew Allen (2005) tested the assumption that HR comes first. Using 45 business units of a large food-service company in the United States and Canada, they correlated a measure of HR practices with operational performance before, during and after the period in which practices were measured. The correlations were "both high and invariant" across all three periods, and controlling for past or concurrent performance "virtually eliminates the correlation of HR practices with future performance." The sample was small and came from one company, and the authors did not claim to have proved reverse causation. What they showed is that the usual evidence cannot rule it out: profitable units may be the ones that can afford better HR practices, and well-led units may do both.

Three alternative explanations deserve a hearing whenever an HR measure is offered as proof of impact:

1. Reverse causation: performance produces the HR practice, as when successful units fund more training.
2. Selection: people who receive an intervention differ from those who do not, as when managers nominate their most promising staff for a development programme.
3. Confounding: a third factor drives both, as when a new unit head introduces better HR practices and also fixes the supply chain.

The designs that address these threats are well established: random assignment, staggered roll-outs, before-and-after comparison against a comparison group, and checks on whether larger doses produce larger effects. They are slower and more demanding than a dashboard, and Module 16 (Articles 76 to 80) develops them properly. The point for now is conceptual: an effectiveness measure, however well chosen, is still a measurement, and whether it shows an effect depends on the comparison behind it.

Bangladesh offers an example of an HR intervention tested by randomized experiment. Rocco Macchiavello, Andreas Menzel, Atonu Rabbani and Christopher Woodruff (2026) worked with 24 large garment factories around Dhaka, all suppliers to one large UK-based buyer. In the sewing sections of these factories, women made up more than 80 percent of workers but only 6 percent of line supervisors. Factories nominated equal numbers of male and female sewing operators as candidates for promotion, and 145 (73 women and 72 men) completed a six-week supervisor course designed by the German development agency GIZ, in rounds that began in 2014. Each trainee then spent two months as a co-supervisor on a line drawn at random from those the factory had nominated. The study is real and peer-reviewed; the reading of it below, for HR measurement, is this article's.

The results expose the weakness of both activity counts and quick effectiveness checks. The course ran six days a week in eight-hour sessions, close to 290 classroom hours per trainee by simple arithmetic, yet a before-and-after test found that it raised neither women's nor men's knowledge. Whatever the programme achieved, an hours figure would not have captured it; what it changed was the factories' willingness to try women in the role. A two-month efficiency check would have misled in the opposite direction. During the trial, lines randomly assigned a female trainee were about 7 percent less efficient than lines assigned a male trainee, a gap the authors linked largely to co-workers' negative beliefs about women as supervisors. A factory judging the programme at that point would have concluded that women make weaker supervisors.

The longer comparison reversed that reading. Factories kept 53 percent of the female and 67 percent of the male trainees as supervisors, and retained women performed similarly to retained men; an earlier version of the paper reported the initial gap closing within four to six months (Macchiavello et al., 2020). Factories that joined an extension of the programme went on to raise the share of women among newly appointed supervisors permanently, compared with similar factories that did not take part.

The study has limits: one buyer's suppliers, one role and one country, a trial that could not use a random set of lines, and incomplete compliance with assignment, which the authors handled by comparing lines as originally assigned. Its value here lies in the contrast with the APA record. One system counts person-hours and stops. The other compares like with like and waits long enough for the answer to settle. Few organizations can run an experiment of this size, but many could randomize the order in which units receive a new programme.

In defence of efficiency. Nothing above makes efficiency measures unimportant. HR administration costs real money: Lawler and colleagues, citing Fitz-enz, put typical HR administrative cost at $1,200 to $1,600 per employee at the time, "a large enough cost" to be worth measuring well. Payroll accuracy, the speed of filling critical roles and the cost of HR services matter to employees and line managers, and an HR function that cannot run them reliably will not be trusted with larger questions. Boudreau and Ramstad treated efficiency as a legitimate anchor point. The problem is substitution: efficiency measures reported as though they answered effectiveness questions.

Against more measurement. The opposite objection comes from critics of metric culture. The historian Jerry Muller (2018) argues that faith in numerical indicators over experienced judgement, and in rewarding measured performance, persists despite its unintended negative consequences, and that the real question is "metrics as informing judgment, which includes knowing how much weight to give to metrics, recognizing their characteristic distortions, and appreciating what can't be measured" (quoted in Kinni, 2018). On this view, replacing activity targets with effectiveness targets merely moves Goodhart's problem to a new measure: a target for skill-gain scores can be gamed as readily as a target for training hours. The objection is sound, and it supports a distinction this article would defend. Effectiveness evidence is most useful for learning, that is, deciding which programmes to keep, change or stop. It is most fragile when it becomes a personal target for the people who produce it.

Rate each statement for the HR report your senior team actually receives, such as a monthly dashboard or an annual HR review. Totals range from 8 to 40. The statements and score bands are this article's teaching tool, based on the distinctions discussed above. They are not a validated instrument.

Reading the result. Statements 1 to 3 concern what is measured, statements 4 and 8 how claims are supported, and statements 5 to 7 whether the measures are used and protected from gaming. A report can score well on the first group and poorly on the second, which is a common way for an effectiveness dashboard to mislead. If HR staff and line managers rate the same report separately, compare their answers: a wide gap is itself a finding worth discussing.

This case is fictional and was constructed for teaching. The company, people and figures are invented.

Karnaphuli Pharma Distribution, a mid-sized distributor based in Chattogram, employs 120 territory sales representatives. Last year its HR team ran a six-month sales coaching programme, and regional managers nominated 40 representatives to take part. At year end the HR head presents the results to the executive committee: sales in the coached representatives' territories grew 14 percent, against 6 percent for the other 80. "The programme added eight points of growth," she says, and proposes extending it to every representative at a cost of Tk 1.2 crore. The finance director asks for your view before the vote.

What the exercise is designed to show. The HR head's claim combines a measurement (growth differed between two groups) with an inference (coaching caused the difference). A useful response separates the two, tests the cheapest alternative explanation first, and turns the next spending decision into a fair comparison.

Module 03 has examined the foundations of HR management: what the function does, how it is organized, and the activities on which its reputation rests. This article closes the module with a question about evidence. The measurement tradition that Jac Fitz-enz began made HR work countable and comparable, a real achievement that still underpins sound HR administration. It did not, and on its own could not, show which HR activities change people's capability or the organization's results. Efficiency measures dominate because they are visible, controllable and quick. They distort once they become targets, as the APA's person-hour evidence illustrates. And moving from an effectiveness measure to a claim of impact requires a comparison that most HR reports lack, as Wright and colleagues showed for the HR-performance research itself. The Bangladesh garment experiment shows that such a comparison can be made in practice.

The unresolved issue is organizational more than technical. Testing whether a programme works means accepting that it might not, and HR functions are often judged by the programmes they launch. David Angrave and colleagues (2016) warned that, without better methods, current HR analytics practice could seal HR's exclusion from board-level influence while doing little for organizations and actively damaging employees' interests. The methods that answer the effectiveness question are the subject of Module 16 (Articles 76 to 80), and Article 20 considers how a strategic HR scorecard can be built around them. The question to carry forward can be asked of any HR number: what would this figure look like if the programme had not worked?

Select an organization whose HR reporting you can examine: your employer (which you may disguise), or a company or public body that publishes human capital information in an annual, sustainability or performance report.

Prepare a report of 2,500 to 3,000 words that:

1. lists the HR measures the organization reports to senior management or the public, and classifies each as efficiency, effectiveness or impact, explaining any borderline cases;
2. identifies at least two measures that could be improved in ways that damage what they are meant to capture, describing the mechanism and any safeguards already in place;
3. selects one HR programme the organization claims has improved a business result and evaluates the evidence for that claim, naming which alternative explanations (reverse causation, selection, confounding) the evidence does and does not rule out;
4. designs a feasible evaluation of that programme that would give stronger causal evidence, stating the comparison group, outcome measures, timescale, cost and likely resistance; and
5. recommends changes to the reporting, including at least one measure to remove or demote, and explains how the revised measures would be protected from becoming distorting targets.

Use at least five academic sources, including at least three from this article's reference list. Do not present any association between an HR measure and a business result as proof of effect without stating the comparison that supports it. Where data are unavailable, say so and explain how the gap limits your conclusions.

Assessment rubric (100 marks)

1. Boudreau, J. W., & Ramstad, P. M. (2002). Strategic HRM measurement in the 21st century: From justifying HR to strategic talent leadership (CAHRS Working Paper 02-15). Cornell University, Center for Advanced Human Resource Studies. https://hdl.handle.net/1813/77386
2. Campbell, D. T. (1979). Assessing the impact of planned social change. Evaluation and Program Planning, 2(1), 67–90. https://doi.org/10.1016/0149-7189(79)90048-X
3. Fitz-enz, J. (1980). Quantifying the human resources function. Personnel, 57(2), 41–52. PubMed record: https://pubmed.ncbi.nlm.nih.gov/10246466/
4. Fitz-enz, J. (2001). The e-aligned enterprise: How to map and measure your company's course in the new economy. AMACOM. Formulas as reprinted in "Formulas for HR's effect on business results," Workforce, 11 December 2001: https://www.workforce.com/news/formulas-for-hrs-effect-on-business-results
5. Goodhart, C. A. E. (1984). Problems of monetary management: The UK experience. In Monetary theory and practice: The UK experience (pp. 91–121). Macmillan. https://doi.org/10.1007/978-1-349-17295-5_4 (first presented in 1975 to a Reserve Bank of Australia conference)
6. Huselid, M. A. (1995). The impact of human resource management practices on turnover, productivity, and corporate financial performance. Academy of Management Journal, 38(3), 635–672. https://doi.org/10.2307/256741
7. Lawler, E. E., III, Levenson, A., & Boudreau, J. W. (2004). HR metrics and analytics: Use and impact. Human Resource Planning, 27(4), 27–35. Figures checked against the working-paper version, USC Center for Effective Organizations Publication G04-8 (460): https://ceo.usc.edu/wp-content/uploads/2004/05/2004_08-g04_8-HR_Metrics_and-Analytics.pdf
8. Macchiavello, R., Menzel, A., Rabbani, A., & Woodruff, C. (2026). Promoting women to managerial roles in the Bangladeshi garment sector. Econometrica, 94(5), 1685–1716. https://doi.org/10.3982/ECTA18759 (details checked against the authors' December 2025 version, CSAE Working Paper WPS/2026-04: https://ora.ox.ac.uk/objects/uuid:b97d0b14-521c-4fef-91ac-90ff3a6a68f1)
9. Strathern, M. (1997). 'Improving ratings': Audit in the British university system. European Review, 5(3), 305–321. https://doi.org/10.1002/(SICI)1234-981X(199707)5:3<305::AID-EURO184>3.0.CO;2-4
10. Wright, P. M., Gardner, T. M., Moynihan, L. M., & Allen, M. R. (2005). The relationship between HR practices and firm performance: Examining causal order. Personnel Psychology, 58(2), 409–446. https://doi.org/10.1111/j.1744-6570.2005.00487.x

Other sources cited

• Angrave, D., Charlwood, A., Kirkpatrick, I., Lawrence, M., & Stuart, M. (2016). HR and analytics: Why HR is set to fail the big data challenge. Human Resource Management Journal, 26(1), 1–11. https://doi.org/10.1111/1748-8583.12090
• Boudreau, J. W., & Ramstad, P. M. (2006). Talentship and HR measurement and analysis: From ROI to strategic organizational change. Human Resource Planning, 29(1), 25–33.
• Boudreau, J. W., & Ramstad, P. M. (2007). Beyond HR: The new science of human capital. Harvard Business School Press.
• Cascio, W. F., & Boudreau, J. W. (2011). Investing in people: Financial impact of human resource initiatives (2nd ed.). FT Press. Sample chapter: https://ptgmedia.pearsoncmg.com/images/9780137070923/samplepages/9780137070923.pdf
• Caudron, S. (2004, April 1). Jac Fitz-enz, metrics maverick. Workforce. https://workforce.com/news/jac-fitz-enz-metrics-maverick
• Fitz-enz, J. (2000). The ROI of human capital: Measuring the economic value of employee performance. AMACOM.
• Kaplan, A. (1964). The conduct of inquiry: Methodology for behavioral science. Chandler.
• Kinni, T. (2018, April 23). How to cure a bad case of metric fixation [Review of The tyranny of metrics, by J. Z. Muller]. strategy+business. https://www.strategy-business.com/article/How-to-Cure-a-Bad-Case-of-Metric-Fixation
• Kirkpatrick, D. L. (1959–1960). Techniques for evaluating training programs (Parts 1–4). Journal of the American Society of Training Directors, 13(11)–14(2).
• Macchiavello, R., Menzel, A., Rabbani, A., & Woodruff, C. (2020). Challenges of change: An experiment promoting women to managerial roles in the Bangladeshi garment sector (NBER Working Paper 27606). https://www.nber.org/papers/w27606
• Muller, J. Z. (2018). The tyranny of metrics. Princeton University Press.
• International Organization for Standardization. (2018). ISO 30414:2018 Human resource management: Guidelines for internal and external human capital reporting. https://www.iso.org/standard/69338.html (replaced by a 2025 edition)
• "PwC acquires benchmarking concern." Accounting Today, 6 March 2003. https://www.accountingtoday.com/news/pwc-acquires-benchmarking-concern
• "ANSI approves first American HR standard." CBIA, 1 June 2012. https://www.cbia.com/news/hr-safety/ansi-approves-first-american-hr-standard/

Bangladesh policy and news sources

• Cabinet Division, Government of the People's Republic of Bangladesh. Barshik Kormoshompadon Chukti pronoyon, bastobayon, poribikkhon o mullayon nirdeshika, 2024-25 [Annual Performance Agreement formulation, implementation, monitoring and evaluation guideline, 2024-25] (in Bangla), pp. 3, 7, 38, 43. Copy hosted on the government portal: https://objectstorage.ap-dcc-gazipur-1.oraclecloud15.com/n/axvjbnqprylg/b/V2Ministry/o/office-bsb/2024/12/ffb650fa33454fba8e33245c09934973.pdf
• "Govt introduces new employee performance evaluation system GPMS." The Business Standard (UNB), 20 October 2025. https://www.tbsnews.net/bangladesh/govt-introduces-new-employee-performance-evaluation-system-gpms-1265401
• "Govt introduces new employee performance evaluation system GPMS." Prothom Alo (English), 21 October 2025. https://en.prothomalo.com/bangladesh/government/4bmsu5saqb

Verification notes: Goodhart's 1975 wording is quoted as it appears, with page reference, in C. J. Moses, Reserve Bank of Australia Research Discussion Paper 8301 (1983); the 1984 chapter itself was not read. Campbell's (1979) wording is consistent across several secondary sources but the original was not consulted, so no page number is given. Kaplan (1964), Kirkpatrick (1959–1960), Muller (2018) and the 1978 "Measurement Imperative" article are cited through the secondary sources named in the text. The diagnostic, the table in Section 1, the Karnaphuli case and its commentary are original teaching material for this series.

Next in The Human Capital MBA series: Module 04 opens with Article 16.