Metrics

September 2026

• Sep 28, 2026ConceptsLeading vs Lagging Indicators ExplainedA leading indicator is a bet: it encodes your theory about what drives the outcome, which is why it has to be both predictive and something you can actually influence. This guide contrasts the two kinds of measurement across timing, actionability and certainty, then covers the three ways leading indicators fail: they get gamed when the number becomes the target, the assumed link to the outcome quietly breaks, and too many of them dilute focus.7 MIN1.1k

June 2026

• Jun 22, 2026ConceptsGoodhart's Law: When a Measure Becomes a TargetCharles Goodhart noticed in 1975 that the Bank of England's stable money-supply relationships broke down the moment it started targeting them, because targeting changed the system that produced the correlation. Marilyn Strathern gave it the phrasing everyone quotes in 1997. Manheim and Garrabrant later split it into four mechanisms, and Choi, Hecht and Tayler found surrogation deepens exactly when pay is tied to the proxy.10 MIN1.6k

February 2026

• Feb 27, 2026ConceptsMeasuring Success: Metrics vs KPIs in OrganizationsGoodhart never said metrics are bad. He said a measure used as a target ceases to be a good measure, and that distinction between a metric that informs and a target that carries consequences is the conceptual hinge. Google Flu Trends shows the cost of missing it: search volume was a defensible proxy for flu prevalence until media-driven searching broke the link, and the system kept reporting confidently. Covers vanity metrics, proxies, validity and reliability.20 MIN14.8k
• Feb 10, 2026ConceptsHow Goodhart's Law Breaks MetricsGoodhart's 1975 observation that a statistical regularity collapses once it is used for control explains teaching to the test, surgeons avoiding high-risk patients, and Compstat crime reclassification. Wells Fargo's cross-selling target produced about 3.5 million potentially unauthorized accounts in the bank's expanded review, $185 million in 2016 fines and a $3 billion settlement in 2020. Also: detection signs and eight harder-to-game metric designs.21 MIN56.1k
• Feb 8, 2026ConceptsKPIs Explained Without BuzzwordsAll KPIs are metrics, but very few metrics are KPIs. Drucker's line that what gets measured gets managed has a corollary: measure everything and you manage nothing well. The 3-7 rule per strategic goal exists because nobody holds more than about seven items in working memory, so if you cannot recall your KPIs without looking them up, you have too many. Leading against lagging indicators, and why counting meetings or lines of code measures activity rather than outcome.18 MIN3.4k
• Feb 6, 2026ConceptsVanity Metrics vs Meaningful MetricsEric Ries defined vanity metrics as numbers that feel good without helping you understand your situation. A startup reporting 100,000 downloads avoids saying 90% never opened the app and 200 people pay. The actionability test is one question: if this metric moves, what do you do differently? Page views up 30% gives no answer; conversion rate down 15% sends you into the funnel. Why they survive, and why context decides.18 MIN20.9k
• Feb 6, 2026ConceptsQuantitative vs Qualitative MetricsThe sharpest illustration is Kalev, Dobbin, and Kelly's study of 708 companies over 31 years: the quantitative data showed diversity training and grievance procedures moved managerial diversity almost not at all while mentoring and task forces did, and the organizational case studies explained why, because programs that generated paperwork without changing decision-making were gamed. Both methods were needed to see it.18 MIN55.3k
• Feb 5, 2026ConceptsWhy Metrics Often MisleadGoodhart noticed every monetary aggregate the Bank of England targeted lost its predictive validity once it became a target, and the pattern repeats. Wells Fargo staff opened an estimated 3.5 million unauthorized accounts chasing cross-selling numbers, and in 2014 the UK Statistics Authority stripped police-recorded crime of its National Statistics designation after an inspectorate report found around 19 percent of crimes under-recorded.21 MIN8.2k
• Feb 5, 2026ConceptsWhat Should Be Measured and WhyA metric deserves tracking only if it aligns with a real goal, is actionable, can be collected reliably, either predicts or measures an outcome, and is not redundant. The article separates outcomes from outputs and drivers, caps each goal at three to seven metrics, and reframes selection through Hubbard's point that a measurement is worth making when the uncertainty it removes is worth more than its cost.20 MIN70.4k

December 2025

• Dec 2, 2025TechnologyUnderstanding Analytics Tools for Your Data NeedsThe usual failure is not missing data but missing analysis capacity, which is why a dashboard nobody opens beats nothing only on paper. The page gives a blunt test for separating real metrics from vanity ones: if this number moved, what would you do differently? If the answer is nothing, stop tracking it. It also covers the GA4 versus cookieless tradeoff, where avoiding consent banners costs you cross-device tracking.21 MIN62.8k

October 2025

• Oct 26, 2025Work SkillsProductivity Metrics: Measuring Impact Beyond ActivityHours worked is the most damaging metric an organization can reward. John Pencavel's study in The Economic Journal found output per hour falls sharply after 49 hours a week and total output barely rises past 55, so a 70-hour week buys roughly the same work with more errors. The article maps three levels of measurement, inputs through outputs to outcomes, and the traps between: Goodhart's Law, the busyness trap and vanity metrics.20 MIN6.9k
• Oct 7, 2025Work SkillsKey Project Metrics That Predict SuccessFlickr's engineering dashboard tracked lines of code committed per week and engineers competed to top the leaderboard. The codebase bloated with code written to boost the metric. Instagram launched in 2010 with a tiny codebase and one metric, daily active users, and within 18 months had more users than Flickr ever reached. Dashboards should favor leading indicators during execution: a board full of lagging ones is a rearview mirror.15 MIN48.0k

August 2025

• Aug 17, 2025IdeasKey Issues in Measurement and MetricsA hospital published ER wait times and they fell dramatically, but patients were triaged faster then waited longer in treatment rooms: the system optimized time to first clinical contact while total treatment time got worse. Charles Goodhart observed in 1975 that any statistical regularity collapses once pressure is placed on it for control. The streetlight effect explains why: lines of code are easy to count, so they stand in for productivity.20 MIN52.5k

July 2025

• Jul 18, 2025ExplainersMetrics and Measurement: Importance in PerformanceWorkers allowed to develop local practices away from supervisory view outperformed those under continuous watch, because observation suppressed the experimentation that produced the best results. Landsberger's reanalysis of the original Hawthorne data points the same way: productivity rose from positive feedback and human attention, not from being monitored. The mechanism is social. This page covers leading versus lagging indicators, vanity metrics and gaming.31 MIN60.3k
• Jul 15, 2025ExplainersMetrics That Destroyed Performance: Case StudiesGoodhart's Law and Campbell's Law play out with near-mechanical predictability here. Wells Fargo's eight-products-per-customer target produced roughly 2 million fake accounts, later revised upward, and $3 billion in fines; NHS hospitals held patients in ambulances so the four-hour clock would not start. Ravitch found high-stakes testing states gaining 22 percentage points on their own tests but only 7 on NAEP, which they did not control.33 MIN38.2k
• Jul 10, 2025ExplainersMetrics Design ChecklistA cross-selling target of eight products per customer was well intentioned, easy to count and catastrophic: Wells Fargo paid $185 million in 2016, later growing past $3 billion. The checklist works through 16 checks in four phases, from goal alignment and a gaming pre-mortem to leading indicators and a retirement process. Flyvbjerg's 258 transport projects across 20 nations make the case: nine in ten overran, by roughly 20 percent on roads and 45 percent on rail.22 MIN46.7k
• Jul 2, 2025ExplainersKPIs vs OKRs: Key Differences ExploredDrucker's MBO had managers and employees set objectives jointly, but it merged the goal with the proof of achievement and left both vague. Grove's requirement in High Output Management was that every objective carry specific measurable key results. That structural separation is why OKRs worked where MBO often failed. KPIs monitor steady-state health with no end date; OKRs are time-boxed and drive change, and conflating them produces systems that do neither well.22 MIN44.6k

June 2025

• Jun 24, 2025ExplainersHow Metrics Influence BehaviorWells Fargo tracked products sold per household as a proxy for engagement and attached bonuses and firings to an eight-product target. Employees hit the number by opening roughly 2 million accounts customers never requested, drawing a $185 million penalty in 2016. The healthcare version is stranger: Dranove and colleagues found in the Journal of Political Economy that publishing New York surgeons' cardiac mortality rates raised overall cardiac mortality.27 MIN3.0k
• Jun 19, 2025ExplainersThe Evolution of Metrics in Science and ManagementThe earliest writing systems, developed in Mesopotamia around 3400 BCE, were accounting records rather than literary documents: clay tablets from Uruk logging grain, livestock and labor. Luca Pacioli formalized double-entry bookkeeping in 1494, imposing mathematical discipline on economic activity. The culture of quantification has a specific history with identifiable choices behind it, running through statistics, scientific management, and Goodhart's law.27 MIN42.3k
• Jun 16, 2025ExplainersStep-by-Step: Designing Effective MetricsMetrics shape behavior whether or not they measure the right thing, so gaming follows wherever stakes are attached. A Department of Justice investigation found Baltimore officers downgrading shootings and pressuring victims to withdraw reports under CompStat. Payment can also crowd out motive: Swiss communities offered compensation for a nuclear waste site accepted it at 25%, down from 51% with no money offered.36 MIN38.4k

May 2025

• May 17, 2025TechnologyWhat Is a KPI: Key Performance Indicators ExplainedEngland's four-hour emergency-department target was clear, measurable and time-bound. Then hospitals began holding ambulances in car parks so the clock would not start, and reclassifying waiting areas so it reset. The target was met in the data while care in a number of cases got worse. That is Goodhart's Law, and the real design problem behind KPIs: choosing proxies for strategy, separating leading from lagging signals, keeping the list short.24 MIN21.3k

October 2024

• Oct 22, 2024ConceptsUnderstanding Goodhart's Law and Its ConsequencesThe memorable paraphrase about measures becoming targets came later, from Strathern writing on British university audit in 1997. The page's most useful contribution is a taxonomy separating four distinct failure modes, Regressional, Extremal, Causal and Adversarial, because each calls for a different response. It is also careful to flag the famous cobra bounty story as possibly apocryphal while noting the Hanoi rat bounty is historically documented.21 MIN9.2k

December 2023

• Dec 19, 2023ConceptsAnalyzing Campbell's Law and Its ImplicationsThat distinction decides how bad the damage is. In Atlanta an investigation found 178 teachers and principals across 44 schools altering answer sheets, and NHS hospitals met the four-hour target by holding patients in units that were not counted as emergency care. The page explains why awareness does not protect you, since anyone who reduces gaming is at a disadvantage relative to those who continue, and why removing a metric rarely restores what it reshaped.16 MIN10.2k

February 2023

• Feb 10, 2023TechnologyThe Net Promoter Score (NPS): Does It Work?How NPS is calculated and what the research says about it. Promoters scoring 9 to 10 minus detractors scoring 0 to 6 produces a figure from minus 100 to plus 100, with passives excluded entirely. Reichheld's founding 2003 analysis of roughly 400 companies was a practitioner article rather than peer-reviewed work, and Bain commercialized the system. Keiningham and colleagues found in the Journal of Marketing that NPS held no statistically consistent advantage.19 MIN5.5k

August 2022

• Aug 4, 2022Work SkillsThe Cobra Effect: Risks in Gaming MetricsThat single distinction explains why the same number cannot serve both purposes, and why diagnostic metrics stay honest while bonus-linked ones drift. The NHS supplies the clearest evidence. After four-hour emergency targets arrived, ambulance handovers were delayed at hospital entrances so the clock did not start, and Bevan and Hood found each new target generated fresh gaming within weeks of implementation. The parallel results are Campbell's Law and the Lucas Critique.17 MIN4.5k