High Growth Orgs Build Metrics Models That Measure What Matters

Third, calculate the Expected Value of Information (EVI) for each variable: which measurements would most reduce uncertainty about your highest-stakes decisions. Hubbard’s research shows organisations routinely collect vast amounts of data on low-value questions whilst remaining ignorant about high-value performance drivers. For stalled-growth companies with limited resources, this prioritisation becomes essential. You can’t afford to measure everything, so measure what matters most for restarting growth.

Fourth, measure enough to reduce uncertainty to the point where the decision becomes clear. Hubbard demonstrates you typically need far less data than you think. His “Rule of Five” [4] shows that a random sample of just five observations has a 93.75% chance of falling between the median of the entire population. Five well-chosen data points can dramatically reduce uncertainty about larger patterns.

Fifth, make your decision and reevaluate periodically. Measurement isn’t one-time but continuous: updating your understanding as new information becomes available.

Hubbard’s decomposition approach is particularly powerful for building analytical capability. When something seems immeasurable, break it into components that are more observable. Revenue growth decomposes into: customers acquired, customers retained, average order value, and purchase frequency. Customer acquisition breaks into: leads generated, conversion rate, and sales cycle length. Each of those decomposes further. Continue until you reach things you can directly observe and measure.

According to Hubbard’s research, most “immeasurables” become quite measurable through decomposition. Organisational effectiveness manifests in observable outcomes: project completion rates, decision-making speed, cross-functional coordination time, and execution quality. Customer satisfaction shows up in: usage frequency, feature adoption, support ticket volume, renewal rates, and Net Promoter Score responses. Innovation capability appears in: ideas generated, experiments run, time to market for new features, and percentage of revenue from recent innovations.

Hubbard warns against “measurement inversion”: measuring what’s easy rather than what matters. His research found organisations spending enormous resources measuring low-value factors whilst remaining ignorant about high-value performance drivers. One organisation spent over $1 million measuring software development productivity using lines of code written, a metric with no correlation to business outcomes that actually encouraged worse code quality. Meanwhile, they spent nothing measuring whether projects delivered intended business benefits.

For you, Hubbard’s framework means starting with the decisions: “What should we change to restart growth?” but also “What early signals predict problems quarters before they hit revenue?” and “Are our strategic initiatives moving the drivers we expected?” Then work backward to identify which measurements would reduce uncertainty about those decisions. Don’t measure things because they’re easy to measure or because “everyone measures them.” Measure what would actually change your actions if you knew the answer.

Hubbard’s work on calibrated estimation addresses another critical challenge: overconfidence. His research shows most managers are poorly calibrated, expressing high confidence about things they’re actually quite uncertain about. Through calibration training, managers can express uncertainty more accurately, which reveals where additional measurement provides most value.

The practical application: When growth is not where it needs to be, resist the temptation to immediately launch initiatives based on intuition. Use Hubbard’s framework to identify your highest-uncertainty, highest-value questions. Quantify current uncertainty about potential root causes. Calculate which additional measurements would most reduce that uncertainty. Then design minimal measurements (often much simpler than expected) to test which factors actually drive the stall.

But Hubbard’s framework isn’t just diagnostic. Once you’ve built these decomposed metrics models with validated causal relationships, they become predictive. The same leading indicators that help you diagnose why something went wrong also signal when problems are emerging before they show up in outcomes. According to Hubbard’s research, organisations that build this measurement capability gain enormous strategic advantage. They can intervene quarters earlier because they’re monitoring the right leading indicators rather than waiting for lagging outcomes to deteriorate.

Hubbard’s research demonstrates that breakthrough analytical insights rarely require sophisticated statistics or big data infrastructure. They require clarity about what you’re trying to decide, honest assessment of current uncertainty, and strategic collection of small amounts of highly relevant information. For CEOs working with limited resources to restart stalled growth, this efficiency in measurement is essential.

Eliyahu Goldratt’s Approach: Systematic Root Cause Identification Through Theory of Constraints

Eliyahu Goldratt’s [2] Theory of Constraints, introduced in “The Goal” (1984), provides systematic tools for causal analysis that cut through organisational complexity. According to Goldratt’s framework, most organisational problems aren’t caused by multiple independent issues but by a small number of core conflicts that generate numerous visible symptoms. Organisations waste enormous energy treating symptoms (what Goldratt calls “firefighting”) whilst root causes persist and multiply.

Goldratt’s Thinking Processes provide structured methods for tracing symptoms back to root causes. The Current Reality Tree (CRT) is the primary diagnostic tool. You start by identifying “Undesirable Effects” (UDEs): the symptoms you’re experiencing. For a Growth-Mandate CEO, UDEs might include: revenue growth stalled, customer acquisition costs increased 30%, sales cycle lengthened, win rates declined, and employee turnover increased.

The CRT methodology connects these UDEs through logical cause-and-effect relationships, reading from bottom to top using “if-then” statements. If marketing generates unqualified leads, then sales spends time on poor-fit prospects. If sales spends time on poor-fit prospects, then qualified leads receive less attention. If qualified leads receive less attention, then conversion rates decline. If conversion rates decline, then revenue growth stalls.

Goldratt emphasises that effective root cause analysis must trace back through multiple levels of causation. Most organisations stop too early. They identify an immediate cause and declare victory. But that immediate cause itself has causes. Marketing generates unqualified leads, why? Perhaps marketing doesn’t understand the current ideal customer profile. Why don’t they understand the ICP? Perhaps the ICP evolved as you moved upmarket but that evolution was never explicitly communicated cross-functionally.

The power of Goldratt’s approach [5] is identifying where multiple UDEs converge on a common cause. If you can trace 70% or more of undesirable effects back to a single root cause, you’ve found what Goldratt calls the “core conflict” or “core problem.” Solving that one issue can eliminate the majority of visible symptoms simultaneously, rather than fighting each symptom individually (crucial for CEOs with limited resources).

Goldratt’s “Effect-Cause-Effect” technique provides validation for causal relationships. Once you propose that factor A causes effect B, test that hypothesis by predicting: if A truly causes B, what other effects should we observe? If those predicted effects are present, your causal theory gains credibility. If they’re absent, you may have identified correlation rather than causation.

According to Goldratt’s framework, root causes typically fall into three categories: policies, measurements, or behaviours. Policies are explicit or implicit rules about how work gets done. Measurements are the metrics people are held accountable for. Behaviours are actions people take in response to those policies and measurements. Goldratt’s research showed most performance problems trace back to policies or measurements that made sense in the past but no longer serve current objectives.

For you, this is particularly relevant. The organisation’s current policies and measurements were optimised for a different growth stage or market condition. What worked at £10M revenue may actively constrain growth at £25M. The sales compensation structure designed for land-and-expand might discourage enterprise deals. The decision-making process that enabled speed at 50 employees creates bottlenecks at 150 employees.

Goldratt’s “Evaporating Cloud” technique addresses deeper conflicts beneath visible problems. Organisations often oscillate between competing approaches because they face genuine dilemmas: two needs that seem mutually exclusive. For example: “We need to standardise processes for efficiency” conflicts with “We need flexibility to serve diverse customer segments.” Root cause analysis must surface and resolve these underlying conflicts rather than simply choosing one side.

The Five Focusing Steps (Goldratt’s core methodology for addressing constraints) provides systematic approach to intervention once root causes are identified. First, identify the constraint (the factor most limiting performance). Second, exploit the constraint (optimise how you use that limiting resource). Third, subordinate everything else to the constraint (align other processes to support maximum throughput through the constraint). Fourth, elevate the constraint (add capacity if needed). Fifth, repeat, because once you’ve addressed one constraint, another will emerge.

Goldratt emphasises that constraints shift over time. What limits growth today won’t limit growth tomorrow. Causal analysis must be continuous rather than one-time diagnostic. The analytical systems you build should help identify when new constraints emerge so you can adapt focus accordingly. This continuous monitoring serves multiple purposes: it helps you diagnose new problems as they emerge, provides early warning when leading indicators shift, and tracks whether your strategic interventions are actually removing the constraints they were designed to address.

Goldratt warns specifically against “local optimisation”: improving individual metrics or departments in ways that suboptimise the overall system. His research demonstrated repeatedly that optimising parts in isolation often makes total system performance worse. If you optimise marketing for lead volume without considering sales capacity to handle those leads, you overwhelm sales and reduce overall conversion rates. Causal analysis must examine system-level dynamics, not just individual component performance.

Goldratt’s approach also provides clarity about when to stop analysis and start action. According to his framework, perfect information is never required or possible. Once you’ve identified a root cause that explains 70%+ of your undesirable effects, you have sufficient understanding to act. The remaining 30% can be addressed as you go. Waiting for complete certainty is itself a constraint on organisational velocity (particularly important for Growth-Mandate CEOs who need to show momentum to the board).

For you, Goldratt’s framework suggests your growth rate itself is a symptom: the visible effect of an underlying constraint that has changed or emerged. Your task isn’t to “fix sales” or “improve marketing” but to identify what constraint now limits throughput (revenue growth) when previously it didn’t. The analytical process is tracing visible symptoms through cause-and-effect chains back to that changed constraint, then systematically addressing it.

The Current Reality Trees and cause-and-effect logic Goldratt developed for diagnosis also serve other purposes once built. They become your framework for tracking strategic progress. You can monitor whether interventions are moving the specific causal factors they were designed to move. They provide structure for team health conversations. Many organisational constraints relate to capability, coordination, or cultural factors that show up as early warning signs before they become crises. The same logical rigour that helps you diagnose problems helps you predict where problems will emerge and track whether solutions are working.

Alistair Croll & Benjamin Yoskovitz’s Framework: Actionable Metrics That Reveal Causation

Alistair Croll and Benjamin Yoskovitz’s “Lean Analytics” (2013) addresses a fundamental problem in organisational measurement: most companies drown in data whilst remaining blind to causation. Their framework, built on Eric Ries’s Lean Startup methodology, focuses on the “measure” portion of the Build-Measure-Learn cycle and provides specific guidance on which metrics actually reveal root causes versus which ones simply make you feel good without improving decisions.

According to Croll and Yoskovitz [3], effective analytical capability requires distinguishing between “vanity metrics” and “actionable metrics.” Vanity metrics make you feel good but don’t guide action: things like total registered users, total page views, or total social media followers. These numbers might grow whilst your business deteriorates. Actionable metrics reveal causal relationships and change how you behave based on what they tell you.

Croll and Yoskovitz identify four properties that make metrics actionable. First, actionable metrics must be comparative. You can compare them across time periods, customer segments, channels, or other dimensions to understand what’s driving changes. Total revenue is less actionable than revenue by customer segment by acquisition channel over time, which reveals which segments and channels are improving or declining.

Second, actionable metrics must be understandable. Everyone in the organisation interprets them the same way. Vague metrics like “engagement” or “satisfaction” prevent effective analysis because different people attribute changes to different causes. Clear operational definitions enable shared understanding of what’s being measured and why it matters.

Third, actionable metrics must be ratios or rates rather than absolute numbers. Absolute numbers hide important patterns. Growing from 100 to 200 customers sounds good, but if you’re also growing from 1,000 to 4,000 trial users, your conversion rate actually declined from 10% to 5%, revealing a problem that raw growth numbers masked. Ratios expose causal relationships that absolutes conceal.

Fourth, actionable metrics must change behaviour. They must be tied to specific actions you can take based on what the metric reveals. If a metric going up or down doesn’t change what you’d do next, it’s not actionable. It’s just interesting information consuming attention and resources.

Croll and Yoskovitz emphasise the importance of cohort analysis for understanding causation. Instead of looking at aggregate metrics across all customers, track groups (cohorts) based on when they were acquired, which channel they came from, or which product version they experienced. Cohort analysis isolates variables and reveals what’s truly causal versus mere correlation.

For example, if overall retention rate is 85%, that tells you nothing about causation. But if you analyse by cohort and discover that customers acquired in Q1 have 92% retention, Q2 customers have 88% retention, Q3 customers have 82% retention, and Q4 customers have 76% retention, you’ve identified that something changed over the year affecting retention. Now you can investigate what changed. Perhaps product quality declined, or you shifted to lower-quality acquisition channels, or your ideal customer profile evolved but messaging didn’t.

According to Croll and Yoskovitz’s framework, effective analytical capability requires building metrics trees that connect leading indicators to lagging outcomes. Leading indicators change before lagging outcomes do, giving you predictive power. If reduced feature adoption in the first week predicts churn six months later, you’ve found a leading indicator enabling earlier intervention before problems show up in revenue.

Croll and Yoskovitz introduce the concept of “One Metric That Matters” (OMTM) [6]: the single metric that captures the most critical factor limiting growth at your current stage. Focusing organisational attention on one primary metric accelerates learning about causation more effectively than tracking dozens of metrics simultaneously. The OMTM should be the metric that, if you moved it, would most directly address your current constraint.

For different business models and growth stages, different metrics become the OMTM. In early stages validating whether you’ve found a problem worth solving, engagement metrics matter most. Are people using the product regularly? In growth stages, acquisition and activation become critical. Are you efficiently converting awareness to active usage? In mature stages, revenue and efficiency metrics dominate. Are you generating sustainable unit economics?

Croll and Yoskovitz’s framework involves three specific analytical approaches. First, variance analysis with context. When a metric changes, don’t just note the change; decompose it to understand which segments or dimensions drove the change. “Revenue declined 15%” is a symptom. “Revenue declined 15%: down 20% in Enterprise segment, up 8% in Mid-Market segment, flat in SMB segment” points towards root causes about market positioning or competitive dynamics.

Second, correlation versus causation testing. Croll and Yoskovitz warn that correlation is everywhere but causation is rare. The discipline required for analytical rigour is systematically testing whether apparent correlations represent genuine causal relationships. Just because metric A and metric B move together doesn’t mean A causes B. They might both be caused by C, or the correlation might be coincidental.

To test causation, Croll and Yoskovitz recommend controlled experiments whenever possible. Change one variable whilst holding others constant and observe whether predicted effects occur. This experimental approach moves you from correlation (these things happened together) to causation (this change produced that effect). For Growth-Mandate CEOs with limited resources, small targeted experiments often reveal more about causation than extensive analysis of historical data.

Third, leading versus lagging relationship validation. Powerful analytical capability builds predictive models by identifying which metrics reliably predict future outcomes. Test whether changes in early-stage metrics (like trial user engagement) predict changes in later-stage metrics (like conversion to paid). If they do, you’ve found a causal pathway enabling earlier intervention.

Croll and Yoskovitz emphasise that the value of analytics isn’t in sophistication of analysis tools but in clarity of thinking about what matters. Companies often implement elaborate analytics platforms generating hundreds of reports whilst leaving fundamental questions about causation unanswered. The framework they provide (distinguishing vanity from actionable metrics, using cohorts to isolate variables, building metrics trees connecting leading to lagging indicators, and focusing on the One Metric That Matters) provides intellectual discipline regardless of technical capabilities.

For you specifically, Croll and Yoskovitz’s framework suggests that stalled growth typically shows up first in leading indicators before appearing in lagging revenue numbers. By the time revenue growth has clearly stalled, the causal factors have been in motion for months. Building analytical capability means instrumenting your leading indicators (engagement, activation, conversion quality) so you can identify and address emerging problems before they show up in revenue.

According to their framework, effective analytics also requires honest examination of your assumptions. Organisations often continue measuring what they’ve always measured even after their business model, market, or strategy has changed. The metrics that mattered when you were product-led growth at £5M aren’t necessarily the metrics that matter now as enterprise sales-led at £25M. Sometimes the analytical work reveals you’re optimising the wrong metrics entirely: measuring things that no longer drive the outcomes you care about.

The metrics trees and leading-lagging relationships Croll and Yoskovitz help you build serve multiple purposes beyond diagnosis. Once you’ve validated which leading indicators predict which outcomes, you have a predictive system. You can see problems coming quarters before they hit revenue. When you connect your strategic initiatives to specific metrics in your tree, you have progress tracking. You can monitor whether initiatives are moving the drivers they’re supposed to move. When you include team health metrics in your trees, you have early warning systems for organisational dysfunction before it becomes a retention crisis.

Practical Integration: The Three-Layer Metrics Framework

The uploaded practical guide provides a specific implementation framework that integrates these expert perspectives. The Three-Layer Metrics Framework [7] structures your analytical system:

Layer 1: Outcome Metrics (Lagging Indicators) – Revenue, profit, customer retention, market share. These tell you what happened but not why. This is where Growth-Mandate CEOs see the symptom: “Growth stalled.”

Layer 2: Performance Metrics (Current Indicators) – Sales pipeline velocity, customer satisfaction scores, product quality metrics, conversion rates. These show what’s happening now and begin revealing causation.

Layer 3: Input Metrics (Leading Indicators) – Activity levels, capability metrics, resource allocation, process efficiency. These reveal what causes the outcomes and where intervention opportunities exist.

The key insight: Most organisations only measure Layer 1, then scramble when numbers decline. Your metrics model must connect all three layers causally, which combines Hubbard’s decomposition, Goldratt’s cause-and-effect logic, and Croll and Yoskovitz’s leading-lagging relationships.

Build complete causal chains. Start with your critical outcome metric (revenue growth). Break it down mathematically: Revenue = Customers × Average Order Value × Purchase Frequency. For each component, identify the performance metrics that drive it. For each performance metric, identify the input activities or resources that create it. This creates a testable causal map.

Apply Goldratt’s “5 Whys” approach systematically: When an outcome metric declines, which performance metric(s) changed? Which input metrics correlate with that change? What organisational factors affect those inputs? What decisions or constraints created those factors? What strategic or structural issues drive those constraints?

Use cohort analysis to isolate variables: track groups over time based on acquisition date, channel, or segment. This removes confounding variables and reveals what’s truly causal versus correlational, exactly as Croll and Yoskovitz recommend.

Perform variance analysis with context. Don’t just report “Sales down 15%.” Decompose it: 10% from losing key customer X, 8% from pricing change in Product A, minus 3% from new market segment growth. Now you know where to focus, which aligns with both Goldratt’s constraint identification and Croll and Yoskovitz’s actionable metrics framework.

Test leading versus lagging relationships deliberately: Does employee engagement survey data predict turnover 2 months later? Do early customer health scores predict churn 6 months out? Does pipeline coverage ratio predict quarterly revenue accuracy? Build a predictive dashboard based on validated leading indicators. This is Hubbard’s measurement approach applied to Croll and Yoskovitz’s metrics framework.

Beyond Root Cause: Multiple Applications of the Same Infrastructure

Whilst root cause diagnosis is often the immediate need, the metrics infrastructure you’re building serves four interconnected purposes:

Root Cause Analytics – When performance declines, trace symptoms through your metrics tree to identify what actually changed and why. The causal chains you’ve built become diagnostic pathways. This is typically the urgent capability when you’re six months in and the board is asking why growth hasn’t restarted.

Predictive Analytics – Leading indicators in Layer 3 (input metrics) predict changes in Layer 2 (performance metrics) which predict Layer 1 (outcomes). Once you’ve validated these relationships, you gain quarters of advance warning before problems show up in revenue. This prevents the next stall rather than just diagnosing the current one.

Strategic Progress Tracking – Key Results in your OKR system should map directly to specific metrics in your tree. This creates visibility into whether strategic initiatives are actually moving the drivers they’re supposed to move, not just generating activity. You can see in real-time whether your interventions are working.

Team Health Monitoring – Many Layer 3 input metrics relate to team capability, coordination, and engagement. Build these into your metrics trees and you have early warning systems for organisational dysfunction before it becomes visible in turnover or execution failures. Capability degradation shows up in leading indicators months before it hits outcomes.

The same discipline (decomposing outcomes, mapping causal relationships, validating leading-lagging connections) serves all four purposes. You’re not building separate systems; you’re building one analytical infrastructure with multiple applications. The investment in building this capability compounds over time because each validated causal relationship serves diagnosis, prediction, progress tracking, and health monitoring simultaneously.

Building Organisational Capability

To build this capability in your organisation:

Start with Metrics Trees – Don’t try to map everything. Pick your most important outcomes. Build complete causal chains for just those. Validate the relationships quarterly. This combines Hubbard’s prioritisation (focus on reducing uncertainty where it has the highest value) with Goldratt’s constraint focus (identify the few factors most limiting performance).

Create a Metrics Review Cadence – Monthly: Review Layer 2 & 3 metrics (forward-looking). Quarterly: Deep-dive on variance and causality. Annually: Validate the model itself. Are these still the right drivers? This rhythm enables continuous root cause analysis, predictive monitoring, and strategic tracking rather than one-time diagnostics.

Invest in Analytics Capability – Hire or develop someone who understands both statistics AND your business. Most “data analysts” report numbers; you need someone who interrogates causality using these frameworks. This role typically pays for itself 10x+ through better resource allocation: stopping you from investing in initiatives that address symptoms rather than causes.

Build the Cultural Element – Train leaders to ask “What data would help us understand why?” not just “What are the numbers?” Reward curiosity about root causes, not just hitting targets. Create psychological safety to surface systemic issues. People must feel safe saying “Our ICP evolved but our messaging hasn’t, which is why conversion rates declined” rather than hiding uncomfortable truths.

ZOKRI can support this work as a precursor to OKRs if you want to speed up this process and do it well.

The Key Question

Analytical capability is how organisations move from reactive firefighting to systematic capability building. The companies that break through growth plateaus do this systematically. They build infrastructure that reveals why their organisation performs the way it does: the actual causal drivers rather than mere correlations. That understanding allows you to allocate scarce resources to what genuinely accelerates growth.

But more than that, the same infrastructure prevents future stalls by providing predictive power (see problems coming quarters before they hit revenue), tracks whether strategies are working (monitor if initiatives move the right drivers), and maintains organisational health (early warning systems for dysfunction before it becomes crises).

Which performance outcome matters most to restarting growth right now? Can you trace the causal chain from that outcome back through intermediate drivers to root causes? Do you have actionable metrics that reveal causation, or are you tracking vanity metrics that simply describe symptoms? Are you measuring what actually drives that outcome today, or optimising metrics that worked at a previous growth stage but no longer determine success at your current scale? And once you diagnose and fix the immediate problem, will you have built the predictive and monitoring capability to prevent the next stall?

If you’re honest about those questions, you probably realise your current analytical capability isn’t where it needs to be. 

The good news: building that capability (decomposing problems into observable components, mapping cause-and-effect relationships, distinguishing actionable from vanity metrics) is itself a skill that compounds over time. 

Each cycle of hypothesis, test, and learning improves your understanding of how your organisation actually works. Each validated causal relationship serves multiple purposes: diagnosis when things break, prediction before they break, progress tracking on whether fixes work, and health monitoring to maintain capability. 

That compounding understanding is what accelerates growth.

Footnotes

[1] Douglas Hubbard’s Applied Information Economics (AIE) methodology has been used across hundreds of organisations in IT portfolios, R&D, military logistics, and strategic decision-making. His work demonstrates that perceived “immeasurables” can be quantified through decomposition and strategic observation.

[2] Eliyahu Goldratt’s Theory of Constraints was initially developed for manufacturing but has been successfully applied across industries. The Thinking Processes (particularly the Current Reality Tree) provide systematic tools for tracing symptoms to root causes through logical cause-and-effect relationships.

[3] “Lean Analytics” by Croll and Yoskovitz builds on Eric Ries’s Lean Startup methodology, specifically addressing the “measure” component of the Build-Measure-Learn cycle. Their framework for distinguishing actionable from vanity metrics has become standard in startup and growth company analytics.

[4] Hubbard’s “Rule of Five” demonstrates that a random sample of five observations has a 93.75% chance of containing the median of the population, meaning small samples can dramatically reduce uncertainty when properly applied.

[5] Goldratt’s research showed that 70% or more of organisational problems typically trace back to a small number of root causes, often policies or measurements that made sense previously but no longer serve current objectives.

[6] Croll and Yoskovitz’s concept of the “One Metric That Matters” (OMTM) helps organisations focus analytical attention on the constraint most limiting current growth, preventing the dispersion of effort across dozens of metrics simultaneously.

[7] The Three-Layer Metrics Framework (Outcome/Performance/Input metrics) synthesises Hubbard’s decomposition approach, Goldratt’s cause-and-effect logic, and Croll and Yoskovitz’s leading-lagging indicator relationships into a practical implementation model that serves root cause diagnosis, predictive analytics, strategic progress tracking, and team health monitoring.