Most business models suffer from diminishing returns. The first million dollars in revenue is cheap to acquire. The second costs more. The tenth costs dramatically more — because easy customers are gone, competitors have responded, and the market segment is saturated. This is the default condition of economic activity, and it's been the organizing assumption of economics since David Ricardo. But a specific class of businesses operates under the opposite dynamic: the more they do something, the better they get at doing it, and the better they get, the more the system rewards them for doing it. These are improving returns — and they are the most powerful force in modern business.
The mathematical expression is Wright's Law. In 1936, Theodore Wright — an aeronautical engineer at Curtiss-Wright Corporation — published a paper documenting a pattern in aircraft manufacturing: every time cumulative production of a given aircraft model doubled, the per-unit labor cost dropped by a consistent percentage, typically 15–20%. The pattern was not a one-time efficiency gain. It was a curve — a predictable, continuous decline in cost as a function of cumulative experience. Wright observed it in airplane manufacturing, but the law turned out to be universal. Semiconductors follow it. Solar panels follow it. Batteries follow it. DNA sequencing follows it. Any production process that involves complex coordination, iterative learning, and cumulative refinement exhibits Wright's Law — cost declines as a power function of cumulative production volume.
The distinction from economies of scale matters enormously. Economies of scale are about size at a point in time — the cost advantage of producing a million units this year versus a thousand. Improving returns are about accumulated experience over time — the cost advantage of having produced ten million units across your entire history versus a competitor who has produced one million. Scale can be purchased with capital. A well-funded competitor can build a factory as large as yours. Improving returns cannot be purchased — they must be earned through cumulative production, cumulative learning, and cumulative refinement of processes that embed knowledge into the system itself. The competitor who builds an identical factory on day one faces a cost curve that the incumbent already descended years ago.
Tesla's battery cost trajectory is the cleanest modern demonstration. In 2010, lithium-ion battery packs cost approximately $1,100 per kilowatt-hour. By 2023, the cost had fallen below $140/kWh — an 87% decline driven primarily by Wright's Law dynamics. Each doubling of cumulative battery production reduced costs by roughly 28%. The learning was embedded in everything: cathode chemistry optimization, cell manufacturing yield improvements, pack design simplification, supply chain negotiation leverage, and factory layout efficiency gained through repetition. A new entrant building batteries in 2024 does not start at $140/kWh. They start somewhere above it — because they lack the cumulative learning that Tesla, CATL, and BYD embedded into their processes through millions of units of production experience. The cost is not in the factory. It's in the history.
TSMC demonstrates improving returns in semiconductor fabrication. Each new process node — from 28nm to 16nm to 7nm to 5nm to 3nm — requires TSMC to solve manufacturing challenges that no competitor has solved at equivalent scale. The learning from each node transfers to the next: lithography techniques, defect reduction methods, yield optimization algorithms, and equipment calibration procedures all improve with cumulative production. Intel, with decades of fabrication experience but lower cumulative production at advanced nodes, has struggled to match TSMC's yields at 7nm and below — not because Intel lacks engineering talent but because TSMC has run more wafers through more advanced processes and accumulated more learning. The improving returns advantage compounds: TSMC's superior yields attract more customers, more customers fund more production, more production generates more learning, and more learning widens the yield gap.
Amazon's logistics network is improving returns applied to distribution. Every package shipped teaches Amazon's routing algorithms something about delivery optimization. Every warehouse processed teaches the robotics and layout teams something about throughput maximization. Every holiday peak teaches the capacity planning systems something about demand forecasting. By 2024, Amazon had shipped over fifteen billion packages cumulatively. That cumulative experience — embedded in algorithms, warehouse designs, driver routing software, and packaging optimization — represents a learning advantage that no competitor can shortcut by building warehouses. The warehouses are commodities. The accumulated intelligence within them is not.
Data flywheels represent the purest form of improving returns in the digital economy. Google Search processes over 8.5 billion queries per day. Each query — and each subsequent click, scroll, and refinement — teaches the algorithm something about relevance. The cumulative dataset of trillions of search interactions produces a search quality advantage that no competitor can replicate by building better algorithms alone, because the algorithms are only as good as the data that trains them, and the data is a function of cumulative usage. Bing can match Google's engineering talent. It cannot match Google's twenty-five years of cumulative query data. The improving return is not in the code. It is in the corpus of human behavior that the code has learned from — a corpus that grows with every query and can never be purchased, only accumulated.
Section 2
How to See It
Improving returns reveal themselves through a specific pattern: the company or system gets better at its core activity with each unit of output, and the improvement is durable — it doesn't reset when conditions change. The signal is not that costs decline (any company can cut costs through negotiation or austerity). The signal is that costs decline because the system is learning, and the learning is embedded in processes, algorithms, or organizational knowledge that accumulates rather than depreciates.
You're seeing Improving Returns when a company's cost per unit or quality per unit improves predictably as a function of cumulative output — not one-time gains, but a continuous curve that steepens the lead over competitors with each passing year.
Technology
You're seeing Improving Returns when NVIDIA's CUDA ecosystem creates a self-reinforcing learning cycle. Each generation of GPU hardware generates more developer experience with parallel processing. More developer experience produces more optimized software libraries. Better libraries attract more researchers and engineers to the platform. More platform users generate more feedback that informs the next hardware generation. The improving return is not just in the silicon — it's in the ecosystem surrounding it. Each iteration of the cycle produces a platform that is harder to replicate because the cumulative learning is distributed across millions of developers, not stored in a single factory.
Manufacturing
You're seeing Improving Returns when Toyota's production system improves year after year despite being studied, documented, and explicitly copied by competitors for five decades. The Toyota Production System generates improving returns because each kaizen cycle — each small, incremental improvement — embeds new knowledge into the system. Competitors can copy the visible artifacts (kanban boards, just-in-time delivery, andon cords) but cannot copy the cumulative learning from seventy years of continuous improvement applied to billions of units produced. The system's advantage is not its design. It's its history.
Investing
You're seeing Improving Returns when a company's margins improve with each year of operation despite constant pricing — signaling that the learning curve is reducing costs faster than competition is reducing prices. The classic example: semiconductor foundries whose gross margins improve from 40% to 55% over a decade as cumulative production drives yield improvements and process efficiencies. The margin expansion is not pricing power. It's learning power — the mathematical consequence of Wright's Law applied to billions of chips.
Data & AI
You're seeing Improving Returns when a machine learning system improves with each user interaction and the improvement attracts more users. Google Search gets better with every query — each click, each query refinement, each bounce teaches the algorithm something about relevance. Better results attract more users. More users generate more training data. More training data produces better results. The improving return is the data flywheel: the system's performance is a function of cumulative usage, and cumulative usage is a function of performance. Competitors cannot buy their way to Google's query history. They must earn it — and earning it requires already having the traffic that quality generates.
Section 3
How to Use It
The strategic question is not whether your business has improving returns — most don't. The question is whether you can design your business to generate them, because improving returns are the most durable competitive advantage available.
Decision filter
"Ask: does each unit of output make the next unit cheaper, better, or faster — and is that improvement embedded in the system permanently? If yes, you're on an improving returns curve. If no, you're in a commodity business where any advantage can be purchased by a well-funded competitor."
As a founder
Design your business to generate cumulative learning from its core activity. This means choosing architectures where every unit of output feeds back into the system: every customer interaction improves the product, every manufacturing run improves the process, every data point improves the algorithm. The founder of a SaaS company who treats customer data as exhaust is leaving improving returns on the table. The founder who builds feedback loops — where usage data improves the product, which drives more usage, which generates more data — is building an improving returns machine.
The second application: invest in the early, expensive part of the learning curve when competitors won't. Wright's Law guarantees that costs will decline with cumulative production, but the first units on the curve are the most expensive. Tesla invested billions in battery production when the cost per kWh made electric vehicles unprofitable. The investment was not in batteries — it was in learning. Each unit produced moved Tesla further down the cost curve, and by the time competitors entered the EV market, Tesla had a cumulative production advantage that translated into a cost advantage no amount of capital could instantly close.
As an investor
Improving returns produce the most asymmetric investment outcomes because the advantage compounds rather than decays. The company operating on a Wright's Law curve gets cheaper with each unit produced, which allows lower prices, which drives more volume, which further reduces costs. The flywheel accelerates. The key diligence question: is this company's cost advantage a function of cumulative production (improving returns, durable) or current scale (economies of scale, replicable)?
Track the slope of the learning curve. Wright's Law predicts a specific cost reduction for each doubling of cumulative output — typically 10–30% depending on the industry. A steeper curve means faster improving returns. Solar panels have exhibited a 28% cost reduction per doubling. Batteries show 28%. Semiconductors historically show 20–25%. The slope tells you how fast the leader's advantage will widen — and how long a challenger will take to reach cost parity, if they ever can.
As a decision-maker
Map your organization's improving returns surfaces — the activities where cumulative experience produces durable advantage — and protect them with investment priority. The most common strategic error is cutting investment in improving returns activities during downturns, which cedes cumulative learning to competitors who maintain investment. Toyota never cut its continuous improvement programs during recessions. The short-term cost savings of suspending improvement activities are real. The long-term cost — stalling on the learning curve while competitors advance — compounds for years after the downturn ends.
The second application: identify where your organization is accidentally generating improving returns and amplify the effect. Many companies produce learning as a byproduct of operations without capturing it systematically. The warehouse team that discovers a more efficient picking pattern through trial and error is generating improving returns — but only if the learning is codified into the routing algorithm and applied across all facilities. Individual learning that stays in one person's head is not an improving return. Institutional learning that embeds in the system is.
Common misapplication: Confusing economies of scale with improving returns. Scale is about spreading fixed costs over more units at a point in time. Improving returns are about the cumulative learning that reduces variable costs over the full production history. A competitor can match your scale by building a factory the same size. A competitor cannot match your improving returns without replicating your cumulative production history — which, by definition, takes time that cannot be compressed with capital.
Second misapplication: Assuming improving returns are permanent. Learning curves can plateau when the process reaches physical or theoretical limits. Moore's Law — the semiconductor industry's most famous improving returns curve — slowed dramatically after 2015 as transistor sizes approached atomic-scale limits. The improving returns machine that drove fifty years of exponential cost reduction began to stall not because learning stopped but because the physics imposed a boundary condition the learning curve couldn't transcend.
Section 4
The Mechanism
Section 5
Founders & Leaders in Action
The leaders below didn't stumble onto improving returns. Each deliberately designed their organizations to generate and capture cumulative learning — and then invested aggressively in the early, expensive portion of the learning curve when the payoff was years away and the skeptics were loudest.
What unites them is patience calibrated to the curve. Both understood that improving returns require upfront investment in volume and process before the cost advantages materialize — and both had the conviction to sustain that investment through years of negative cash flow while the learning accumulated.
Huang built NVIDIA into the most valuable semiconductor company in the world by riding an improving returns curve that he recognized two decades before the market did. NVIDIA's GPU architecture — designed for massively parallel computation — improved with each generation not just through better transistors but through cumulative software optimization. The CUDA platform, launched in 2006, created a feedback loop: developers who wrote code for NVIDIA's architecture generated libraries, tools, and educational materials that made the next generation of developers more productive. Each generation of hardware accumulated more software investment. Each generation of software accumulated more hardware optimization. By 2024, NVIDIA had over four million CUDA developers — a cumulative ecosystem asset that no competitor could replicate by building better chips alone. AMD and Intel could match NVIDIA's silicon performance on a spec sheet. They could not match the fifteen years of cumulative developer learning embedded in the CUDA ecosystem. Huang invested in CUDA for nearly a decade before it generated meaningful revenue. That patience — sustaining investment in the early, expensive part of the improving returns curve — is what separated NVIDIA from every GPU competitor that focused on hardware specifications without building the cumulative ecosystem that transforms hardware into a platform.
Musk bet Tesla's entire strategy on Wright's Law before most of the automotive industry had heard of it. The bet was explicit: battery costs would decline on a predictable curve as cumulative production increased, and the company that descended the curve fastest would achieve cost positions that latecomers could not match. Tesla's Gigafactory — a $5 billion investment announced in 2014 when EV skepticism was mainstream — was a Wright's Law machine. The factory was designed not for current demand but for the cumulative production volume required to drive battery costs below the threshold where electric vehicles achieve price parity with internal combustion engines. Musk published a version of the learning curve in investor presentations, showing the specific relationship between cumulative production and cost per kWh. By 2023, Tesla had produced over five million electric vehicles — cumulative experience that translated into battery pack costs roughly 30% lower than competitors entering the market at lower volumes. The Cybertruck, the Semi, and the anticipated affordable Model Q all depend on the same improving returns thesis: each vehicle produced drives further cost reduction that funds the next vehicle category. The strategy only works because the improving returns are real. Every competitor who enters the EV market starts higher on the cost curve than Tesla occupies — and closing the gap requires not capital but cumulative production that takes years to accumulate.
Section 6
Visual Explanation
The curve tells the story. The leader — the company that started earlier and has produced more cumulative units — sits lower on the cost curve at any given point in time. The entrant, even building an identical factory with identical technology, starts higher on the curve because they lack the cumulative production experience that drove the leader's cost position down. The vertical gap between the two curves at any given moment is the improving returns moat — a cost advantage that exists not because of current scale but because of historical learning. The three boxes at the bottom capture the mechanism: Wright's Law provides the mathematical relationship, the feedback loop explains why the advantage accelerates, and the moat explains why competitors cannot close the gap with capital alone.
The critical insight is the curve's trajectory over time. The gap between leader and entrant does not narrow as the entrant accumulates production. It widens — because the leader continues to produce and learn while the entrant is catching up. The only way to close the gap is to produce at a rate that exceeds the leader's, which requires either massive capital investment (Tesla's approach against incumbents) or a technological discontinuity that resets the curve entirely (digital cameras resetting Kodak's chemical film learning). Absent either condition, the leader's improving returns advantage compounds indefinitely.
Section 7
Connected Models
Improving returns sits at the intersection of compounding dynamics, strategic positioning, and organizational learning. The connected models below explain the mechanisms that generate improving returns, the frameworks that describe their effects, and the boundary conditions where they weaken or reverse.
The reinforcing connections show how improving returns compound through flywheels, network effects, and experience curves — each mechanism feeding the others. The tension connection reveals the critical distinction between improving returns and the scale effects they are most often confused with. The leads-to connection traces how firm-level improving returns produce market-level increasing returns dynamics.
Reinforces
[Compounding](/mental-models/compounding)
Improving returns are compounding applied to organizational capability rather than capital. Just as compound interest turns a small initial investment into a large sum through reinvestment of returns, improving returns turn early cumulative production into a widening cost advantage through reinvestment of learning. The mathematics are structurally identical: both involve exponential growth from repeated application of a percentage gain to an accumulating base. The difference is that financial compounding can be replicated by any investor with capital, while improving returns compounding requires cumulative experience that takes years to accumulate and cannot be shortcut.
Reinforces
[Flywheel](/mental-models/flywheel)
The flywheel is the operational mechanism through which improving returns manifest. Amazon's flywheel — lower prices drive more customers, more customers drive more volume, more volume drives lower costs, lower costs enable lower prices — is an improving returns machine expressed as a circular process. Each rotation of the flywheel embeds additional learning into the system: the logistics network gets smarter, the recommendation algorithms get better, the supply chain relationships get tighter. The flywheel accelerates because each rotation generates both financial returns and learning returns — and the learning returns feed back into the next rotation.
Tension
Economies of Scale
Economies of scale and improving returns are frequently confused but structurally different. Scale is a static concept — the cost advantage of being large at a single point in time. Improving returns are dynamic — the cost advantage of having accumulated more experience over time. A competitor can match your scale by building an equivalent facility. A competitor cannot match your improving returns without accumulating equivalent experience, which requires time that capital cannot compress. The tension emerges in strategic planning: companies that pursue scale without improving returns build large commodity businesses. Companies that pursue improving returns build durable advantages that scale effects alone cannot replicate.
Section 8
One Key Quote
"Increasing returns are the tendency for that which is ahead to get further ahead, for that which loses advantage to lose further advantage. They are the mechanism by which the economy selects its structure."
— W. Brian Arthur, 'Increasing Returns and the New World of Business,' HBR (1996)
Arthur spent nearly a decade fighting to get this idea taken seriously by mainstream economics. The prevailing orthodoxy — rooted in Alfred Marshall's equilibrium models — assumed that all returns diminish: any advantage a firm accumulates will erode as competitors imitate and the market corrects. Arthur argued that this was true for the old economy of resource extraction and physical manufacturing but false for the emerging economy of knowledge, technology, and networked systems. In knowledge-based industries, the leader's advantage doesn't erode. It compounds. The firm that is ahead gets further ahead because learning, data, and network effects accumulate with experience and use.
The quote's sharpest edge is the second half: "for that which loses advantage to lose further advantage." Improving returns are not neutral. They are polarizing. The leader and the follower are on the same curve, but the leader is further down it — and each unit of additional production widens the gap. The follower is not standing still. They are falling further behind with each unit the leader produces, because the leader's marginal learning generates cost reductions that the follower cannot match without equivalent cumulative volume. This is why improving returns industries tend toward concentration: the curve punishes the second-mover as severely as it rewards the first.
The strategic implication Arthur drew — and that most businesses still underappreciate — is that in improving returns industries, early investment in volume is not a cost. It is the purchase of a permanent asset: a position on the learning curve that no subsequent entrant can occupy without years of equivalent production. The companies that understand this invest aggressively in the early, expensive portion of the curve. The companies that don't understand it optimize for short-term profitability and watch the improving returns leader build an advantage that eventually becomes insurmountable.
Section 9
Analyst's Take
Faster Than Normal — Editorial View
Improving returns is the single most important dynamic in technology investing, and the one most frequently conflated with economies of scale. The distinction matters for valuation. Economies of scale are replicable — a well-funded competitor can build an equivalent operation. Improving returns are not replicable — they require cumulative experience that takes years to accumulate. A company with improving returns deserves a structural premium because its advantage widens with each unit produced. A company with only scale economies deserves a temporary premium that erodes as competitors reach equivalent size.
The diagnostic I use: can a new entrant with $10 billion match this company's cost position within three years? If yes, the advantage is scale — and the moat is shallow. If no, the advantage is improving returns — and the moat deepens with each year of additional production. TSMC's fabrication yields cannot be matched with $10 billion and three years. Amazon's logistics intelligence cannot be purchased. Tesla's battery cost curve cannot be replicated by building identical factories. These are improving returns moats, and they compound.
The pattern in AI that most observers are missing: data flywheels are the improving returns of the intelligence era. The AI company that processes more queries accumulates more training data, which produces better models, which attracts more users, which generates more queries. The cycle is Wright's Law applied to algorithmic learning rather than manufacturing. The company that reaches meaningful scale first will descend a learning curve that later entrants cannot shortcut — because the training data is a function of cumulative usage, not current compute capacity. Compute can be purchased. Cumulative usage data cannot.
The most underrated improving returns machine in the current market: SpaceX's Starship program. Each Starship launch generates manufacturing learning (how to build stainless steel rockets cheaper), operational learning (how to catch boosters), and regulatory learning (how to accelerate FAA approvals). SpaceX has launched more orbital rockets than any organization in history — cumulative experience that translates into reliability improvements, cost reductions, and cadence advantages that no competitor can match by building equivalent hardware. The hardware is visible. The cumulative learning is the moat.
The most misunderstood aspect of improving returns: the early portion of the curve looks like waste. Tesla spent years burning cash on battery production that was uneconomical by any traditional analysis. TSMC invested billions in process nodes before they generated positive returns. Amazon ran logistics at a loss for over a decade. In each case, traditional financial analysis said the investment was irrational. Improving returns analysis said the investment was purchasing a position on a learning curve that would compound for decades. The founders who understood this invested through the losses. The analysts who didn't understand it wrote "sell" ratings that aged catastrophically.
Section 10
Test Yourself
The scenarios below test whether you can distinguish genuine improving returns from ordinary scale effects, and whether you can identify when the improving returns curve is the dominant competitive dynamic in a market. The key diagnostic: is the advantage a function of cumulative experience (improving returns, durable) or current size (scale, replicable)?
The most common analytical error is conflating the two. Scale advantages — bulk purchasing power, fixed cost spreading, market presence — are real but replicable. Improving returns advantages — cumulative manufacturing learning, accumulated data assets, embedded process intelligence — are real and structurally unreplicable without equivalent time and experience. The distinction determines whether a competitor can close the gap with money alone or whether the gap widens with every passing quarter regardless of the competitor's capital.
Pay particular attention to the scenario where the answer is counterintuitive — where what appears to be an improving returns advantage is actually a scale effect that a well-funded competitor can match.
Is this improving returns at work?
Scenario 1
A cloud computing provider operates fifty data centres globally. A well-funded competitor announces plans to build fifty equivalent data centres over three years, matching the incumbent's capacity. The competitor's CEO declares: 'We will have equivalent infrastructure within thirty-six months. The incumbents' advantage is purely capital, and we have $40 billion to deploy.'
Scenario 2
Two autonomous vehicle companies are testing self-driving technology. Company A has logged 30 million miles of autonomous driving data. Company B has logged 500,000 miles. Company B argues that their sensor hardware is superior and their machine learning team has more PhD researchers. Company A's CEO says: 'Miles are the moat.'
Scenario 3
A restaurant chain with 2,000 locations has lower food costs per unit than a chain with 200 locations — the larger chain negotiates better supplier pricing due to higher volume. A third chain with 500 locations announces plans to reach 2,000 locations within five years through aggressive expansion, arguing that 'scale parity will eliminate the cost advantage.'
Section 11
Top Resources
The improving returns literature spans manufacturing economics, complexity theory, and technology strategy. Start with Arthur for the economic theory of why improving returns dominate knowledge-based industries, move to Wright for the original empirical observation, and read the BCG experience curve work for the strategic application. The combination provides the theoretical framework (Arthur), the empirical foundation (Wright), and the strategic playbook (BCG) for building improving returns businesses.
The most influential articulation of why knowledge-based industries operate under improving returns rather than the diminishing returns that classical economics assumed. Arthur argues that positive feedback mechanisms — learning, network effects, and adaptive expectations — create winner-take-most dynamics in technology markets. The article changed how a generation of technology strategists and investors evaluated competitive advantage, shifting focus from current market share to cumulative learning position on the improving returns curve.
The foundational paper that documented the learning curve in manufacturing. Wright's observation — that per-unit labor cost declines by a predictable percentage with each doubling of cumulative production — became the empirical basis for what BCG later called the experience curve and what technology strategists now call improving returns. The paper's mathematical formulation remains the standard model for predicting cost declines as a function of cumulative output, and its predictions have proven accurate across industries from aviation to solar panels to batteries.
BCG's experience curve work transformed Wright's manufacturing observation into a strategic framework. The central argument: if cost declines predictably with cumulative production, then the company that achieves the highest cumulative volume first will achieve the lowest cost position — and the strategic imperative is to pursue volume aggressively, even at the expense of short-term profitability, to descend the curve faster than competitors. The framework directly informed Texas Instruments' semiconductor pricing strategy, Toyota's manufacturing investments, and Tesla's battery production scaling.
Arthur's deeper exploration of how technologies evolve through recombination and how improving returns emerge from the interaction between technology components. The book extends his economic theory into a framework for understanding technological progress itself — arguing that technologies improve not through isolated invention but through cumulative recombination of existing components, each improvement creating the conditions for the next. Required reading for understanding why improving returns are structural features of technological systems rather than temporary advantages.
A systematic comparison of Wright's Law (cost declines with cumulative production) and Moore's Law (performance improves with time) across sixty-two technology domains. The researchers found that Wright's Law — the improving returns formulation — predicted cost declines more accurately than Moore's Law across the majority of technologies studied. The finding is significant for strategists: it confirms that cumulative production, not merely the passage of time, is the primary driver of technology cost improvement — and that companies actively pursuing volume have a structural advantage over those waiting for time-based improvements to materialise.
Improving Returns — the more you produce, the better you get. Wright's Law shows cost declining as a function of cumulative experience, creating a durable advantage that capital cannot buy.
Reinforces
Network Effects
Network effects and improving returns often operate simultaneously, creating compounding advantages. Each user who joins a network improves the product for existing users (network effect) and generates data that improves the product's algorithms (improving return). Facebook's social graph created network effects that attracted more users, whose interactions generated data that improved the News Feed algorithm, which increased engagement, which attracted more users. The network effect and the improving return are distinct mechanisms — one operates through connectivity, the other through learning — but they reinforce each other, creating combined advantages that are nearly impossible to replicate.
Leads-to
Increasing Returns (Brian Arthur)
Brian Arthur's theory of increasing returns is the economic formalization of improving returns applied to market dynamics. Arthur demonstrated that in knowledge-based industries, positive feedback mechanisms — learning by doing, network effects, and adaptive expectations — cause the leading firm to accumulate advantages that compound over time, creating winner-take-most market structures. The improving returns curve describes the mechanism at the firm level. Arthur's increasing returns theory describes the market-level consequence: industries with strong improving returns dynamics tend toward concentration because the leader's cumulative advantage grows faster than followers can close it.
Reinforces
The Experience Curve
The experience curve, formalized by the Boston Consulting Group in the 1960s, is the strategic application of Wright's Law. BCG demonstrated that a company's total cost per unit declines by a predictable percentage with each doubling of cumulative production — and argued that firms should pursue volume aggressively to descend the curve faster than competitors. The experience curve is improving returns expressed as a strategic prescription: invest in volume today to achieve cost positions tomorrow that competitors cannot match. The reinforcement is direct — the experience curve is the business strategy derived from improving returns dynamics, and every company that has ridden the curve (Texas Instruments in semiconductors, Toyota in manufacturing, Tesla in batteries) has done so by treating cumulative production as a strategic weapon.
One warning: improving returns create fragility at paradigm shifts. The company that has descended the furthest on a given learning curve has the most to lose when the curve itself becomes obsolete. Kodak's cumulative learning in chemical film was the deepest in the industry — and worthless when digital imaging replaced it. Intel's cumulative learning in x86 fabrication was immense — and insufficient when the market shifted to architectures optimized for mobile and AI. The improving returns leader should always be monitoring whether the underlying curve will remain relevant. The deepest position on a deprecated curve is not an asset. It's a sunk cost dressed as competitive advantage.
