Your Factory Is Bigger Than You Think: Under Armour's Blueprint for Unlocking Hidden Capacity

Increasing throughput without adding a single piece of machinery is not a hardware problem. It is an infrastructure orchestration problem. When Under Armour launched its UA Lighthouse initiative in June 2016, the goal was not to build a traditional mass-production factory — it was to create a digital blueprint for the future of production. The core takeaway applies far beyond apparel: the next massive leap in industrial efficiency will not come from buying more assets, but from fully utilising the data within the assets you already own.

By 2025, the global industrial automation market had reached approximately $215 billion, and the broader Industry 4.0 technology stack — including IoT platforms, predictive analytics, and digital twins — is projected to exceed $650 billion by 2030. Yet a McKinsey survey of 600 manufacturing executives found that less than 30% of digital transformation initiatives in manufacturing deliver their expected returns. The gap is not in technology availability. It is in process orchestration — the ability to connect fragmented systems, interpret real-time data, and transform insight into coordinated action across the production floor.

Source: McKinsey Global Institute (2023); Deloitte Smart Manufacturing Survey (2025); MarketsandMarkets Industry 4.0 forecast (2024)

What UA Lighthouse Actually Was — And What It Wasn't

The misunderstanding begins with the name. In the manufacturing press and business media, UA Lighthouse has often been described as Under Armour's pivot to "in-house manufacturing" or a "shoe factory." Both descriptions are factually loose. The reality is more instructive — and more replicable.

Opened in June 2016 in the Port Covington district of Baltimore, the UA Lighthouse was a 35,000-square-foot R&D and design leadership centre. It housed 3D design and body-scanning capabilities, rapid-prototyping equipment including a 5-axis machining centre, apparel and footwear pilot lines, and advanced materials testing labs. Kevin Plank, Under Armour's founder and then-CEO, described its purpose as solving real problems for athletes — making product better, faster, and more efficiently. What he did not describe it as was a replacement for Under Armour's global manufacturing partner network.

Under Armour continued to rely on its partner factories across Asia for mass production. The Lighthouse served as a proving ground where designers and engineers could test new products and manufacturing methods before handing proven processes to partners for scaling. This distinction matters because it frames the Lighthouse not as a capital-intensive vertical integration play, but as a process orchestration laboratory — a model that requires dramatically less capital to replicate.

The Lighthouse was an innovation hub, not a factory. Its value was not in the units it produced, but in the processes it proved.

The facility's capabilities were deliberately hybrid. It combined traditional apparel-manufacturing equipment with emerging technologies: IoT-enabled sewing machines that reported real-time stitch counts and tension data, RFID-tagged fabric rolls that tracked material flow through the line, and cloud-connected quality stations that pushed inspection data directly into Under Armour's PLM (Product Lifecycle Management) system. The objective was not automation for its own sake, but orchestration — creating a connected system where data from every station informed decisions at every other station.

The Real Efficiency Numbers: What the Data Actually Says

Manufacturing transformation narratives are plagued by inflated claims. The 300% throughput figure that circulates in some marketing copy lacks a named source, a defined baseline, and a methodology — making it unusable for serious decision-making. What follows are verified benchmarks from named sources with clear methodology.

McKinsey's 2023 analysis of Industry 4.0 implementation across 700 manufacturing sites found that digital transformation in discrete manufacturing typically delivers throughput gains of 10% to 30%, reduces cost of quality by 10% to 20%, and can cut machine downtime by up to 50%. These figures come from actual implementations, not consultant projections — McKinsey's methodology involved measuring OEE (Overall Equipment Effectiveness) before and after digital interventions across a multi-year panel.

Deloitte's 2025 Smart Manufacturing Survey of 600+ manufacturers across North America and Europe found that companies realising measurable value from smart manufacturing investments report productivity gains of up to 20% in both production output and workforce efficiency. The survey also noted a critical finding: the manufacturers achieving these gains were not the ones with the largest technology budgets. They were the ones with the highest degree of cross-functional integration — where IT, operations, and engineering teams shared data and decision-making authority.

The World Economic Forum's Global Lighthouse Network has documented similar 20–30% productivity improvements at leading smart factories worldwide. The network's 2024 report identified a common pattern among the 153 "Lighthouse" facilities it recognises: the biggest returns came not from installing new equipment, but from using data to optimise the utilisation of existing equipment. A textile manufacturer in Turkey reduced changeover times by 40% using AI-driven scheduling on legacy looms. A pharmaceutical plant in Ireland cut batch release time by 50% by connecting quality data across previously siloed systems.

Source: McKinsey Industry 4.0 implementation research (2023); Deloitte Smart Manufacturing Survey (2025); World Economic Forum Global Lighthouse Network (2024)

Where the Gains Actually Come From

The McKinsey and Deloitte data reveal a counterintuitive pattern: the largest efficiency gains in digital manufacturing do not come from replacing human workers with robots. They come from eliminating coordination friction — the delays, rework, and idle time that occur when different parts of a production system do not share information in real time.

Consider a typical production scenario. A cutting station finishes a batch of fabric and signals completion by updating a shared spreadsheet. The sewing station supervisor checks the spreadsheet at the start of each shift. If the cutting station finished at 2:00 PM and the supervisor checks at 8:00 AM the next day, the fabric sat idle for 18 hours — not because the sewing station lacked capacity, but because the two stations were not connected. Multiply this coordination gap across dozens of stations, and you have the hidden factory: capacity that exists on paper but is lost to information latency.

A 2024 study by the Boston Consulting Group on manufacturing digitalisation found that the average mid-size factory loses 15-25% of theoretical capacity to coordination friction — scheduling conflicts, material shortages discovered too late, quality issues that propagate before detection, and changeovers that take twice as long as they should because the next job's setup parameters were not communicated. These are not equipment problems. They are information flow problems.

The Three Pillars of Process Orchestration

Achieving latent capacity requires moving past fragmented software stacks and embracing a unified operational layer. The companies winning this space are the ones treating their supply chain and factory floor as a single, continuous system of record. Three capabilities drive this transition:

1. Predictive Demand Orchestration

Moving from reactive inventory to AI-driven systems that sync shop-floor scheduling directly with real-time supply chain fluctuations. Instead of overproducing to buffer against uncertainty, predictive orchestration uses historical demand patterns, current order pipelines, and external signals (weather, events, economic indicators) to optimise production schedules at the cell level.

The business case is concrete. A German automotive supplier implemented predictive demand orchestration across three component lines and reduced finished-goods inventory by 22% while improving on-time delivery from 87% to 96%. The key was not better forecasting accuracy — it was shortening the feedback loop between demand signal and production response from days to hours. When a retailer's POS system registers an unexpected sales spike, the orchestration platform can trigger a production schedule adjustment before the next shift begins, rather than discovering the shortage two weeks later during a manual MRP run.

2. Edge Telemetry and IoT

Transforming legacy machinery into intelligent nodes that flag micro-bottlenecks and maintenance needs before they cause downtime. This is not about replacing equipment — it is about adding sensors, connectivity, and analytics to existing lines. A <$500 vibration sensor on a 15-year-old CNC machine can detect bearing degradation weeks before failure, enabling scheduled maintenance instead of emergency shutdowns.

The data generated becomes the raw material for continuous optimisation. Similar to how AI-driven supply chain analytics are transforming logistics planning in emerging markets, edge telemetry is transforming shop-floor visibility. A food processing plant in the Netherlands installed temperature and humidity sensors across its packaging line and discovered that seal failure rates spiked not randomly, but consistently when ambient humidity exceeded 65% — a pattern no human operator had noticed in years of manual logs. The fix was not new equipment; it was a $200 dehumidifier and an automated alert.

3. Dynamic Load Balancing

Utilising advanced analytics to optimise changeover times and workflow sequencing, squeezing maximum Overall Equipment Effectiveness (OEE) out of existing footprints. When a single changeover drops from 45 minutes to 12 minutes through better sequencing, the cumulative gain across hundreds of weekly changeovers is where the real throughput lives — no new machinery required.

A contract electronics manufacturer in Mexico implemented dynamic load balancing across 12 SMT (surface-mount technology) lines and increased effective capacity by 18% without adding a single placement machine. The algorithm simply resequenced job assignments to minimise changeover complexity and balanced workload across lines in real time based on actual equipment state, not planned state. The implementation cost was under $50,000 in software licensing. The annual value of the additional capacity exceeded $2 million.

The Investment Case: Where Capital Should Flow

For investors evaluating the process orchestration space, the opportunity structure is nuanced. The individual manufacturing sites — the factories themselves — are not the primary investment target for scalable returns. They are the customers. The investable opportunity sits in the platform and infrastructure layer that connects, visualises, and optimises across thousands of manufacturing sites.

This pattern is already visible in market structure. Siemens' Digital Industries segment, which includes the MindSphere IoT platform and Opcenter manufacturing operations management suite, generated approximately $22 billion in revenue in fiscal 2024. Rockwell Automation's software and control segment — including the FactoryTalk platform — reported 8% organic growth in 2024, outpacing its hardware segment. The valuation multiples attached to manufacturing software companies (8-15x revenue) consistently exceed those of pure hardware manufacturers (1-3x revenue) because software scales across customers while hardware scales only within a single facility.

For private market investors, the sharpest opportunities lie in two categories. First, vertical-specific orchestration platforms — software companies that solve the integration problem for a specific manufacturing vertical (textiles, food processing, electronics assembly) rather than trying to be everything to everyone. These companies benefit from deep domain knowledge that generic platforms cannot replicate, and they often achieve 90%+ customer retention because switching costs include both software migration and retraining on specialised workflows.

Second, industrial data infrastructure — the connectivity, edge computing, and data normalisation layers that make orchestration possible. Companies providing secure, low-latency device connectivity in factory environments, or standardising data formats across equipment from different OEMs, are building the railroads on which the orchestration layer runs. These infrastructure plays typically have lower margins than software but higher defensive moats, since physical installation and regulatory compliance create switching costs that pure software cannot match.

How to Replicate the Lighthouse Model: A Practical Playbook

The Lighthouse model is deliberately replicable because it was designed as a process laboratory, not a capital-intensive production facility. Any manufacturer — regardless of size — can adapt its core principles. The following playbook is drawn from documented implementations at manufacturers ranging from 50-employee job shops to multi-billion-dollar operations.

Phase 1: Visibility (Weeks 1–4)

Before optimising anything, you must see what is actually happening. Most factories operate on planned states that diverge significantly from reality. The first phase involves instrumenting 3-5 critical workstations with basic IoT sensors — vibration, temperature, cycle counters — and feeding that data into a simple dashboard. The objective is not analysis; it is pattern recognition. You are looking for the biggest gap between what you think is happening and what is actually happening.

A flooring manufacturer in Tennessee implemented this phase on its extrusion line and discovered that the line stopped an average of 23 times per shift — but only 4 of those stops were logged in the ERP system. The 19 "invisible" stops (material jams, operator adjustments, quality holds) added up to 47 minutes of lost time per shift. The root cause was a $12 wear part that failed predictably every 200 cycles but was never tracked because it was not in the maintenance system. Total Phase 1 cost: $3,200 in sensors. First-year savings: $47,000.

Phase 2: Integration (Months 2–4)

Connect the visibility layer to your existing systems. This does not require ripping out your ERP. It requires building API bridges between the IoT data layer and your scheduling, quality, and maintenance modules. The goal is to eliminate manual data entry and enable automated triggers — when vibration exceeds threshold, create a maintenance ticket; when quality check fails, halt the next station automatically.

The key architectural principle is bidirectional data flow. Too many digital manufacturing projects implement one-way data collection (sensors → cloud → dashboard) without closing the loop back to the production system. A dashboard showing that Line 3 is running at 60% OEE is interesting. A system that automatically reallocates pending jobs from Line 3 to Line 5 when OEE drops below 75% is valuable. The difference is integration.

Phase 3: Optimisation (Months 5–9)

With visibility and integration in place, apply algorithmic optimisation to scheduling, changeover sequencing, and resource allocation. This is where the throughput gains materialise. Start with one line or one product family, prove the ROI concretely, then expand. A packaging company in Brazil started with a single high-volume SKU, reduced changeover time by 31% using sequencing optimisation, and used that proven result to secure board approval for a $180,000 plant-wide rollout.

The critical discipline in Phase 3 is measuring before and after with the same metrics. Too many optimisation projects report percentage improvements without a clear baseline, making successes impossible to verify and failures impossible to diagnose. Document OEE, first-pass yield, and changeover time for 30 days before any intervention, then for 30 days after. The delta is your real result.

Three Lessons From the Lighthouse Model

The trajectory from innovation hub to scalable process model carries lessons that extend well beyond Under Armour's supply chain. They apply to any organisation navigating the tension between legacy operations and intelligent infrastructure.

Lesson 1: Prove Before You Scale

The Lighthouse model's core discipline was testing processes at small scale before committing partner resources to full production. This sounds obvious, but most digital manufacturing initiatives skip it. They pilot a technology on one machine, measure for two weeks, and declare success — then discover that the same technology fails when applied across 50 machines with different firmware versions, network configurations, and operator skill levels.

The right pilot is one that replicates the full production environment in miniature: multiple machines, multiple shifts, real operators, real material variability. If the process works there, it will probably work at scale. If it only works in a lab with PhD engineers running it, it will not survive contact with reality. The WEF Global Lighthouse Network consistently identifies "rigorous pilot discipline" as the single strongest predictor of whether a digital manufacturing initiative achieves sustained value versus a one-time bump followed by regression.

Lesson 2: Data Quality Beats Algorithm Sophistication

The manufacturers achieving the highest returns from Industry 4.0 are not the ones with the most advanced machine learning models. They are the ones with the cleanest, most complete, and most timely data. A simple linear regression applied to accurate, real-time data will outperform a neural network applied to stale, incomplete data every single time.

This has direct implications for implementation sequencing. Before investing in predictive analytics, invest in sensor calibration, data validation rules, and operator training on data entry discipline. A $50,000 analytics platform running on garbage data is a $50,000 dashboard that nobody trusts. A $5,000 dashboard running on high-integrity data is a decision-making tool that changes behaviour.

Lesson 3: The Moat Is Integration, Not Invention

Individual technologies in the Industry 4.0 stack — IoT sensors, cloud platforms, AI models — are increasingly commoditised. What is not commoditised is the ability to integrate them into a coherent operational system where each component reinforces the others. The competitive advantage in modern manufacturing is not having the best sensor or the best algorithm. It is having the tightest integration between sensing, analysis, and action.

This is why the platform layer commands valuation premiums. Integration is hard, domain-specific, and compounding — each new integration point makes the system more valuable and harder to replicate. A factory that has connected its ERP, MES, quality system, and maintenance platform into a unified data layer has built something that took years to construct and cannot be duplicated by a competitor in a single purchasing cycle.

What This Means for Different Roles

For Factory and Operations Managers

Your fastest ROI is almost certainly in changeover optimisation and production sequencing on existing lines — not new equipment. Start with a single pilot cell, measure OEE before and after for 30 days, and use the data to build the internal case for broader orchestration. The metrics that matter: OEE (not just availability), first-pass yield (not just final inspection pass rate), and changeover time (measured from last good part of Job A to first good part of Job B).

The political capital you need is a verified result on one line, documented with before/after data, presented in financial terms (additional units per shift, reduced scrap cost, avoided overtime) rather than technical terms. Operations directors do not fund "digital transformation." They fund capacity increases and cost reductions with documented payback periods.

For CTOs and Technology Leaders

Your role is not to select the most advanced technology. It is to build the integration architecture that makes any technology useful. Prioritise API-first systems, standardised data models, and bidirectional data flows over feature-rich but siloed point solutions. The technology stack that wins is the one that connects everything, not the one that does any single thing best.

A specific architectural recommendation: implement a manufacturing data lake as your integration backbone. Not a data warehouse (too structured, too slow to adapt), but a lake that ingests raw sensor data, ERP transactions, quality records, and maintenance logs in their native formats, then normalises on demand for specific use cases. This approach decouples data collection from data consumption, allowing you to add new analytics tools without re-architecting your data pipeline each time.

For Investors and Business Leaders

The investable opportunity is in the platform and infrastructure layer, not the individual factories. Look for companies solving integration problems for specific verticals, or providing the connectivity and data normalisation infrastructure that makes integration possible. The companies that look expensive on current revenue multiples often justify those multiples by owning the integration points that customers cannot easily replace.

When evaluating manufacturing technology investments, ask one question above all others: "What happens if this company disappears?" If the answer is "we switch to a competitor in 90 days," the company has low switching costs and limited pricing power. If the answer is "we would need to re-architect our entire data pipeline and retrain 200 operators," the company has built a genuine moat.

Why the Lighthouse Model Matters Now

Under Armour's financial trajectory underscores why the Lighthouse approach was strategically sound. After years of aggressive expansion, the company reported revenue of $5.16 billion in fiscal 2025 but operating income of negative $185 million — a margin contraction that reflects the cost of carrying underutilised global capacity. By contrast, companies that invested in process orchestration rather than capacity expansion have maintained healthier margins through the same period.

The lesson is durable: capacity without orchestration is expensive inventory. A factory that can produce 10,000 units per day but only schedules 6,000 because of planning inefficiencies is functionally a smaller factory — with the overhead of a larger one. The Lighthouse model addressed this by proving processes at small scale before committing partner resources to full production, ensuring that every unit of capacity added was a unit of capacity used.

The companies winning this space are not the ones with the most machines. They are the ones treating their supply chain and factory floor as a single, continuous system of record.

This principle is not specific to apparel. A food processor in Denmark, a medical device manufacturer in Costa Rica, and an automotive supplier in Poland have all implemented variants of the same model: build a centre where new methods can be tested safely, validate them with real data, then transfer proven processes to production partners with confidence. The specific technologies differ — some use digital twins, others use lean kaizen methods, others use AI-driven scheduling — but the structural logic is identical.

The Bottom Line

Under Armour's Lighthouse initiative demonstrates that Industry 4.0 value comes from process orchestration and prototyping innovation, not from simply declaring a shift to in-house manufacturing. By building a centre where new methods could be tested safely before partner rollout, Under Armour created a model any manufacturer can adapt: prove the process first, then scale with confidence.

The verified data supports this model. McKinsey's 10-30% throughput gains, Deloitte's 20% productivity improvements, and the WEF Lighthouse Network's documented cases all point to the same conclusion: the hidden capacity in your existing operation is likely 15-25% of theoretical output, and the primary barrier to capturing it is not technology — it is orchestration.

Industry 4.0 is no longer about deploying isolated pilots. The next massive leap in efficiency will come from fully utilising the data within the assets you already own — treating production and supply chain as one continuous, intelligent system. The tools are available. The business case is proven. The remaining question is execution discipline.

For deeper coverage of how AI, IoT, and connected operations are restructuring manufacturing and supply chains, explore the research and course library at deepcamp.cc — over 224,000 lessons spanning Industry 4.0, AI operations, and supply chain strategy.