Algorithmic Competition and Market Power Study Notes
Course Architecture and Learning Outcomes
Curriculum Architecture: Digital Transformation Strategy
- Week 1: What Is Digital Transformation (and What It Is Not?) | Case Study: Kroger
- Week 2: Data | Case Study: Amazon.com
- Week 3: Artificial Intelligence | Case Study: AI and Strategy: Lessons from Real-World Cases
- Week 4: Economics of Digital & AI Technologies | Case Study: Uber Competing Globally
- Week 5: Platforms, Network Effects, and Algorithmic Competition | Case Study: Uber Technologies Inc
- Week 6: AI as a General-Purpose Technology: Decision Making, Coordination, and Vision | Case Study: Moderna
- Week 7: AI and the Individual: Work, Skill, Identity | Case Study: Cybersecurity at FireEye
- Week 8: Algorithmic Management and Control | Case Study: Study of Uber Drivers
- Week 9: AI and Organization Design | Case Study: JPMorganChase – Leadership in the age of AI
- Week 10: Algorithmic Competition and Market Power | Case Study: X Fire Paintball & Airsoft
- Week 11: Ethics, Governance, and Regulation of AI | Case Study: Mastercard
- Week 12: Strategy in an AI-Dominated World | Case Study: Microsoft
Week 10 Pedagogical Objectives
- Objective 1: Define algorithmic competition and evaluate how automated systems challenge core assumptions regarding rival observability and competitive stability.
- Objective 2: Detail the economics of attention and analyze the precise structural mechanisms through which ranking algorithms and visibility control market value capture.
- Objective 3: Define competitive bottlenecks and explain how they drive severe profit concentration and generate modern regulatory interventions.
- Objective 4: Evaluate winner-take-most dynamics by diagnosing the market conditions under which platforms tip into monopoly or fragment into competitive oligopolies, focusing on multi-homing costs, product differentiation, and feedback loop strength.
- Objective 5: Apply algorithmic competition frameworks to identify where structural power resides in a digital market ecosystem and assess the strategic forces capable of shifting that power.
Paradigm Shift: Traditional Strategy vs. Algorithmic Competition
Foundations of Classical Strategic Frameworks
- Frameworks such as Porter's Five Forces, game theory, and scenario planning (Porter, 1980) rest on specific industrial-era assumptions:
- Competitors and industry rivals are identifiable.
- Competitive actions and market moves are observable.
- Responses are deliberate and formulated by human decision-makers.
- Competition operates at human speed, characterized by deliberate formulation, strategic reviews, and operational rollouts across weeks, quarters, or years.
- These frameworks reflect physical markets where production, supply chain allocation, physical distribution, and retail pricing decisions require human deliberation.
The Algorithmic Breakdown of Classical Assumptions
- When machine-learning systems continuously set prices, re-rank catalogs, and allocate inventory in real time, classical assumptions collapse:
- Unit of Competition Shifts: The primary competitive unit moves from the firm to the system. Competitors do not face an opposing management team; they face a computational optimization architecture.
- Decoupling of Signals: Strategic analysts can no longer rely on observable signals such as formal product launches, published price schedules, or public corporate announcements.
- Opaque Signal Generation: Critical competitive moves are embedded within proprietary algorithmic outputs—real-time pricing micro-adjustments, dynamic search rankings, and personalized recommendation weightings.
- Systemic Diagnosis: Strategic analysis must transition from observing discrete competitor actions to diagnosing complex, interacting competitive systems (Ezrachi & Stucke, 2016).
Comparative Analysis: Strategic Eras
- Traditional Competition:
- Competitor interface: Firm vs. firm.
- Execution: Move vs. move.
- Operational cycle: Observe Analyze Respond (measured in weeks or months).
- Basis of competitive power: Scale and asset accumulation.
- Algorithmic Competition:
- Competitor interface: System vs. system.
- Execution: Signal vs. signal.
- Operational cycle: Sense React Adapt (measured in milliseconds).
- Basis of competitive power: Control of strategic bottlenecks, ranking mechanisms, and system design, rather than asset scale.
The Mechanics of Algorithmic Competition
Formal Definition
- Algorithmic competition is a competitive dynamic in which firms deploy algorithms that continuously monitor market signals—including rivals' prices, rankings, inventory levels, and consumer engagement outputs—and automatically adjust their own strategic decisions in response, without human deliberation (Ezrachi & Stucke, 2016; Calvano et al., 2020).
- The operational unit of analysis shifts from the deliberate strategizing of human executives to the automated, high-frequency execution of machine learning models.
- This evolution represents a qualitative transformation in market behavior: intentionality, corporate accountability, and information observability are altered when autonomous algorithms interact.
Real-World Mechanism: Digital Advertising Auctions
- In automated ad markets (e.g., Google and Meta), algorithms run continuous, real-time auctions.
- Bid parameters, ad placements, targeting criteria, and clearing prices adjust dynamically based on millions of streaming signals without human oversight.
- System-to-system transactions execute billions of times per day.
- Advertisers seeking audience reach compete directly against automated optimization engines rather than human marketing strategists (Varian, 2009).
The Reactive Loop Architecture
- Algorithmic competition functions via an automated closed loop across market participants:
- Step 1 (Signal Monitoring): Organization A's algorithm continuously monitors live market signals, including rival prices, rank position shifts, inventory thresholds, and granular user behaviors.
- Step 2 (Autonomous Output Adjustment): Organization A's algorithm calculates an optimal response and automatically modifies its output parameters (e.g., updating a retail price, adjusting a keyword ad bid, or altering a search ranking coefficient).
- Step 3 (Signal Ingestion): Organization B's algorithm instantly detects this output shift, treating Organization A's adjustment as a fresh market input signal.
- Step 4 (Automated Counter-Adjustment): Organization B's system executes a corresponding counter-adjustment.
- Step 5 (Continuous Iteration): Organization A detects Organization B's reaction, resetting the cycle. The loop iterates continuously at computational speed.
Strategic Consequences: Speed, Opacity, and Tacit Collusion
- Speed Differential: Human deliberation requires hours, days, or months. Algorithmic loops execute in milliseconds (). This temporal compression eliminates the window required for human strategic intervention. Firms using human pricing desks face severe operational disadvantages against autonomous software, driven by latency rather than analytical acumen.
- Opacity and Strategic Ambiguity: Algorithmic logic and internal weights are proprietary and black-boxed. Rivals observe outputs (e.g., price movements), but cannot view underlying loss functions or decision trees. Consequently, observers cannot easily determine whether an observed price drop represents deliberate competitive aggression, automated inventory clearing, cost optimization, or an unintended model artifact (Ezrachi & Stucke, 2016).
- Tacit Algorithmic Collusion: Reinforcement learning pricing agents can learn that matching competitors' prices is individually rational. Over time, algorithmic feedback loops can converge on sustained supra-competitive prices without explicit communication between executives (Calvano et al., 2020). This dynamic challenges modern antitrust and competition laws, which require evidentiary proof of intentional, coordinated agreements.
Weakening of Core Strategic Assumptions
| Traditional Assumption | Algorithmic Competition Reality |
|---|---|
| Rivals are observable | Algorithmic logic is proprietary; only external outputs are visible. |
| Moves are deliberate | System outputs can be emergent properties rather than intended moves. |
| Responses can be planned | Machine execution speeds eliminate deliberate human planning windows. |
| Competition is firm vs. firm | Dynamic competition occurs as system vs. system and signal vs. signal. |
| Accountability is clear | No individual human formulated the micro-decision; it emerged from algorithmic optimization. |
The Economics of Attention: Ranking and Visibility as Power
The Attention Scarcity Model
- In physical retail, prime physical shelf space is scarce; retail shelf owners capture economic rents from packaged goods manufacturers seeking customer visibility.
- In digital environments, human attention is the scarce economic resource (Wu, 2016).
- Digital ranking algorithms serve as virtual shelf space, dictating the discovery, positioning, and purchase conversion of goods, services, and media.
- Structural Power Chain:
- Control over the ranking algorithm leads directly to control over user attention.
- Control over user attention leads directly to control over economic value capture.
Empirical Quantification of Visibility
- Consumer click-through rates (CTR) decline rapidly across organic search rankings (Ahrefs, 2023):
- Position 1 CTR: Approximately .
- Position 10 CTR: Approximately .
- Marginal Rank Value: Moving up a single position in Google search results increases CTR by an average of approximately .
- This steep decay curve proves that ranking is not an operational or technical detail; it is the primary economic mechanism converting consumer attention into commercial transactions.
Ranking as a Structural Allocation Mechanism
- A ranking algorithm operates as an automated market allocation system that determines:
- Which product or service options are made visible.
- The vertical hierarchy and visual sequence of display.
- Which specific consumer profiles see which listings.
- The temporal context of display down to the millisecond.
- Because these systems allocate attention autonomously, algorithmic power is structural rather than behavioral. Platform operators do not need to intervene manually against individual vendors; the underlying platform architecture directs commercial discovery automatically.
- Example (Amazon Sponsored Product Ads): Because organic search ranking dictates transaction volumes regardless of underlying product quality, third-party sellers are compelled to purchase sponsored placement ads. Amazon converts its algorithmic gatekeeping power into high-margin ad revenue by monetizing access to scarce buyer attention (Zhu & Liu, 2018).
Defaults: The Ultimate Concentration of Visibility
Strategic Definition and Mechanics of Defaults
- A default is an automated, pre-selected operational choice delivered to an end-user unless that user takes conscious, active measures to alter it.
- Defaults represent the most concentrated form of algorithmic ranking: they elevate a single option to Position 1 while suppressing or concealing alternative choices entirely.
Behavioral Drivers of Default Power
- Default settings capture market share due to predictable behavioral factors:
- Consumer Inertia: The physical and cognitive effort of searching for alternatives leads users to accept the initial configuration.
- Cognitive Offloading: Users delegate the complexity of product evaluation to the system.
- Endorsement Bias: Users interpret defaults as an implicit quality endorsement from the platform provider.
- Across varied operational environments, empirical default acceptance rates frequently exceed (Johnson & Goldstein, 2003).
Monetizing Default Control: Google and Apple Case Analysis
- The financial value of default status is demonstrated in US Department of Justice v. Google (2023).
- Google pays Apple tens of billions of dollars annually to secure its position as the pre-installed, default search engine on iOS safari devices.
- This capital outlay is not purchased for conventional advertising exposure; it is the acquisition of an exclusive cognitive default.
- Because defaults capture more than of consumer queries, that default positioning generates search volumes, which feed user data back into Google's advertising engine to produce billions in high-margin ad revenue.
Competitive Bottlenecks and Market Gateways
Theoretical Framing of Competitive Bottlenecks
- A competitive bottleneck occurs when one group of market participants (e.g., commercial merchants, app developers, advertisers) must navigate through a single, centralized platform intermediary to access another vital consumer group (Rochet & Tirole, 2003; Armstrong, 2006).
- Bottlenecks are structural properties of multi-sided market architectures combined with cross-side network effects, rather than manifestations of abusive corporate behavior.
- Bottlenecks are self-reinforcing: as more end-consumers cluster on a single platform gateway, that platform becomes essential for upstream commercial providers, deepening the bottleneck.
Profit Concentration and Value Extraction Mechanics
- When a platform successfully establishes a bottleneck gateway:
- Sellers face high multi-homing costs and cannot reach consumers via alternate routes.
- The platform gains market pricing power independent of its marginal cost structure or product innovation.
- The platform extracts economic rent through access fees, commissions, mandatory advertising fees for basic discoverability, and restrictive contractual terms.
- Long-term profitability concentrates around the control of user access rather than the operational excellence or manufacturing quality of underlying goods.
Structural Bottlenecks Across Dominant Digital Platforms
Operating System App Stores (Apple iOS App Store / Google Play Store):
- Bottleneck Structure: Software developers cannot distribute mobile applications at commercial scale to smartphone users without complying with Apple's or Google's closed distribution rails.
- Monetization & Control: Platforms levy commissions ranging from to on digital transactions and enforce proprietary in-app payment processors.
General Internet Search (Google Search):
- Bottleneck Structure: Controlling approximately of global web search volume, Google serves as the primary gateway for cross-industry customer acquisition.
- Data Compounding Loop: Query scale generates click data, which trains ranking algorithms to enhance search utility, which attracts incremental users and solidifies advertiser dependency on Google Ads.
E-Commerce Marketplaces (Amazon):
- Bottleneck Structure: Generating approximately of all United States e-commerce volume, Amazon is an essential gateway for online retailers.
- Ecosystem Moats: Third-party sellers cannot duplicate Amazon's Prime membership base, proprietary logistics network, and embedded consumer trust, granting Amazon unilateral control over commission fees, search visibility, and operating policies.
Market Distortions and Regulatory Interventions
- Bottlenecks create severe economic distortions:
- Sellers: Experience margin compression due to rising platform take-rates and discoverability costs.
- Consumers: Absorb higher retail prices passed down by squeezed merchants, while discovery is biased toward paid platform sponsors.
- New Entrants: Face distribution blockades, suppressing downstream technological innovation.
- These distortions trigger targeted regulatory responses, such as the European Union's Digital Markets Act (DMA):
- Mandatory API and service interoperability.
- Strict data silos separating native platform operations from third-party merchant intelligence.
- Explicit legal prohibitions against platform self-preferencing and discriminatory ranking.
The Conditional Economics of Winner-Take-Most Markets
Deconstructing the Platform Scale Fallacy
- A common strategic error is assuming that strong network effects and algorithmic feedback loops make a single "winner-take-all" platform monopoly inevitable.
- Market tipping is conditional rather than guaranteed. Market structure—whether an industry consolidates into a single monopoly, stabilizes as a competitive oligopoly, or fragments—is determined by three specific, testable conditions (Caillaud & Jullien, 2003; Iansiti & Lakhani, 2020).
Condition 1: Multi-Homing Costs
- Definition: The financial, technical, or cognitive friction incurred when a user or supplier simultaneously uses multiple competing platforms.
- Market Dynamics: When multi-homing is low-cost and frictionless, users switch platforms easily, which disperses aggregate volume, weakens network effects, and prevents single-firm tipping.
- Examples: Ride-hailing (drivers and riders simultaneously use Uber and Lyft) and accommodation (hosts list properties on both Airbnb and Booking.com).
- Strategic Rule: Low multi-homing costs promote stable oligopolies or fragmented markets; high multi-homing costs lock participants to one platform, driving winner-take-most outcomes.
Condition 2: Demand and Product Differentiation
- Definition: The degree to which market participants possess diverse preferences, specialized use cases, or differentiated supply requirements.
- Market Dynamics: When platforms can carve out distinct, non-overlapping niches, heterogeneous consumer demand prevents a single generalized platform from capturing the entire market.
- Examples: Audio streaming (Spotify's algorithmic discovery vs. Apple Music's integrated ecosystem) and professional networking vs. public discourse (LinkedIn vs. X/Twitter).
- Strategic Rule: Broad differentiation insulates platforms against market tipping; highly commoditized, homogeneous markets tip rapidly toward a single low-cost or high-scale winner.
Condition 3: Feedback Loop Strength and Compounding Velocity
- Definition: The rate at which incremental transaction data refines model output accuracy, directly boosting product utility for subsequent users.
- Market Dynamics: Strong feedback loops compound market share advantages rapidly, creating an expanding performance gap between market leaders and challengers.
- Examples: TikTok's high-frequency algorithmic engagement loop and Netflix's hyper-personalized content recommendations exhibit high loop strength. In contrast, mapping applications (Google Maps vs. Apple Maps) demonstrate moderate or decelerating feedback loop returns once core mapping accuracy is achieved.
- Strategic Rule: Rapid, steep data feedback loops trigger fast market tipping; weak or easily saturating data loops leave room for challenger entry.
Strategic Framework for Market Structure Diagnostics
Structural Diagnostic Decision Tree
- Inquiry 1: Is multi-homing easy and low-cost for both sides of the platform?
- If YES: Evaluate Market Differentiation.
- Inquiry 2A: Is there meaningful differentiation across consumer segments?
- If YES: Market likely fragments by specialized niche.
- If NO: Market stabilizes as a competitive, multi-platform oligopoly.
- If NO (Single-Homing Locks In): Evaluate Feedback Loop Strength.
- Inquiry 2B: Are the data and algorithmic feedback loops compounding and strong?
- If YES: Market tips directly into a Winner-Take-Most / Monopoly architecture.
- If NO: Market may stabilize as a concentrated, high-barrier oligopoly.
Systemic Implications for Strategic Planning
- Enterprise leadership must avoid the assumption that scale alone guarantees market dominance.
- Defensive strategy focuses on raising rival multi-homing costs, establishing distinct specialized differentiation, and maximizing the compounding velocity of algorithmic feedback loops.
Case Analysis: X Fire Paintball & Airsoft

Enterprise Context and Operating Profile
- Company Profile: X Fire Paintball & Airsoft is an independent specialty retailer offering paintball equipment, airsoft weapons, tactical uniforms, protective gear, maintenance accessories, and specialized field repair services.
- Operational History: Established in the late 1990s, the company operates physical brick-and-mortar retail storefronts across Massachusetts and New Hampshire while catering to both casual hobbyists and professional competitive teams.
- Strategic Crossroads: Navigating its commercial exposure to third-party digital platforms, framing the dilemma: "Is Amazon a Friend or Foe?"
Analytical Diagnostic Framework: Unpacking Structural Power
- Dimension 1: Bottleneck Identification
- Diagnostic Focus: Identify where access to digital consumers is controlled. Assess whether X Fire can achieve national reach without selling through Amazon's marketplace rails.
- Structural Assessment: While X Fire maintains localized physical storefronts and a direct-to-consumer e-commerce site, aggregate consumer discovery in the United States concentrates on Amazon ( market share). Amazon acts as a distribution bottleneck; bypassing it cuts X Fire off from broad consumer search traffic.
- Dimension 2: Ranking as Market Power
- Diagnostic Focus: Evaluate the criteria determining product visibility and checkout conversions.
- Structural Assessment: Sales volume concentrates within the "Buy Box" and the top three search results. Algorithmic ranking rules dictate sales allocations across competing merchants selling identical SKUs. Price, fulfillment speed via Fulfillment by Amazon (FBA), and ad spend dictate ranking, converting seller margins into platform ad revenue.
- Dimension 3: Platform as Competitor
- Diagnostic Focus: Analyze structural power asymmetries when the marketplace host competes directly by retailing identical tactical inventory.
- Structural Assessment: Amazon operates as both referee and player. It leverages platform-wide transaction data to identify high-margin SKUs, source inventory directly from manufacturers, and allocate default visibility to its own retail operations or private labels.
- Dimension 4: The Algorithmic Operating Layer
- Diagnostic Focus: Define the true competitive counterparty facing X Fire.
- Structural Assessment: X Fire does not compete against human merchants; it faces algorithmic pricing scrapers and programmatic inventory managers. To maintain sales, X Fire must deploy automated repricing software, entering a reactive algorithmic loop that risks triggering automated price wars down to zero economic profit.
- Dimension 5: Winner-Take-Most Market Testing
- Multi-homing Analysis: Consumers single-home on Amazon due to Prime shipping and convenience; sellers attempt to multi-home across Shopify, eBay, and independent sites, but high customer acquisition costs elsewhere force them back to Amazon.
- Differentiation Potential: While commodity airsoft pellets and standard protective gear are undifferentiated, custom airsoft tuning, local marker repairs, and specialized team consulting cannot be replicated algorithmically by an e-commerce platform.
- Feedback Loop Analysis: Amazon's transaction volume compounds its operational scale, logistics speed, and predictive search accuracy, widening the competitive gap against independent specialty retailers.
- Dimension 6: Formulating Strategic Response
- Strategic Assessment: Pure price competition on commoditized SKUs against Amazon's algorithmic marketplace is unsustainable for independent retailers.
- Strategic Reorientation: X Fire must treat Amazon merely as a high-cost customer acquisition funnel while insulating its core business through non-algorithmic, high-touch services: custom field maintenance, regional competitive event leagues, bespoke equipment tuning, and community-driven retail experiences.
Questions, Insight Checks, and Strategic Deliberations
Inquiry 1: Millisecond Moves and Strategic Assumptions
- Prompt: In traditional competition, a firm observes a rival's move and deliberates on a strategic response. What happens when competitive moves occur in milliseconds and are executed by autonomous machines? Which core assumption breaks first?
- Detailed Deliberation: The assumption of deliberate planning windows breaks down completely. When market moves execute in milliseconds, human decision-makers cannot participate in the operational cycle. Strategy shifts from planning periodic tactical moves to designing real-time, autonomous optimization rules, loss functions, and guardrail parameters.
Inquiry 2: Tacit Algorithmic Collusion vs. Standard Profit Maximization
- Prompt: If two competing retailers deploy pricing algorithms that independently learn that matching each other's prices above marginal cost maximizes long-run profits, does that constitute illegal collusion?
- Detailed Deliberation: Under current antitrust frameworks, this behavior generally does not meet the legal standard for illicit collusion because it lacks explicit human communication or deliberate coordination. Independent agents operating reinforcement learning algorithms learn to sustain supra-competitive price floors simply by pursuing individual profit maximization, exposing an enforcement gap in modern competition policy.
Inquiry 3: Challenger Dynamics Against Entrenched Pricing Systems
- Prompt: As a new market entrant, how do you compete against an incumbent whose dynamic pricing algorithm has been trained on five years of continuous transactional data?
- Detailed Deliberation: Directly challenging an incumbent on price optimization is ineffective due to the data asymmetry. Entrants must avoid the incumbent's algorithmic playing field by introducing non-linear business models: offering unbundled services, building proprietary distribution channels, introducing novel product architectures, or competing on unique attributes that the incumbent's optimization models are not designed to measure.
Inquiry 4: Platform Self-Preferencing vs. Competitive Strategy
- Prompt: If Amazon positions its own private-label products at the top of organic search results above third-party merchants, is that legitimate competitive strategy or an anticompetitive abuse of power?
- Detailed Deliberation: From a traditional business standpoint, firms routinely feature proprietary house brands to boost retail margins. However, when a platform operates an unavoidable digital bottleneck, self-preferencing leverages control over the market access gateway to extract value from competitors who have no alternative distribution path, which regulatory bodies like the EU classify as an abuse of platform power.
Inquiry 5: The ROI of Ranking vs. Product Quality
- Prompt: Why would a consumer goods manufacturer pay hundreds of thousands of dollars to purchase top search engine ranking placement rather than investing that capital directly into product quality enhancements?
- Detailed Deliberation: Due to the steep decay curve of consumer attention, even an exceptional product generates zero sales if it sits on Page 2 or lower in search results. Securing Position 1 (capturing CTR) guarantees immediate, high-volume consumer traffic. In digital markets, access to user attention frequently yields higher immediate revenue returns than marginal gains in physical product quality.
Inquiry 6: Explaining TikTok's Rise Against YouTube
- Prompt: How did TikTok achieve rapid scale and market capture despite YouTube's extensive historical head start and massive content library?
- Detailed Deliberation: Applying the Three Conditions framework explains this market entry:
- Differentiation: TikTok pioneered algorithmic micro-video feeds optimized entirely for immediate interest discovery rather than relying on explicit user subscriptions.
- Multi-homing: Viewers and content creators faced near-zero switching costs, allowing them to cross-post and multi-home across both platforms simultaneously.
- Feedback Loop Strength: TikTok created a real-time behavioral feedback loop: tracking exact watch times, video loops, and scrolls to personalize content feeds within a single browsing session, outpacing YouTube's slower, click-based feedback loops.
Inquiry 7: Challenging the Search Monopoly
- Prompt: Could an emergent tech firm mount a credible challenge to Google Search, and what specific structural conditions would be required to succeed?
- Detailed Deliberation: To break Google's structural bottleneck, a challenger must bypass traditional search behaviors entirely:
- Disrupt default distribution agreements by creating alternative hardware interfaces or conversational AI operating environments.
- Lower consumer switching frictions by integrating search directly into operating-system-level workflows.
- Deploy generative AI capabilities that bypass traditional ten-blue-link web results, providing direct answers that render organic search ranking less relevant.
Inquiry 8: Independent Market Diagnostic Exercise
- Prompt: Select a digital market not covered in the session: identify its primary competitive bottleneck, specify one condition that sustains it, and detail one condition that could dismantle it.
- Detailed Deliberation: Consider the Mobile Digital Gaming Market (dominated by the Apple iOS App Store and Google Play Store):
- Bottleneck: The native mobile operating system distribution rails and associated mandatory in-app billing engines.
- Sustaining Condition: High consumer single-homing costs (due to hardware investment and operating-system lock-in) coupled with contractual bans against side-loading third-party application stores.
- Dismantling Condition: Government regulatory action (e.g., European DMA enforcement requiring support for third-party app stores) combined with the growth of browser-based cloud streaming services that bypass native device software distribution entirely.