1/80
Looks like no tags are added yet.
Data Network Effects
Value from data-driven learning, the more the platform learns from your data via AI, the more valuable it is to each user.
Clough & Wu test
Two-part test for whether a data network effect is real: one user's data improves other users' experience, and the improvement reaches the current product quickly enough.
other, current, quickly
Clough & Wu test: one user's data must improve the experience of _____ users, and the improvements must reach the _____ product _____ enough to matter.
data, service, usage
Data flywheel: more _____ → better _____ → more _____ and repeat.
Chicken-and-egg problem
Each side of a platform wants the other side to be there first, so the platform must seed one side.
Multihoming
Users or providers being active on several platforms at once, which weakens lock-in.
Tipping
A platform market collapsing toward one dominant winner.
Liquidity
Enough participants on both sides for matches to happen quickly.
Lock-in
Switching costs that keep users on a platform.
infrastructure, foundation models, applications, agents
Three layers where AI value can be captured: _____ → _____ _____ → _____/_____.
Commoditization
Risk that AI models become interchangeable, pushing prices and profits down.
Gatekeepers
Role agents can take by sitting between firms and customers and controlling access.
Value creation
Potential value data enables (better decisions, lower cost, higher willingness to pay), whether or not the firm keeps it.
Value capture
The profit a firm actually keeps from the value created, after competition, partners and costs.
Externality
A cost (or benefit) that falls on others, such as society or the environment, and not on the firm.
sense, seize, reconfigure
Dynamic capabilities: _____ opportunities and threats; _____ them by mobilising resources; _____ resources and processes.
Gold mines
S4 matrix cell: high attractiveness, easy to do.
Moon shots
S4 matrix cell: high attractiveness, hard to do.
Quick wins
S4 matrix cell: lower attractiveness, easy to do.
Questionable
S4 matrix cell: lower attractiveness, hard to do.
problem, data, decision, mechanism, owner, capabilities, KPIs, constraints
A value-creation opportunity is specified by the business _____, the _____ needed, the _____ it improves, the revenue or cost _____, the _____, the required _____, the _____ and the _____.
trust, pay, society
Three questions on limits: Can we _____ the output? Does it _____, and who pays? What does it cost _____?
firm, other firms, society, now, later
Second-order costs are mapped by who pays (_____, _____ _____ or _____) and when (_____ or _____).
adopt, mitigate, abstain
Three options when facing AI's costs: _____, _____ or _____.
stockouts, food waste, RFID
Foodles, a smart-fridge caterer, suffers both _____ and _____ _____ despite rich _____ and sales data.
KPI ownership, opaque, governance, forecasting
Foodles root causes: missing _____ _____, _____ models, weak _____ and weak _____ skills.
outsource, internal, hybrid
Three ways Foodles could get forecasting: _____, _____ or _____.
Fountain9
External forecasting vendor that cut Foodles' stockouts by 50% and waste by 40%.
Rappi
Company that acquired Fountain9, after which Fountain9 announced its shutdown in two months.
forecast horizon, freelance, in-house, before
Foodles debrief: negotiate the _____ _____ extension, hire a _____ expert, leverage _____ models and planners, and go live _____ the cutoff.
driver, enabler, coach, empowerer
Four roles Data can play depending on urgency and maturity: _____, _____, _____, _____.
reversibility, business teams, decisions
Foodles lessons: _____, trained _____ _____ and better _____ over better models.
definition, mechanism, importance, example, limitation
Part 1 answer structure: _____ → _____ → strategic _____ → course _____ → _____.
storage, processing, sensors, network
Why data volumes exploded, four factors: drop in _____ costs, rise in _____ power, more _____, mature _____ infrastructure. Together they made advanced analytics cost-effective.
Labour
Source of value: value from the effort to create, store, analyze, transport data. Explains cost-plus pricing of datasets. High usage of this source does not guarantee usage value.
Usage
Source of value: value from what data is utilized for (decisions, revenue, cost). Explains why the same data is worth different amounts to different firms.
Scarcity
Source of value: value from limited access to the data. Explains exclusive datasets.
data, information, knowledge, actions, results
Data value chain (value of information):
_____ → _____ → _____ → _____ → _____
descriptive, diagnostic, predictive, prescriptive
Analytics types:
_____ (what happened?)
_____ (why did it happen?)
_____ (what could happen?)
_____ (what should be done?)
Commodity
Form of value: data sold or traded as a product. Slide examples: US geolocation data market; detailed data on 700 million people; data products from flight and telecom data; a 250 M$ deal over 5 years. Kroger monetizing loyalty-card data (96% of purchases tied to a card; 2,000+ variables; 60 M households).
Lever
Form of value: data improves an existing activity. Predictive maintenance saving tens of millions $ via lower failure rates; 50% of 10,000 products sold per minute come from personalized recommendations.
Asset
Form of value: distinctive data at the core of the competitive advantage. Slide example: 4 M cars each sending video from 8 cameras, 1,000+ labelers, "shadow mode" training whether Autopilot is on or not, against ~100 cars for GM Cruise and 1,000 for Waymo.
bigger, marketing, operations, value
Trends for forms of value:
Commodity: gets _____
Lever: shifting from _____ to _____
Asset: biggest form of _____
monetizes, privacy, safest, copy, defensible, slow
Trade-off for forms of value:
Commodity: _____ fast but raises _____ and legitimacy risks
Lever: _____ but easiest to _____
Asset: most _____ but _____ and costly to build
attractiveness, competitive, cost, differentiation
Profit above the cost of capital comes from industry _____ (corporate strategy, where to play) and _____ advantage (business strategy, how to win). Advantage is either _____ or _____.
lower, higher
Advantage shows as a gap between ROIC and WACC.
Similar price, _____ cost → cost advantage.
Similar cost, _____ price → price advantage.
Both differ → advantage from both.
income, investment, turnover, operational, tax, fixed, working
ROI Tree:
ROI = _____ / _____. Income splits into _____ (volume: market increase, market share, diversification; price: pricing policy, higher value-added markets) minus _____ and other costs (productivity, cost reduction) and _____. Investment splits into _____ assets (capacity use) and _____ capital (stocks, client and supplier payment terms).
Revenue
Value creation target: capture emerging demand via social analysis; value-adding services, recommendation, customization; market share through differentiation; loyalty via targeted coupons; sell data services; better price mix with customized pricing.
Cost
Value creation target: automate with ML; reduce errors and scrap with better predictions; better throughput via planning and maintenance; reduce inventory with predictive planning.
electricity, CO2, water
Every stage of the value chain (generation, collection, analysis, exchange) runs on devices, networks and data centres, across production, usage and end of life. Impacts: _____, resources, _____, _____. Value creation is not to be isolated from environmental considerations.
Direct
Same-side network effects, each new user makes the product more valuable to users on the same side (messaging, social networks).
Indirect
Cross-side network effects, more users on one side make it more valuable to the other side (riders ↔ drivers, buyers ↔ sellers).
congestion, variety, boomerang
Not all network effects are positive: _____, low-quality _____ and "_____" effects create endogenous failures.
Utility
Network effect type: identical users, transient value. Example: an Uber ride.
Variety
Network effect type: heterogeneous users, transient value. Example: Ebay, app stores.
Quantity
Network effect type: identical users, persistent value. Example: the telephone network.
Accumulation
Network effect type: heterogeneous users, persistent value. Example: Wikipedia, Reddit.
User Network Effects
Value from the network's size.
Data Network Effects
Value from data-driven learning, the more the platform learns from your data via AI, the more valuable it is to each user.
attract, match, trust, liquidity
Jobs data do for a platform:
_____ (personalization, targeting)
_____ (ranking, search, recommendation)
_____ (ratings, fraud detection, verification)
_____ & balance (forecast supply and demand, pricing and incentives).
connected, consumption, moderate, profitability, rebound
_____ products can be shared, so fewer are produced. But when products become services, _____ grows. The positive impact holds only if service consumption stays _____, which means the company must reduce its service's desirability, and so its _____. This is the _____ effect.
Resources
What the firm HAS: tangible (financial, physical), intangible (technology, reputation, culture), human (skills, communication, motivation).
Capabilities
What the firm DOES: e.g. the ability to integrate, explore and analyze big datasets using specific resources.
valuable, rare, inimitable, substitutable
Resources and capabilities create sustained advantage only if _____, _____, _____, non-_____ and embedded in the organization's processes.
generation, collection, analysis, exchange
Data value chain (operational):
_____ (capture & record: structured, semi-structured, unstructured; apps, ERP, IoT, social)
→ _____ (acquire batch or stream, pre-process, store, secure)
→ _____ (semantic analysis, predictive/descriptive/prescriptive models, visualisation)
→ _____ (personal use for decisions, trading).
insight, responsiveness
"The useless data scientist": _____ is only one component of the ability to sense, seize and reconfigure. The organization must have high _____: flexible operations, fast redeployment, no inertia. These together create value.
collection, labelling, storage, compute, human
Direct costs of data resources and capabilities:
_____: automated or crowdsourced.
_____: specialist annotation (Alibaba health); micro-workers, ~1 hour and $16 per image for self-driving; Scale AI, 100,000+ contractors.
_____ & security: cyber insurance rising >100% a year by end-2021; 2025 breaches at Dior, Coinbase (up to $400 M), Co-op, M&S; insurance for AI underperformance.
_____: cloud ~25% of revenue for AI firms; retraining is ongoing (data drift); inference costs more than classic software. Llama 3.1: 54 days, 16,000 H100s ≈ $500 M.
_____ cost: preparation (10–15% of revenue on cleaning and labeling), verification, "artificial artificial intelligence"; a robotaxi needs ~1.5 humans.
misuse, discrimination, loyalty
Indirect costs of data resources and capabilities:
_____ of data.
PR damage from price _____.
_____ loss when consumers feel fooled.
Make
Option for acquiring resources: Internal human and technical resources.
Pros: Full customization; total control of value capture.
Cons: No scale effect; no external expertise; needs high internal expertise.
Source
Option for acquiring resources: Off-the-shelf products and services.
Pros: Scale effect; external expertise.
Cons: Lower customization; lock-in and dependence; low capitalization of expertise.
Buy
Option for acquiring resources: Acquire a company and its resources.
Pros: Time to delivery; external expertise; control of value capture.
Cons: Integration execution risk; cost.
Partner
Option for acquiring resources: Pool resources with another company.
Pros: Flexible over time; lower risk (shared costs).
Cons: Medium-term alignment of interests; value sharing; limited customization.
migration, retrenchment, establishment, cost, product, diffusion
Why open a resource? Value _____ ("razor and blades": opening increases diffusion and profits in a complement); resource _____ (volunteers maintain it: IBM & Linux at 20% of the cost of a proprietary OS); standard _____ / platform. Conditions: Opening drastically cuts its _____; the firm has a protected complementary _____; opening increases _____.
quick wins, gold mines, moon shots, marginal, capabilities, flexibility
Sequencing rules:
High short-term competitive pressure or falling performance → start with a mix of _____ _____ and _____ _____.
Less pressure → start with _____ _____ to pave the way for a sustainable advantage.
Leverage: if opportunities share the same resources (human, technical, cultural), launching them together lowers _____ cost through scale.
Strategic fit: data delivers more value when it applies to the firm's distinctive _____ (manufacturing leaders gain more from automation and optimisation). The stronger the existing capability, the higher the priority.
_____: the assessment is dynamic; opportunities get easier (bigger asset base, new tech) or more attractive (client habits change). Its main benefit is a shared vision, not a fixed plan. It also eases acquisition and partnership decisions.
categories, causation, proxy
Ways output misleads:
_____ are human choices. Is a tomato a fruit or a vegetable? Botanically a fruit, legally a vegetable, a fruit for the EU jam directive. Labels encode decisions, and models inherit them.
Correlation ≠ _____. Spurious correlations. A strong correlation is a question to investigate: go and see the cause, form new hypotheses, run experiments.
Optimizes the wrong _____. Algorithms define a success criterion (fastest route) and ignore others (most beautiful, safest, greenest)
opaque, scalable, feedback, proxies, contested
"Weapon of math destruction" warning signs:
_____
_____
damaging _____ loop
crude _____ (who people are, not what they did)
cannot be _____.
The more signs, the more harm at scale without accountability.
Eroded signals
Way AI destroys value: Everyone uses AI to write and to screen, so polish stops signalling talent and strong applicants get missed.
Broken markets
Way AI destroys value: Insuring against AI errors, or faking AI-proofed documents, pushes risk onto the whole market.
Crowded-out human effort
Way AI destroys value: Machines flood a creative category; people withdraw; new work falls.
Quality & reliability debt
Way AI destroys value: More code shipped early, then slower reviews and rising incidents.
Arms races that cancel out
Way AI destroys value: One AI generates objections, another processes them; costs rise on both sides for no net gain.