Every concept, defined.

also: quasi-hyperbolic discounting · beta-delta model · present bias

Hyperbolic discounting is a time-inconsistent preference pattern where people place disproportionate weight on immediate rewards and steeply discount near-term future payoffs, then flatten their discount curve for more distant periods. It explains why subscribers sign up enthusiastically but churn when the first renewal charge arrives.

also: cumulative prospect theory · CPT

Prospect theory, developed by Kahneman and Tversky, describes how people choose under risk: outcomes are evaluated as gains or losses relative to a reference point, losses weigh roughly 2.25× more than equivalent gains, and both gains and losses exhibit diminishing sensitivity.

also: loss aversion coefficient · lambda parameter

Loss aversion is the empirical finding that the psychological impact of a loss is roughly twice the impact of an equivalent gain — the loss aversion coefficient λ is typically estimated at 2.0 to 2.5 in controlled settings but varies with stakes, platform investment, and framing context.

The endowment effect is the finding that people demand significantly more to give up an object than they would pay to acquire it. In digital products, activated accounts and populated workspaces create psychological ownership that makes downgrade and cancellation substantially harder than the symmetric purchase decision.

The sunk cost fallacy is the tendency to continue investing in a project, product, or relationship because of cumulative prior investment, regardless of whether additional investment produces positive expected value. In product adoption, it creates both productive lock-in (customization investment) and destructive persistence (continuing to use a failing tool).

Choice architecture is the deliberate design of the context in which decisions are made — the ordering of options, the default selection, the framing of trade-offs, and the number of alternatives presented. In digital products, it determines which outcome occurs when users operate on low cognitive engagement.

also: asymmetric dominance effect · attraction effect

The decoy effect is the phenomenon where adding a third, asymmetrically dominated option to a choice set shifts preference toward the target option. In dynamic pricing it is operationalized by introducing a decoy tier designed to be chosen against, increasing conversion to the intended price point by 20–40% in controlled experiments.

Mental accounting, formalized by Richard Thaler, is the tendency to segregate money into non-fungible categories based on source, intended use, or currency, and to apply different decision rules to each. In multi-currency e-commerce it explains why identical purchases feel costlier in a native currency than in a secondary one.

also: construal level theory · CLT · psychological distance

Temporal Construal Theory (Trope and Liberman, 2003) holds that people mentally represent distant events in abstract, high-level terms and near events in concrete, low-level terms. Landing-page copy that matches the reader's psychological distance — abstract for cold traffic, concrete for hot — converts systematically better than generic copy.

also: effort justification · labor-induced valuation

The IKEA Effect (Norton, Mochon, Ariely, 2012) is the finding that people place disproportionately high value on objects they have partially created themselves — a 63% WTP premium over identical pre-assembled items. In product adoption it converts early configuration effort into durable retention via effort justification.

also: network effect · direct network effects · indirect network effects

A network effect exists when the value of a product to each user increases with the number of other users. Direct network effects (messaging apps) work within a single user side; indirect or cross-side network effects (marketplaces) operate across distinct user groups. Strong network effects create winner-take-most markets.

also: switching cost engineering · lock-in

Switching costs are the total economic, procedural, and psychological burden a customer incurs when moving from one vendor to another. Klemperer's taxonomy decomposes them into transaction, learning, contractual, uncertainty, and psychological components, each with different decay rates and defensive strategies.

Zero marginal cost describes the economics of digital goods, where serving one additional user imposes essentially no incremental production cost. This property inverts classical pricing theory: optimal pricing becomes a discovery problem over willingness-to-pay distributions, not a cost-plus calculation.

also: two-sided market · multi-sided platform · platform economics

Platform dynamics are the economic mechanics of businesses that connect distinct user groups and extract value from the interactions between them. Platforms face chicken-and-egg launch problems, pricing asymmetries across sides, and governance choices around cannibalization and competitive entry.

Multi-homing describes users or suppliers that participate on multiple competing platforms simultaneously. High multi-homing on one or both sides of a two-sided market erodes the winner-take-most dynamic traditionally assumed for platforms, and is a critical predictor of vertical-SaaS competitive outcomes.

Platform cannibalization is the strategic decision by a platform to compete directly with its own complementors, entering adjacent product categories the platform's complementors originally filled. The decision follows a predictable envelope pattern once a category exceeds a threshold share of platform traffic.

also: API pricing · zero-marginal-cost pricing · value-based pricing

Information goods pricing is the pricing theory for products with near-zero marginal cost: APIs, SaaS, content. Optimal pricing departs from cost-plus and instead discovers willingness-to-pay distributions via bundling, tiering, two-part tariffs, and usage-based metering that align price to the value delivered per customer segment.

also: data network effects · data feedback loop

Data moats are defensive advantages that compound as a product accumulates proprietary usage data: each incremental interaction improves the model, which improves the product, which attracts more users, which generates more data. Unlike network effects they operate without direct user-to-user interaction.

also: MTA · touchpoint attribution

Multi-touch attribution assigns credit for a conversion across all marketing touchpoints in the customer journey, using rules (linear, time-decay, U-shaped) or data-driven models. Most MTA implementations confuse correlation with causation and systematically overvalue bottom-funnel channels that intercept already-decided buyers.

also: MMM · media mix modeling

Marketing mix modeling is a top-down econometric approach that estimates the causal contribution of each marketing channel to sales using aggregate historical data and controls for non-marketing drivers (price, distribution, seasonality, competition). Privacy-first MMM has returned as deterministic user-level tracking has eroded.

also: causal graph · DAG · directed acyclic graph

Causal inference is the statistical machinery for estimating causal effects from data rather than just describing correlations. In marketing it involves directed acyclic graphs (DAGs) for identifying confounders, instrumental variables for unobserved confounding, and quasi-experimental methods like difference-in-differences and synthetic control.

also: geo-lift · conversion lift test

Incrementality testing measures the causal lift produced by a marketing intervention by comparing treated units to untreated control units, typically using geographic randomization (geo-lift), user holdouts, or ghost bids. It is the empirical gold standard for answering 'what did this spend actually cause?'

also: PC algorithm · constraint-based causal discovery

Causal discovery is the family of algorithms that infer directed causal structure from observational data alone, using conditional-independence tests (PC, FCI) or score-based search (GES). Applied to business data it produces testable DAGs that replace ad-hoc causal intuitions with falsifiable hypotheses.

also: BSTS · Bayesian structural time series · synthetic control

CausalImpact is Google's open-source implementation of Bayesian structural time series for measuring the causal effect of a known intervention on a time-series outcome. It constructs a counterfactual from correlated control series and reports the posterior distribution of the cumulative effect with calibrated uncertainty.

also: DAG · Bayesian network · structural causal model

A Directed Acyclic Graph is the formal representation of a causal structure: nodes are variables, directed edges are direct causal effects, and no cycles are permitted. DAGs encode the identification assumptions needed to estimate causal effects from observational data via the back-door and front-door criteria.

also: UMM · triangulation · MMM plus MTA plus experimentation

A Unified Measurement Architecture combines Marketing Mix Modeling, Multi-Touch Attribution, and randomized incrementality experiments into a single decision stack, with each method calibrated against the others. MMM supplies top-down causal anchors, MTA supplies in-flight optimization, experiments supply the ground truth.

also: geographic split test · geo experiment · market holdout

Geo-lift testing is an experimental design that randomizes marketing treatment across geographic markets (DMAs, metros, countries) to measure incremental impact when user-level randomization is infeasible. Synthetic control and open-source libraries like GeoLift construct matched control markets from pre-period covariates.

also: Bayesian statistics · Bayes' theorem

Bayesian inference updates prior beliefs about a parameter using observed data via Bayes' theorem to produce a posterior distribution. In A/B testing it directly answers 'what is the probability that B beats A?' — the question product teams actually ask — unlike the indirect counterfactual framing of frequentist p-values.

also: Cox proportional hazards · hazard model · time-to-event

Survival analysis models time-to-event data — how long until a customer churns, a subscription renews, a machine fails — accounting for censored observations where the event has not yet occurred. Cox proportional hazards is the standard semi-parametric model; deep recurrent survival models handle non-proportional hazards.

also: cohort economics · retention curve

Cohort analysis groups users by a shared origin event (acquisition month, first purchase, signup source) and tracks behavior over time for each group. It separates true retention from the compositional distortion caused by new-user dilution, and is the foundational unit of analysis for subscription economics.

also: outlier detection · isolation forest

Anomaly detection identifies observations that deviate meaningfully from expected behavior, accounting for trend, seasonality, and variance. In revenue data it separates true incidents (payment outages, pricing bugs) from normal fluctuation. Isolation forests and Prophet-based decomposition are the practical workhorses.

also: PMF

Product-market fit is the empirical condition where a cohort's retention curve flattens above zero — a group of users has found sufficient value to make the product a persistent part of their behavior. It is not a feeling; it is a quantifiable property of retention, NPS decomposition, and usage depth.

also: dbt · modern data stack · ELT

Analytics engineering is the discipline of building reliable, tested, version-controlled transformations on top of a cloud warehouse, bridging data engineering and analysis. Tools like dbt, Dagster, and Airbyte formalize a software-engineering workflow for SQL transformations with tests, documentation, and lineage.

also: semantic layer · metric definition framework

A metric ontology is a versioned, centrally-governed definition of every metric an organization uses, specifying the grain, filters, time-window, and source tables so that the same metric produces identical values regardless of tool, dashboard, or analyst. It prevents the drift that silently corrupts data-driven decisions.

also: CAC · LTV · LTV to CAC ratio · payback period

Unit economics is the financial performance of a single customer (or transaction) decomposed into acquisition cost, gross margin per period, retention, and payback period. Cohort-level unit economics — computed per acquisition cohort rather than rolled up — is the only form that survives growth-driven distortion.

also: optional stopping · sequential testing

The peeking problem is the inflation of false-positive rates that occurs when a frequentist A/B test is repeatedly evaluated before reaching its pre-registered sample size. A nominal 5% false-positive rate can become 20–30% under daily peeking. Bayesian testing and sequential-analysis methods eliminate the problem.

also: Cox PH · Cox regression

The Cox proportional hazards model is the semi-parametric workhorse of survival analysis: it estimates how covariates multiply the baseline hazard rate without requiring a parametric form for the baseline. It yields interpretable hazard ratios under the assumption that the ratio is constant over time.

also: iForest · tree-based anomaly detection

Isolation Forest is a tree-based anomaly detection algorithm that scores observations by how easily they can be isolated via random recursive partitioning. Anomalies are isolated in few splits; normal points require many. The algorithm handles mixed feature types without density estimation or distance calculations.

also: decision systems · automated analytics

The dashboards-to-decisions gap is the structural failure of analytics investment: teams produce more dashboards but decisions don't get better or faster. Closing the gap requires moving from descriptive reports to decision systems — pre-specified trigger thresholds, automated action routing, and outcome logging for calibration.

Contextual bandits are online learning algorithms that choose an action (a price, a layout, a recommendation) given a context (user features), observe a reward, and update their policy to balance exploration and exploitation. They are the modern foundation of real-time personalization and dynamic pricing.

also: item embeddings · transformer embeddings

Product embeddings are dense vector representations of items in a learned semantic space, such that geometrically close items are similar in the behavioral or content sense. Transformer-based embeddings trained on session sequences capture nuanced substitute/complement relationships that simple collaborative filtering misses.

The cold-start problem describes recommendation and ranking systems' inability to serve new users or new items with no interaction history. Few-shot learning, meta-learning (MAML), and prototypical networks address it by learning initializations that adapt quickly from sparse signals.

Dynamic pricing is the practice of adjusting prices in real time based on demand, inventory, user context, competition, or time. Machine-learned pricing uses contextual bandits and demand models, but introduces fairness, perception, and regulatory considerations that static pricing avoids.

Conformal prediction is a model-agnostic framework for producing calibrated prediction intervals with finite-sample coverage guarantees. Applied to demand forecasting it replaces opaque point predictions with intervals that provably contain the true demand at a specified confidence level.

also: heterogeneous treatment effects · HTE · incremental response modeling

Uplift modeling estimates the heterogeneous causal effect of a treatment — a promotion, a feature, a message — on each individual. Unlike propensity or response models, it explicitly targets the difference in outcome between the treated and untreated counterfactual, enabling promotion budgets to be spent only on the persuadable segment.

also: LTR · L2R · search ranking

Learning to Rank is the class of supervised machine learning algorithms that optimize the ordering of a result set — search results, recommendations, product rankings — for revenue, engagement, or relevance. Pairwise (RankNet) and listwise (LambdaMART, ListNet) objectives are the dominant training paradigms.

also: GNN · GraphSAGE · graph convolution

Graph Neural Networks learn representations over graph-structured data by message-passing between nodes. In e-commerce cross-sell, GNNs ingest the user-item-category graph and produce recommendations that respect product hierarchy, co-purchase relationships, and session structure — outperforming flat collaborative filtering by 15–25% on business metrics.

also: session-based recommendation · in-session personalization

Real-time personalization adapts product recommendations, content, and pricing within a session based on immediate behavior signals — dwell time, scroll depth, added items, search queries. Contextual-bandit systems with streaming feature stores enable policy updates in milliseconds, producing 10–25% lift over batch-trained models.

also: transaction fraud scoring · payment fraud ML

Real-time fraud detection scores checkout transactions within latency budgets of 50–200 ms to decide allow, challenge, or block. Production systems combine gradient boosting (feature-rich), graph features (linked-device, shared-card), and autoencoder-based anomaly scoring under extreme class imbalance and adversarial adaptation.

also: LLM catalog enrichment · catalog enrichment · AI catalog enrichment

LLM-powered catalog enrichment uses large language models to generate product descriptions, attributes, categorization, and structured data from sparse inputs (SKU name, supplier feed) at scale. It eliminates the manual-curation bottleneck that has historically limited catalog coverage in marketplace and retail businesses.

also: brand salience · Ehrenberg-Bass

Mental availability is the propensity of a brand to be thought of in buying situations. It is a network property across the category's entry points, not a single recall score, and is the dominant predictor of market share according to Ehrenberg-Bass research.

also: CEPs

Category Entry Points are the specific cues — occasions, needs, moods, locations, social contexts — that trigger a consumer to think about a category. CEPs are the measurable unit of mental availability. A brand's CEP portfolio breadth is a stronger driver of share than the depth of association with any single CEP.

also: brand value · brand strength

Brand equity is the premium — in sales, pricing power, and LTV — attributable to the brand identity beyond functional product attributes. It is an asset that compounds with consistent investment and decays with over-optimization toward short-term performance marketing, making brand-vs-performance allocation a portfolio problem.

also: JTBD

Jobs-to-be-Done is a segmentation framework that groups customers by the underlying 'job' they hire a product to do, rather than by demographic or firmographic attributes. Modern implementations use NLP on customer interviews and reviews to derive job spaces at scale.

also: cost per attention second · CPAS · attention-based media buying

Attention economics prices media against the attention it actually produces, not impressions or clicks. A display banner costing 2.50 CPM may generate zero attention seconds, making its true cost per attention second far higher than a connected-TV ad priced at 25 CPM. Correcting for attention reshuffles media-plan ROI rankings.

also: ad fatigue · entropy-based creative decay

Creative fatigue is the decay of advertising performance as the same creative is repeatedly shown to the same audience. Entropy-based detection metrics identify the inflection point earlier than CTR decline — typically 40–60% through the CTR curve — enabling proactive rotation before wasted spend accumulates.

also: topical authority · compounding content · SEO moat

A content moat is a defensive advantage built from a deep, interconnected portfolio of articles on a topic such that search engines treat the publisher as the topical authority. Returns compound: each new article benefits from the authority of the existing set, while competitors cannot catch up without a comparable investment.

also: brand-vs-performance allocation · Markowitz for marketing

Brand-performance portfolio optimization applies Markowitz mean-variance portfolio theory to the allocation of marketing budget between brand-building (long-duration, uncertain ROI) and performance (short-duration, tightly measurable). The efficient frontier reveals allocations that dominate the common 60/40 heuristic in both expected return and variance.

also: OKR cascade · goal decomposition

The strategy-execution gap is the systematic distance between stated strategy and observed team behavior. The reason most strategic plans fail is not bad strategy but poor decomposition into measurable objectives at the team and individual level. Quantitative goal cascades (OKRs done well) close the gap.

also: competitive intelligence · AI-driven competitor monitoring

Market sensing with LLMs is the practice of using large language models to continuously monitor competitive messaging, product changes, and strategic positioning across unstructured signal sources (news, SEC filings, job listings, changelogs). It replaces quarterly competitive-intelligence reports with continuous, structured signal feeds.

also: agent-based marketing simulation · funnel model · full funnel simulation · simulation model · marketplace simulation

Full-funnel simulation models the connected dynamics of awareness, consideration, conversion, and retention using agent-based or system-dynamics methods. It reveals nonlinear budget allocation effects — that doubling top-of-funnel spend may not double conversions if mid-funnel conversion rates are the binding constraint.

also: Goodhart's Law · brand erosion from performance

The hidden cost of over-optimization is Goodhart's Law applied to marketing: when a bidding algorithm targets short-term conversion KPIs exclusively, it systematically degrades brand equity and long-term customer value. The damage is invisible in the optimization metric but visible in downstream retention and organic traffic.