Assortment Elasticity Modeling

Measures how the demand for Product A changes when Product B is added or removed from the assortment. Unlike price elasticity, this focuses on substitution effects and cannibalization risk.

• Substitutes: Adding a similar private-label item may decrease demand for the national brand by 15%.
• Complements: Removing a specific charger may reduce the sales velocity of the associated device.
• Calculation: Often derived from multinomial logit (MNL) models using historical transaction data.

Predicts the likelihood that a customer will purchase an alternative item if their first choice is unavailable, rather than abandoning the session.

• No-Compromise Users: High brand loyalty segments show near-zero transfer probability.
• Attribute-Based Transfer: Customers often switch to items sharing the same color, size, or price band.
• Revenue Impact: Models must distinguish between transfers to higher-margin items versus forced discounting to retain the sale.

Isolates the net financial impact of an assortment change by separating cannibalized revenue from incremental category growth.

• Gross Lift: The total sales of the newly introduced item.
• Cannibalization Loss: The sales reduction in existing items directly attributed to the new introduction.
• Net Incrementality: Gross Lift minus Cannibalization Loss. A positive value indicates true category expansion, not just market share shifting.

Relies on econometric frameworks like the Multinomial Logit (MNL) or Nested Logit models to simulate consumer decision-making.

• Utility Maximization: Assumes a consumer selects the product that provides the highest latent utility.
• Independence of Irrelevant Alternatives (IIA): A key property of MNL models requiring that the ratio of choice probabilities between two items is unaffected by other alternatives.
• Hierarchical Nesting: Nested logit relaxes IIA by grouping similar products into nests (e.g., 'Diet Drinks' vs. 'Regular Drinks') to model correlation.

Decomposes elasticity not just by SKU, but by specific product attributes to guide assortment depth tuning.

• Color Elasticity: Adding a 'Midnight Black' variant may grow the category by 8%, while 'Neon Green' adds only 1%.
• Size Curves: Models predict the optimal ratio of small, medium, and large units to stock based on regional demographic elasticity.
• Price Tier Gaps: Identifies 'white space' where no product exists at a specific price point, quantifying the revenue lost by not serving that value-conscious or premium segment.

A computational layer that runs 'what-if' analyses on virtual assortment configurations before physical implementation.

• Virtual Delisting: Simulates removing a low-margin item to verify if 90% of its revenue transfers to higher-margin alternatives.
• Range Extension: Projects the total category volume if a new organic line is introduced alongside the conventional line.
• Constraint Optimization: Solves for the optimal product mix that maximizes revenue subject to hard constraints like shelf-space capacity or minimum supplier quotas.

A predictive framework that estimates which alternative product a customer will purchase if their first choice is out of stock. While elasticity modeling quantifies the revenue impact of an item's presence, transference modeling predicts the cannibalization or substitution behavior that occurs when an item is removed. It uses product affinity graphs and historical session data to calculate the probability that demand will flow to a specific substitute rather than being lost entirely.

A reinforcement learning agent that dynamically selects which products to display by balancing exploration of new items with exploitation of known high-performers. Unlike static elasticity models that pre-calculate item importance, the bandit learns in real-time, conditioning decisions on user context and session signals. The reward function often integrates elasticity coefficients to prioritize items with high incremental revenue potential.

An analytical method that identifies when the introduction or promotion of one product reduces the sales of another similar item in the same catalog. This is the negative counterpart to elasticity modeling. Key techniques include:

• Incremental lift analysis isolating the new item's effect
• Cross-elasticity matrices measuring pairwise substitution effects
• Market basket decomposition to detect shifting share of wallet Without this detection, a positive elasticity score for a new item may mask zero-sum revenue outcomes.

A causal inference technique that isolates the incremental revenue generated by a specific merchandising change from confounding factors like price changes, marketing campaigns, or seasonality. Elasticity models provide the predicted lift; attribution validates it post-hoc. Common methods include difference-in-differences analysis and synthetic control modeling, which construct a counterfactual baseline to measure the true causal effect of an assortment adjustment.

A ranking signal that down-weights or up-weights a product's search score based on its real-time inventory position. Elasticity models inform which items are critical to display, but availability-weighted relevance ensures those items are actually purchasable. The combined system prevents high-elasticity items with zero stock from occupying valuable screen real estate, instead routing demand to the next-best available alternative.

The strategic balancing of carrying a wide variety of product categories versus deep selection within a single category. Elasticity modeling quantifies the marginal revenue contribution of each additional SKU, enabling a data-driven trade-off. The goal is to satisfy heterogeneous local demand without bloating inventory carrying costs. Optimization solvers use elasticity curves to find the point where the next item added yields diminishing returns.