Merchandising, Pricing and Revenue Management

Introduction and Context

Digital commerce has transformed merchandising from a periodic exercise in selecting products for physical shelves into a continuous process of shaping customer choice across websites, mobile applications, marketplaces, search engines, social platforms and personalised communication. In an e-commerce or direct-to-consumer (D2C) business, the commercial result depends not only on what is sold, but also on how products are grouped, ranked, priced, promoted, recommended and replenished.

Merchandising, pricing and revenue management are therefore interdependent capabilities. Assortment decisions determine the products available to customers; category management determines how those products are organised and presented; pricing establishes perceived and realised value; promotion affects short-term demand and long-term price expectations; and revenue management coordinates demand, inventory, margin and capacity over time.

The central managerial challenge is to maximise customer value and profitable growth simultaneously. A product that generates high sales but requires heavy discounting, costly fulfilment and frequent returns may destroy value. Conversely, a high-margin product that is difficult to discover may remain commercially irrelevant. Effective digital retail management integrates customer analytics, economics, operations, technology and governance.

Learning Objectives

By the end of this chapter, the reader should be able to:

1. Explain digital assortment planning and category management.
2. Evaluate products using demand, margin, availability and customer-value measures.
3. Apply pricing, promotion and revenue-management principles in digital channels.
4. Interpret recommendation logic, demand sensing and competitive intelligence.
5. Analyse the strategic and ethical implications of algorithmic merchandising and pricing.
6. Design an integrated commercial decision process for an e-commerce or D2C business.

1. Digital Merchandising and Assortment Planning

1.1 Meaning and scope

Merchandising is the managerial process of selecting, presenting, pricing and promoting products so that customer needs are met and commercial objectives are achieved. Digital merchandising extends this definition by including search ranking, navigation, digital content, product discoverability, personalisation, availability messaging and channel-specific presentation.

An assortment is the set of products offered to a defined customer segment through a defined channel and time period. Its principal dimensions are:

• Breadth: number of product categories offered.
• Depth: number of variants within a category, such as sizes, colours or specifications.
• Price architecture: distribution of entry-level, mainstream, premium and luxury products.
• Role structure: traffic builders, profit generators, basket builders, image products and convenience products.
• Lifecycle coverage: new, growing, mature, seasonal and declining products.

Digital retailers can offer very broad assortments because online shelf space is relatively inexpensive. However, excessive choice can create search costs, decision difficulty, catalogue duplication and inventory complexity. The relevant question is not “How many products can be listed?” but “Which products create incremental customer and commercial value?”

1.2 Assortment economics

A product should be evaluated on its contribution to the total customer proposition rather than on sales alone. Important measures include gross margin, contribution margin, conversion, repeat purchase, return rate, fulfilment cost, stock-out risk and incremental basket value.

Gross margin is calculated as:

Gross-margin percentage is:

Net sales exclude discounts, refunds and cancellations where appropriate. Cost of goods sold includes the product acquisition or manufacturing cost. Gross margin does not necessarily represent profit because fulfilment, payment processing, marketing, customer service and return costs may still be incurred.

A more useful digital measure is contribution after variable commercial costs:

The formula assumes that the listed costs vary sufficiently with the transaction to support short- or medium-term decision-making. Fixed corporate costs are excluded. If a retailer uses different cost definitions, those definitions must be documented to preserve comparability.

1.3 Assortment planning process

A robust assortment-planning process generally includes the following stages:

1. Define the decision unit. Specify category, market, channel, customer segment and planning period.
2. Analyse customer missions. Determine whether customers are replenishing, browsing, comparing, gifting, solving a problem or seeking inspiration.
3. Segment products. Use need state, price tier, brand role, lifecycle and profitability.
4. Estimate demand and substitution. Consider whether removing one product transfers demand to another or causes the customer to leave.
5. Evaluate operational feasibility. Include supplier reliability, lead time, minimum order quantities, storage, returns and fulfilment constraints.
6. Set range and space rules. Digital space may be allocated through search rank, category pages, recommendation modules and advertising placements.
7. Test and revise. Use controlled experiments, seasonal reviews and post-period evaluation.

Digital assortment planning differs from traditional store planning because visibility is not fixed. A product can be physically available but commercially invisible if its content is poor, its search terms are incorrect or its ranking is weak. Consequently, digital availability has at least three dimensions: inventory availability, discoverability and suitability for the customer’s mission.

2. Category Management in Digital Retail

Category management treats a category as a strategic business unit managed according to customer needs and economic performance. In a digital environment, categories are simultaneously taxonomies, navigation systems, analytical units and merchandising interfaces.

2.1 Category roles

Common category roles include:

A category may perform different roles for different customer segments. For example, premium skincare can be an image category for a beauty specialist but a routine category for a loyal replenishment customer.

2.2 Digital category architecture

The digital category tree should reflect customer language and shopping behaviour rather than only internal purchasing structures. A technically correct but customer-incomprehensible taxonomy increases search abandonment. Product information management is therefore central to merchandising. Attributes such as size, compatibility, material, use case, colour and sustainability claims must be complete, standardised and searchable.

Search and browse systems typically combine lexical matching, attribute filters, behavioural signals, commercial rules and sometimes machine-learning ranking. Commercial rules must not override relevance so aggressively that customers receive unsuitable products merely because they are profitable or sponsored.

2.3 Category scorecards

A category scorecard should combine customer, commercial and operational indicators. Examples include:

• sessions and search volume;
• click-through and conversion rates;
• average order value;
• gross and contribution margin;
• repeat purchase rate;
• stock-out and cancellation rates;
• return and refund rates;
• delivery promise accuracy;
• customer ratings and review sentiment;
• percentage of sales from new products;
• incremental sales from recommendations or promotions.

A single metric can mislead. High conversion may reflect a narrow assortment that excludes potential customers; high revenue may result from deep discounting; high recommendation clicks may not create incremental orders. Metrics should therefore be interpreted within a causal and financial framework.

3. Pricing Architecture and Digital Price Setting

3.1 Pricing objectives

Pricing may pursue several objectives: profit maximisation, market entry, traffic generation, inventory liquidation, price image, customer acquisition, retention or fairness. These objectives can conflict. A low introductory price may attract customers but create reference-price problems and future resistance to normal pricing.

The price architecture should communicate a coherent ladder. Typical levels are opening price point, good/better/best, premium and exclusive. Gaps between levels should be meaningful in customer-perceived benefits, not merely numerical.

3.2 Basic demand and price elasticity

Price elasticity of demand measures the responsiveness of quantity demanded to a change in price:

For a continuous approximation:

Here, Q is quantity demanded, P is price, and ∆ denotes a small change. Demand elasticity is usually negative. If the absolute value of E_p is greater than 1, demand is elastic; if it is less than 1, demand is inelastic. The measure is local and can vary by customer segment, competitor prices, season, stock availability and promotional context.

If demand is represented by Q = a − bP, where a and b are positive constants, revenue is:

The unconstrained revenue-maximising price in this simple linear model is:

This result applies only to the specified linear demand model and ignores costs, competition, inventory, capacity, customer lifetime value and price discrimination. A profit-maximising price requires cost information and a relevant demand model.

3.3 Unit economics and break-even pricing

If variable cost per unit is c and price is P, unit contribution is:

If fixed costs are F and expected unit volume is Q, break-even quantity is:

The formula assumes P is greater than c and that fixed and variable costs are correctly classified. It is not a demand forecast; it identifies the volume required to cover fixed costs under an assumed price and unit contribution.

3.4 Dynamic pricing

Dynamic pricing changes prices in response to demand, inventory, time, competition, customer context or capacity. It is common in travel, food delivery, event tickets, ride-hailing and some retail categories. In product retail, dynamic pricing may operate at SKU, seller, region or customer-segment level.

A responsible pricing system should include:

• a clearly defined objective function;
• minimum margin and maximum discount constraints;
• inventory and replenishment conditions;
• competitor and market-quality controls;
• price-change frequency limits;
• approval rules for sensitive products;
• audit logs and rollback mechanisms;
• monitoring for unfair or discriminatory outcomes.

Prices should not change so frequently that customers perceive instability or manipulation. In many categories, a stable price combined with targeted value-added benefits is preferable to constant price movement.

4. Promotion Optimisation

Promotions include price reductions, coupons, bundles, loyalty benefits, free shipping, gifts, financing and limited-time offers. A promotion should be assessed on incremental contribution, not reported sales during the promotional period.

Let incremental units be the additional units attributable to a promotion, and let promotional contribution per unit be the selling price after discount minus variable costs. Promotional profit should also deduct campaign costs:

The principal analytical difficulty is estimating the counterfactual: what would have happened without the promotion? Comparing promoted sales with zero sales is incorrect because many customers would have purchased at the regular price.

Promotion evaluation should consider:

• baseline demand;
• cannibalisation of related products;
• forward buying and post-promotion decline;
• new-customer quality and repeat rate;
• increased returns;
• operational congestion;
• voucher leakage and fraud;
• supplier funding;
• changes in brand price perception.

A controlled holdout group, where commercially feasible, provides a stronger estimate of incrementality than simple before-and-after comparison. Randomisation must be designed carefully to avoid contamination between treated and untreated customers.

5. Recommendations and Personalised Merchandising

Recommendation systems help customers discover relevant products and help retailers increase basket size, conversion and retention. Common approaches include:

• Content-based recommendation: recommends products with attributes similar to those previously viewed or purchased.
• Collaborative filtering: uses similarities among customers or products based on behavioural data.
• Association rules: identify co-occurrence patterns, such as products frequently bought together.
• Contextual recommendation: incorporates device, location, time, season, mission and current session.
• Hybrid systems: combine several methods to reduce weaknesses of any single approach.

Association-rule analysis often uses support, confidence and lift. For products A and B:

Support measures prevalence, confidence measures the conditional frequency of B when A occurs, and lift compares that frequency with the baseline prevalence of B. A lift above 1 suggests positive association, although it does not prove causation. High lift based on very few transactions may be unstable.

Recommendation quality must be assessed beyond click-through rate. Relevant measures include add-to-cart rate, conversion, incremental revenue, contribution, diversity, novelty, coverage, cancellation and customer satisfaction. A recommendation can receive many clicks because it is visually prominent without producing incremental value.

Recommendation systems also create risks. Popularity bias can suppress new or niche products; feedback loops can make already visible products increasingly dominant; and personalisation can narrow customer choice. Governance should include explanations where appropriate, opt-out controls, privacy protection and fairness monitoring.

6. Demand Sensing and Forecasting

Demand sensing uses recent and high-frequency signals to update forecasts more rapidly than conventional periodic forecasting. Signals may include searches, page views, add-to-cart events, conversion, social activity, weather, competitor availability, advertising exposure, delivery promises and marketplace trends.

A simple exponential-smoothing model is:

where F_{t+1} is the forecast for the next period, A_t is actual demand in the current period, F_t is the previous forecast, and α is the smoothing parameter with 0 ≤ α ≤ 1. A high α responds quickly to recent changes; a low α produces a more stable forecast. The model assumes a relatively stable level without explicitly modelling trend or seasonality.

Forecast error can be measured as:

where n is the number of observations. Mean absolute error is expressed in units of demand. Mean absolute percentage error can be problematic when actual demand is zero or very small; therefore, metric selection should reflect the business context.

Demand sensing should not be confused with demand creation. A sudden increase in search activity may indicate genuine demand, a viral event, a competitor stock-out or paid-media exposure. Human interpretation and causal analysis remain important.

7. Competitive Intelligence

Competitive intelligence involves the systematic collection, validation and interpretation of information about competitors, substitutes, suppliers, platforms and market conditions. Relevant data include price, assortment, availability, delivery promise, reviews, promotions, content quality, marketplace ranking and seller activity.

A competitor price index may be calculated as:

An index of 100 indicates parity with the reference price; an index below 100 indicates a lower own price. The result is meaningful only when products are genuinely comparable and the reference competitor, taxes, shipping, pack size, warranty and service conditions are aligned.

Competitive intelligence should inform—not mechanically determine—pricing. Matching every competitor can create a race to the bottom, particularly when the competitor has a different cost structure or a temporary clearance motive. Strategic response may instead involve better content, faster fulfilment, bundles, service, authenticity guarantees or loyalty benefits.

8. Integrated Revenue Management

Revenue management coordinates price, demand, inventory and capacity to sell the right product to the right customer at the right time and through the right channel. In e-commerce, capacity may include inventory, fulfilment labour, delivery slots, advertising impressions, seller quality and customer-service resources.

A useful decision sequence is:

1. Forecast demand under alternative prices and promotions.
2. Estimate available and incoming inventory.
3. Calculate contribution under each commercial option.
4. Account for substitution, cannibalisation and future demand.
5. Apply customer, legal and brand constraints.
6. Select, test and monitor the decision.

The marginal value of inventory is particularly important. When stock is scarce, selling one unit today may prevent a higher-value sale later. When inventory is perishable or obsolete quickly, delaying a sale may destroy value. Consequently, the same product can rationally have different prices or promotional intensity across its lifecycle, subject to fairness and channel rules.

Worked Example: Promotion and contribution

An online retailer sells a product at £50. Variable product and fulfilment costs total £32 per unit. During a coupon campaign, the price becomes £45 and campaign cost is £2,000. Forecast sales without the campaign are 1,000 units; expected campaign sales are 1,300 units.

Regular contribution per unit is £50 − £32 = £18. Promotional contribution per unit before campaign cost is £45 − £32 = £13. The incremental volume is 1,300 − 1,000 = 300 units. Incremental contribution is therefore 300 × £13 = £3,900. After campaign cost, incremental promotional contribution is £3,900 − £2,000 = £1,900.

The campaign appears commercially attractive under the assumptions. However, the conclusion changes if the baseline forecast is overstated, if returns increase, or if many campaign customers would have purchased at £50. The example demonstrates why promotion evaluation must focus on incremental economics rather than total promotional revenue.

9. Applied Case: NovaHome Direct (Illustrative Case)

Background

NovaHome Direct is an illustrative D2C retailer selling home-organisation products. Its catalogue contains 4,500 active SKUs. Sales have grown rapidly through paid social advertising, but contribution margin has declined. Customers complain that search results are repetitive, popular products are frequently unavailable and discount codes appear inconsistent.

Problem

Management observes three symptoms: a high overall conversion rate on promoted landing pages, low repeat purchase among first-time customers and rising return costs. The merchandising team recommends adding more products, while finance recommends reducing discounts. Neither recommendation addresses the interaction among assortment, ranking, price and fulfilment.

Analysis

The firm constructs a category-level scorecard. It discovers that 18% of SKUs generate less than 1% of sales, but several of these products improve basket completion and reduce customer switching. It also finds that products with attractive gross margin have high return rates because dimensions and compatibility information are incomplete. Finally, a competitor-monitoring exercise shows that NovaHome is not consistently the cheapest retailer, but its delivery promise is stronger in major cities.

A promotion analysis using customer-level holdouts shows that the existing 15% discount generates substantial sales but only modest incremental contribution. Many loyal customers would have purchased without the discount. A recommendation test shows that “frequently bought together” modules increase basket value when they include complementary products, but generic popularity modules mainly shift clicks among products already visible on the page.

Approach and solution

NovaHome adopts an integrated programme:

• It classifies products as destination, routine, convenience, image or clearance items.
• It removes duplicate and low-information listings, while retaining strategically important niche products.
• It improves product attributes, dimensions, photography and compatibility information.
• It introduces a price ladder rather than uniform discounting.
• It replaces blanket discounts with targeted benefits for genuinely incremental segments.
• It uses inventory-aware recommendations that suppress unavailable products and protect scarce stock.
• It creates a competitor dashboard comparing comparable products, delivery and total checkout cost.
• It establishes a governance committee for pricing and recommendation changes.

Outcome

In the first evaluation period, conversion remains broadly stable, average contribution per order increases and return rates decline. The retailer also observes improved repeat purchase among customers who receive replenishment reminders rather than generic discount messages. The result is not attributed to a single algorithm; it reflects coordinated improvements in content, assortment, price design and customer experience.

Learning points

1. More listings do not necessarily create more value.
2. Gross margin is insufficient without return and fulfilment economics.
3. Promotional sales must be evaluated against a counterfactual.
4. Recommendation systems should optimise incremental value, not clicks alone.
5. Competitive intelligence should include service and comparability, not price alone.
6. Governance is a commercial capability, not merely a compliance requirement.

10. Technology, AI and Governance

Artificial intelligence can support demand forecasting, product classification, image enrichment, price recommendations, fraud detection, assortment optimisation and natural-language search. However, AI outputs are only as reliable as the data, objectives and controls surrounding them.

Key governance requirements include:

• clear ownership of models and decisions;
• documented data sources and feature definitions;
• testing for accuracy, drift and subgroup disparities;
• human review for high-impact pricing or customer outcomes;
• explainability appropriate to the decision;
• privacy-compliant data use and retention;
• cybersecurity and access controls;
• audit trails, version control and rollback procedures.

Retailers should distinguish personalisation from discriminatory treatment. Using legitimate contextual information to improve relevance is not automatically equivalent to charging different customers different prices. Nevertheless, differential treatment based on sensitive or protected characteristics, inferred vulnerability or exploitative timing can create legal, ethical and reputational risks. Transparent policies, impact assessments and legal review are essential.

11. Challenges and Limitations

Several limitations constrain digital merchandising and revenue management:

• Data quality: missing attributes and inconsistent product identifiers undermine analysis.
• Endogeneity: price, ranking and promotion influence the behaviour used to estimate demand.
• Selection bias: observed purchases reflect previous merchandising choices, not pure preference.
• Cold-start problems: new products and new customers have limited behavioural history.
• Short-term optimisation: algorithms may sacrifice loyalty, brand equity or supplier relationships.
• Channel conflict: marketplace prices may undermine D2C positioning or retailer partnerships.
• Operational constraints: forecasts cannot create inventory or delivery capacity.
• Measurement error: attribution systems may over-credit last-click advertising or recommendations.
• Customer trust: opaque experimentation and unstable prices can damage confidence.

Managers should therefore combine quantitative models with qualitative insight, experiment where possible and review outcomes over an appropriate time horizon.

Best-Practice Framework

A high-performing digital merchandising and revenue-management system should:

1. Start with customer missions and category roles.
2. Use contribution, not revenue alone, as a core commercial measure.
3. Maintain a coherent price and value architecture.
4. Estimate promotion incrementality using control or holdout designs.
5. Integrate inventory, delivery and return information into ranking and pricing.
6. Evaluate recommendations on incremental value, diversity and customer outcomes.
7. Use demand sensing as an input to judgement, not an unquestioned decision engine.
8. Compare total customer value, not only headline competitor prices.
9. Apply guardrails, auditability and fairness review to automated decisions.
10. Reassess models as customer behaviour, competition and operating conditions change.

Chapter Summary

Digital merchandising integrates assortment selection, product information, search, ranking, recommendations, pricing and promotion. Effective assortment planning balances breadth and depth with customer missions, substitution, margin, inventory and fulfilment economics. Category management treats categories as strategic units and requires scorecards that combine customer, commercial and operational outcomes.

Pricing decisions should consider elasticity, cost, competition, inventory, brand positioning and customer lifetime value. Dynamic pricing can improve revenue or margin, but requires constraints, monitoring and governance. Promotions should be evaluated using incremental contribution and a credible counterfactual. Recommendation systems can increase relevance and basket value, yet must be assessed for incrementality, diversity, privacy and bias.

Demand sensing updates forecasts using recent behavioural and market signals, while competitive intelligence supports informed—not automatic—responses to rival actions. Revenue management coordinates all these decisions over time. The strongest practice is integrated, experimentally informed and ethically governed.

Key Terms

Assortment: The set of products offered for a defined market, channel, segment and period.

Breadth: The number of categories in an assortment.

Depth: The number of variants offered within a category.

Category management: Managing a product category as a strategic business unit based on customer needs and performance.

Contribution: Sales remaining after specified variable commercial costs.

Price elasticity: The responsiveness of quantity demanded to a change in price.

Dynamic pricing: Adjusting prices in response to changing demand, inventory, competition or context.

Promotion incrementality: The sales or contribution caused by a promotion beyond the expected baseline.

Cannibalisation: The diversion of sales from one product or channel to another within the same business.

Recommendation system: A technology that selects and ranks products likely to be relevant to a customer.

Collaborative filtering: Recommendation based on similarities among customer or product behaviours.

Association rule: A statistical relationship describing the co-occurrence of products or events.

Demand sensing: Rapid updating of demand expectations using recent and high-frequency signals.

Competitive intelligence: Systematic collection and analysis of information about competitors and market conditions.

Revenue management: Coordinating price, demand, inventory and capacity to maximise commercial value.

Price architecture: The structured relationship among product price tiers and value propositions.

Baseline demand: Expected demand in the absence of a specific intervention.

Price index: A ratio comparing an own price with a defined competitor or reference price.

Feedback loop: A process in which an algorithm’s output changes behaviour and subsequently reinforces the algorithm’s future output.

Contribution margin: Contribution expressed as a proportion of sales or as a monetary amount, depending on the stated definition.

Review Questions

Short Answer Questions

1. Distinguish assortment breadth from assortment depth.
2. Why is contribution generally more informative than revenue alone?
3. Define price elasticity of demand and explain its sign in normal circumstances.
4. What is the purpose of a holdout group in promotion evaluation?
5. Identify two risks associated with recommendation feedback loops.

Long Answer Questions

1. Explain how digital category management differs from traditional store-based category management.
2. Discuss the economic and ethical conditions under which dynamic pricing may be appropriate.
3. Compare content-based recommendation, collaborative filtering and association-rule recommendation.
4. Explain how demand sensing can improve merchandising decisions and identify its limitations.
5. Develop an integrated framework for evaluating a digital assortment.

Analytical / Application Questions

1. A retailer sells a product for £80 with variable costs of £52. A promotion reduces the price to £72 and is expected to increase sales from 2,000 to 2,500 units. The campaign costs £3,000. Calculate the incremental promotional contribution and state whether the campaign is attractive under the stated assumptions.
2. An online category has high conversion but low repeat purchase and high returns. Design a diagnostic plan using at least six customer, commercial and operational metrics.
3. A competitor reduces price by 10% for one week while offering slower delivery. Recommend whether a D2C retailer should match the price, and justify the decision using value architecture, contribution, inventory and customer experience considerations.

Multiple Choice Questions with Answers

Question 1

Which measure most directly captures the responsiveness of quantity demanded to a price change?

A. Gross margin

B. Price elasticity of demand

C. Price index

D. Average order value

Correct Answer: B. Price elasticity of demand

Question 2

A product sells for £50 and has variable costs of £32. What is its unit contribution?

A. £18

B. £32

C. £50

D. £82

Correct Answer: A. £18

Question 3

Why is a holdout group useful in promotion evaluation?

A. It guarantees higher sales

B. It estimates what would have happened without the promotion

C. It eliminates all sampling error

D. It measures only customer satisfaction

Correct Answer: B. It estimates what would have happened without the promotion

Question 4

In association-rule analysis, lift above 1 generally indicates that:

A. The products never occur together

B. The consequent is always purchased

C. The products co-occur more than expected from the baseline

D. The rule proves causation

Correct Answer: C. The products co-occur more than expected from the baseline

Question 5

Which category role primarily gives customers a reason to visit a retailer?

A. Destination

B. Clearance

C. Convenience

D. Routine

Correct Answer: A. Destination

Question 6

In the exponential-smoothing equation F_t+1 = αA_t + (1 − α)F_t, a higher α generally makes the forecast:

A. Less responsive to recent demand

B. More responsive to recent demand

C. Independent of actual demand

D. Always equal to zero

Correct Answer: B. More responsive to recent demand

Question 7

Which statement about recommendation click-through rate is most accurate?

A. It always proves incremental profit

B. It is irrelevant to merchandising

C. It can be high without creating incremental orders

D. It measures return-rate reduction directly

Correct Answer: C. It can be high without creating incremental orders

Question 8

A price index of 100 means that:

A. The own price equals the defined reference price

B. The own price is double the reference price

C. The product has no demand

D. The competitor is out of stock

Correct Answer: A. The own price equals the defined reference price

Question 9

Which is a major limitation of using revenue alone to evaluate an assortment?

A. Revenue cannot be recorded digitally

B. Revenue ignores costs such as fulfilment, returns and discounting

C. Revenue is always lower than contribution

D. Revenue measures only inventory age

Correct Answer: B. Revenue ignores costs such as fulfilment, returns and discounting

Question 10

Which governance practice is most appropriate for automated pricing systems?

A. Remove all constraints to maximise speed

B. Avoid retaining decision records

C. Use audit logs, guardrails and rollback procedures

D. Change objectives without monitoring

Correct Answer: C. Use audit logs, guardrails and rollback procedures