Bundling & Product Line Design

When a firm sells multiple products, combining them into a bundle at a single price can capture more surplus than selling each product separately. The key driver is not product synergy but the statistical structure of customer preferences: bundling works best when willingness-to-pay values are negatively correlated across products.

This topic is part of the Nonlinear Pricing & Product Design chapter. See also: Two-Part Tariffs & Volume Discounts and Versioning & Quality Tiers.

Introduction

Bundling is a form of second-degree price discrimination. The firm cannot directly observe each customer’s willingness to pay for individual products, but by offering a bundle at a single price it induces customers to self-select: those whose combined valuations exceed the bundle price buy the bundle, while others may purchase individual products (under mixed bundling) or forgo the purchase entirely (under pure bundling).

The formal analysis of bundling as a profit-maximizing strategy begins with Adams and Yellen (1976), who showed that bundling can increase profit even when individual products have independent costs and demands. The critical insight is that bundling reduces the dispersion of effective valuations across customers, enabling the firm to set a price that extracts a larger fraction of the total surplus.

Three Selling Strategies

Definition — Pure Bundling

Under pure bundling, products are available only as a bundle. A customer who wants just one product must still purchase and pay for the entire package. There is a single price $P_{\text{bundle}}$ and no individual prices.

Definition — Mixed Bundling

Under mixed bundling, the firm offers each product individually and as a bundle. Three prices are set: $P_A$ for product A alone, $P_B$ for product B alone, and $P_{\text{bundle}}$ for both together, typically with $P_{\text{bundle}} < P_A + P_B$ . Each customer chooses whichever option maximizes their surplus.

Definition — Unbundled Selling

Under unbundled (separate) selling, each product is priced independently with no bundle discount. The firm sets $P_A$ and $P_B$ to maximize total revenue across both products without offering a combined option.

A customer with willingness to pay $\text{WTP}_A$ for product A and $\text{WTP}_B$ for product B purchases the bundle if and only if:

\text{WTP}_A + \text{WTP}_B \ge P_{\text{bundle}}

(1)

Under mixed bundling, a customer who prefers not to buy the bundle will additionally compare buying A only (if $\text{WTP}_A \ge P_A$ ) or B only (if $\text{WTP}_B \ge P_B$ ) against buying nothing.

When Bundling Helps

Consider two products and two customers. Customer 1 has $\text{WTP}_A = 80$ and $\text{WTP}_B = 20$ . Customer 2 has $\text{WTP}_A = 30$ and $\text{WTP}_B = 70$ . Assume zero production costs.

Separate selling at low prices ( $P_A = 30, P_B = 20$ ): both customers buy both products. Revenue $= 2(30) + 2(20) = \$100$ .
Separate selling at high prices ( $P_A = 80, P_B = 70$ ): each customer buys only the product they value most. Revenue $= 80 + 70 = \$150$ .
Pure bundle at $P_{\text{bundle}} = 100$ : both customers have total WTP of exactly 100, so both purchase. Revenue $= 2 \times 100 = \$200$ .

Bundling achieves $200 because the negative correlation in WTPs makes total valuations uniform: even though individual valuations are dispersed (ranging from 20 to 80), the sums are identical at 100.

This example illustrates the central mechanism. The individual WTP values are highly dispersed, making it impossible to set any single price that captures both customers for either product. But the sums of WTPs are identical, so a single bundle price captures the entire market. Bundling has transformed a heterogeneous market into a homogeneous one for the purpose of pricing.

Adams-Yellen Theory

Adams and Yellen (1976) provided the first systematic analysis of bundling as a profit-maximizing strategy. Their framework considers a 2D WTP space, with $\text{WTP}_A$ on the horizontal axis and $\text{WTP}_B$ on the vertical axis. Each customer is a point in this space. The three pricing strategies partition the space into regions determining which customers buy what.

Under mixed bundling with prices $P_A$ , $P_B$ , and $P_{\text{bundle}}$ , the WTP space divides into five regions:

Buy bundle: customers with $\text{WTP}_A + \text{WTP}_B \ge P_{\text{bundle}}$ and for whom the bundle provides more surplus than buying either product alone.
Buy A only: customers with $\text{WTP}_A \ge P_A$ but $\text{WTP}_B < P_B$ and the bundle does not provide additional surplus.
Buy B only: the symmetric case.
Buy both separately: customers with $\text{WTP}_A \ge P_A$ and $\text{WTP}_B \ge P_B$ but $\text{WTP}_A + \text{WTP}_B < P_{\text{bundle}}$ (the bundle is not worth the premium over buying separately).
Buy nothing: all remaining customers whose valuations are too low.

The key result of Adams and Yellen (1976) is that the optimal strategy depends on the joint distribution of WTP values. When WTPs are negatively correlated, the bundle price can be set high enough to capture most customers (because few have very low combined valuations), while the individual prices capture customers at the extremes of the preference distribution.

Bundling and WTP Variance

Let $V = \text{WTP}_A + \text{WTP}_B$ denote total valuation. If $\text{WTP}_A$ and $\text{WTP}_B$ have equal marginal distributions with variance $\sigma^2$ and correlation $\rho$ , then the variance of total valuation is:

\text{Var}(V) = 2\sigma^2(1 + \rho)

When $\rho = -1$ , total valuation variance is zero: all customers have identical combined WTP, enabling the firm to perfectly extract the full market surplus with a single bundle price. When $\rho = +1$ , total valuation variance doubles, and bundling offers no statistical advantage over separate selling.

This variance-reduction property is the statistical heart of bundling. Negative correlation converts a pricing problem with heterogeneous customers into one with near-homogeneous total valuations, making it much easier to extract surplus through a single bundle price.

Mixed Bundling Dominance

Mixed Bundling Weakly Dominates (Schmalensee 1984)

Under mild regularity conditions, mixed bundling weakly dominates both pure bundling and separate selling. That is, the maximum profit attainable under mixed bundling is at least as high as under either of the other two strategies.

The intuition for this dominance result is straightforward. Mixed bundling can always replicate pure bundling by setting $P_A$ and $P_B$ prohibitively high, so no one buys individual products. It can replicate separate selling by setting $P_{\text{bundle}} = P_A + P_B$ , eliminating the bundle discount. Since mixed bundling encompasses both strategies as special cases, it cannot do worse than either.

McAfee, McMillan, and Whinston (1989) extended this analysis to show that when customer valuations are independently distributed, pure bundling is never strictly optimal. Mixed bundling always weakly dominates, with strict dominance when the distribution of valuations has sufficient probability mass near the individual prices. Their result holds for a wide class of continuous distributions and does not require the negative correlation that makes bundling most attractive.

For large product lines, Hanson and Martin (1990) showed that optimal bundle pricing with many products can be formulated and solved as an integer programming problem, providing a computational approach when the combinatorial space of possible bundles is large.

Interactive Explorer

Each dot in the chart below represents a customer with willingness to pay for products A and B. The vertical dashed line marks $P_A$ , the horizontal line marks $P_B$ , and the diagonal marks $\text{WTP}_A + \text{WTP}_B = P_{\text{bundle}}$ . Customers are colored by their purchase decision under the current prices.

Adjust the WTP correlation slider from positive to negative and observe how the profit advantage of bundling grows as correlation decreases. With strong negative correlation, most customers have nearly identical total WTP, allowing the bundle price to capture almost the entire market. Compare total revenues under pure bundling, mixed bundling, and separate selling as you move the sliders.

Extensions & Applications

Personalized Bundling

With sufficient data on customer preferences, firms can offer individualized bundle recommendations. Recommendation systems that suggest complementary products exploit the same variance-reduction logic: the system identifies customers whose valuations for a set of products are negatively correlated and targets bundle offers accordingly. This bridges second-degree discrimination (a uniform menu) and first-degree discrimination (individualized pricing).

Tying and Leverage

Tying is a form of pure bundling in which the purchase of one product (the tying good) is contingent on purchasing a second product (the tied good). Unlike voluntary bundling, tying may have anticompetitive effects if the tying product is a monopoly good and the tied good has competitive alternatives. The welfare analysis of tying is therefore distinct from the welfare analysis of voluntary bundling.

Subscription Bundles

Streaming services (music, video, software) often bundle many goods into a single subscription. With a large number of independent goods, the law of large numbers implies that total WTP has low variance even when individual WTPs are dispersed. This is the basis for the subscription pricing model: at scale, a fixed monthly fee can extract a high fraction of total customer surplus without requiring fine-grained knowledge of individual preferences. For more on subscription design, see the Subscription Pricing topic.

Versioning as a Dimension of Bundling

When products differ in quality rather than kind, the bundling framework extends naturally to quality-tier design. A firm offering basic and premium versions of the same product is effectively designing a menu of quality–price pairs. The optimal menu satisfies incentive-compatibility constraints analogous to those governing the optimal bundle. See Versioning & Quality Tiers for a full treatment.

Capstone: Bundling Lab

In Play mode, you set prices for two products and their bundle, then watch a stream of heterogeneous consumers decide what to buy. Compare separate selling, pure bundling, and mixed bundling strategies head-to-head. In Design mode, explore how WTP correlation, market size, and cost structure determine when bundling dominates.

References

Adams, W. J. & Yellen, J. L. (1976). “Commodity Bundling and the Burden of Monopoly.” Quarterly Journal of Economics, 90(3), 475–498.
Hanson, W. & Martin, R. K. (1990). “Optimal Bundle Pricing.” Management Science, 36(2), 155–174.
McAfee, R. P., McMillan, J. & Whinston, M. D. (1989). “Multiproduct Monopoly, Commodity Bundling, and Correlation of Values.” Quarterly Journal of Economics, 104(2), 371–383.
Phillips, R. L. (2021). Pricing and Revenue Optimization, 2nd ed.. Stanford University Press.
Schmalensee, R. (1984). “Gaussian Demand and Commodity Bundling.” Journal of Business, 57(1), S211–S230.
Wilson, R. (1993). Nonlinear Pricing. Oxford University Press.