The Economics of Zero Marginal Cost Bundling: When Adding Products Decreases Revenue

TL;DR: In digital markets where the marginal cost of adding a product to a bundle is zero, conventional wisdom says bundle everything -- but past a threshold, each addition dilutes the bundle's perceived value and total willingness to pay drops. The companies capturing disproportionate value subtract instead of add, optimizing for value density (value per item) rather than bundle size, because attention scarcity trumps production abundance.

---

The Counterintuitive Truth About Free Addition

Here is something that should bother every product manager building a digital product: adding a feature that costs nothing to produce can destroy revenue.

Not reduce margins. Destroy revenue.

We have been trained to think in physical-goods logic, where mental accounting categories align neatly with production costs. Adding a widget to a box costs raw materials, labor, shipping weight. There is a natural brake on bundling — cost. But in digital markets, that brake is gone. The marginal cost of including one more app in a software suite, one more show in a streaming catalog, one more song in a subscription is functionally zero.

So we add everything. And then we wonder why conversion drops, why willingness to pay compresses, why the market treats our ever-expanding bundle like a commodity.

The economics of zero marginal cost bundling are not the economics of abundance. They are the economics of attention scarcity dressed in abundance clothing. And the companies that understand this distinction — the ones that subtract instead of add — are capturing disproportionate value.

This article builds a framework for understanding when bundling creates value, when it destroys value, and how to find the inflection point between the two. We draw on three decades of bundling theory, from Adams and Yellen's original taxonomy to Bakos and Brynjolfsson's information goods model to recent experimental work by Brough and Chernev on the dilution paradox. We present original simulation data. And we propose a metric — value density — that we believe should replace bundle size as the primary design variable.

---

A Brief Taxonomy of Bundling

Adams and Yellen established the canonical framework in 1976, dividing bundling into three forms: pure components (everything sold separately), pure bundling (everything sold together, no individual purchase option), and mixed bundling (both options available). Their central finding, which has held up across five decades of subsequent research, is that mixed bundling weakly dominates both pure strategies under most demand conditions.

The reason mixed bundling dominates is mathematical, not intuitive. Consider two consumers: Consumer A values Product X at $100 and Product Y at$20. Consumer B values Product X at $20 and Product Y at$100. Under pure components priced at $100 each, you sell one unit of each —$200 total. Under pure bundling at $120, you sell two bundles —$240 total. Under mixed bundling at $100 individual /$120 bundle, you sell two bundles — still $240, but with the added insurance that a third consumer who values only Product X at$100 can still purchase.

This logic scales. Bundling works because it reduces variance in willingness to pay across a heterogeneous consumer base. The more products in the bundle, the tighter the distribution of total valuations becomes. This is the law of large numbers applied to pricing.

But there is a catch. And the catch is everything.

---

The Bakos-Brynjolfsson Theorem and Its Limits

Yannis Bakos and Erik Brynjolfsson published their landmark paper in 1999, extending Adams and Yellen's framework to information goods — products with near-zero marginal cost. Their core theorem is elegant: as the number of goods in a bundle increases, the distribution of consumer valuations for the bundle converges, allowing the seller to extract a larger share of total surplus.

In formal terms, if individual valuations are independently distributed with mean $\mu$ and variance $\sigma^2$, then the total willingness to pay for a bundle of $n$ items is:

$$WTP_{\text{bundle}}(n) = \sum_{i=1}^{n} v_i, \quad \text{where } v_i \sim \mathcal{D}(\mu, \sigma^2)$$

The per-good variance in a bundle of $n$ items is $\sigma^2 / n$. As $n$ grows, the distribution of per-good bundle valuations concentrates around $\mu$. Dead-weight loss shrinks. The seller captures more.

This is a beautiful result. It is also dangerously incomplete.

The model assumes three things that rarely hold in practice:

First, it assumes additive valuations — that a consumer's willingness to pay for a bundle equals the sum of their willingness to pay for each component. Behavioral research consistently shows this is false. Consumers discount items they do not want. Worse, they sometimes subtract value.

Second, it assumes independent valuations — that how much you value Product X does not affect how much you value Product Y. In reality, valuations are often negatively correlated within bundles (a consumer who wants a professional video editor probably does not also want a children's drawing app, yet both might appear in a "creative suite").

Third, it assumes consumers can evaluate all items in the bundle. As bundle size grows, evaluation becomes impossible. Consumers fall back on heuristics. And those heuristics almost never favor the seller.

The gap between theoretical and observed revenue is not noise. It is the dilution paradox at work.

---

The Dilution Paradox

Aaron Brough and Alexander Chernev published a series of experiments between 2012 and 2016 that should have reshaped how every digital company thinks about bundling. Their finding: adding low-value items to a bundle reduces the perceived value of the entire bundle, even when the added items have positive standalone value.

Read that again. A product with positive value, added to a bundle at zero cost, reduces what consumers will pay for the bundle.

This is not rational in the classical sense. But it is consistent and replicable. Brough and Chernev call it the "presenter's paradox" — the intuition that more is better conflicts with how evaluators actually form judgments.

The mechanism is averaging. When consumers evaluate a bundle, they do not sum the values of individual components. They form an overall impression that functions more like a weighted average. Adding a $5 item to a bundle of$50 items drags the average impression downward. The $5 item signals that the bundle contains filler. And filler contaminates everything around it.

Consider a concrete scenario. A SaaS company sells three products individually: a project management tool ($30/mo), a document editor ($25/mo), and a spreadsheet app ($25/mo). Customers evaluate the bundle at roughly$65/mo — a slight discount from the $80 component sum, consistent with standard bundling theory.

Now the company adds a basic note-taking app (standalone value: $3/mo) and a simple timer widget ($2/mo). The five-product bundle should be worth at least $65/mo — the added items have positive value and cost nothing to include. Instead, experimental data from Brough and Chernev's framework predicts willingness to pay drops to approximately$55/mo. The note-taking app and timer have reframed the bundle from "three professional tools" to "a mixed bag of stuff."

The pattern is clear. The first three additions increase total willingness to pay. The fourth begins destroying it. By the seventh product, the bundle is worth less than the original two-product combination.

This is not a theoretical curiosity. This is the revenue curve of every digital platform that has ever asked the question "why not add one more thing?"

---

Value Density: A New Framework

We propose a metric that captures the dilution effect and gives product teams a concrete design target. We call it value density: the ratio of perceived bundle value to the number of items in the bundle.

$$V_d = \frac{WTP(n)}{n}$$

where $V_d$ is value density, $WTP(n)$ is the perceived willingness to pay for a bundle of $n$ components. The consumer surplus of the bundle can then be expressed as:

$$CS = WTP(n) - P_{\text{bundle}} = n \cdot V_d - P_{\text{bundle}}$$

Value density is not willingness to pay. It is not revenue. It is the per-item signal strength of the bundle — a measure of how concentrated the value proposition feels to the buyer.

The critical insight: revenue is maximized not at peak WTP, but near peak value density. This is because value density correlates with conversion rate. A bundle with high value density feels like a good deal. A bundle with low value density feels like bloatware, regardless of its absolute price.

The optimal bundle, in this simulation, is three items. Not because three is a magic number, but because the third item is the last one that increases total WTP without catastrophically reducing value density. The fourth item crosses the threshold — it adds marginal perceived value below the bundle's current average, dragging everything down.

We formalize this as the Value Density Threshold Test:

Add item item (k+1) to a bundle of k items only if v(k+1) > WTP(k)/k, where v(k+1) is the standalone perceived value of the new item and WTP(k)/k is the current value density.

This rule is conservative. It will sometimes reject items that would increase total WTP slightly. But it protects against the far more common and costly error: adding items that feel free to produce but cost real revenue through dilution.

---

Case Study: Microsoft Office vs Google Workspace

Microsoft and Google have pursued opposite bundling strategies for the same market, and the revenue outcomes illuminate the dilution paradox with unusual clarity.

Microsoft's approach: curated density. Microsoft 365 Personal includes six core applications: Word, Excel, PowerPoint, Outlook, OneNote, and OneDrive. The company has resisted adding every product it makes to the core bundle, a strategy consistent with how switching costs through deep integration create more durable lock-in than breadth alone. Teams was added in 2020, but notably removed from default EU bundles in 2023 following antitrust pressure — and Microsoft did not fight particularly hard to keep it bundled. Why? Internal data almost certainly showed what the dilution model predicts: Teams' inclusion was compressing per-seat willingness to pay in segments that did not want a communication tool.

Google's approach: everything included. Google Workspace bundles Gmail, Drive, Docs, Sheets, Slides, Meet, Chat, Forms, Sites, Keep, Currents, Apps Script, Calendar, and more. The bundle has grown steadily since its launch as "Google Apps for Your Domain" in 2006.

The pricing tells the story.

Google offers roughly twice as many applications for a similar or lower price. Microsoft offers fewer applications and commands higher willingness to pay at the enterprise tier — $36 versus$25 per seat per month. The company with the smaller bundle extracts more revenue per user.

This is not because Microsoft's individual products are necessarily better. It is because the bundle signal is different. Six carefully selected professional tools communicate a different value proposition than fourteen loosely related tools plus whatever Google shipped last quarter.

Microsoft's bundle says: "These are the tools professionals need." Google's bundle says: "Here is everything we make." The first framing commands a premium. The second invites commodity pricing.

Google seems to understand this partially — their tiered approach essentially uses feature gating to create sub-bundles within the mega-bundle. But the top-level positioning still suffers from the density problem. When a CIO evaluates the two offerings, Microsoft's pitch is tighter.

---

Netflix and the Content Paradox

Netflix spent $17 billion on content in 2024. Its library contains over 17,000 titles. And its biggest retention problem is not a lack of content — it is the perception that there is nothing to watch.

This is the dilution paradox at platform scale -- and it illustrates the cannibalization dynamics that platforms face when first-party additions compete with existing quality signals.

The mechanism works through two channels. First, choice overload — Barry Schwartz's "paradox of choice" applied to content catalogs. When faced with 17,000 options, consumers spend more time browsing and less time watching. The browsing experience itself becomes a source of dissatisfaction.

Second, and more relevant to our framework: the mere presence of low-quality content reduces perceived catalog value. Netflix's library includes prestige dramas, award-winning documentaries, and also a vast quantity of low-budget filler produced to hit content quotas. A consumer who encounters three mediocre titles while browsing revises their estimate of the entire catalog downward. The filler is not neutral. It is actively corrosive.

The data shows a consistent inverse relationship. As Netflix's library grew from 2020 to 2024, subscriber satisfaction declined. The slight uptick in 2025 coincides with Netflix's shift toward fewer, higher-budget productions — a tacit acknowledgment that more is not better.

HBO's counter-strategy is instructive. Max (formerly HBO Max) has always maintained a smaller, more curated library. Its content spend is a fraction of Netflix's. Yet its brand carries a quality premium that translates directly into pricing power. Consumers will pay more per title for a curated selection than for an exhaustive one.

This is value density applied to content. HBO maximizes perceived value per title. Netflix has historically maximized total title count. The market rewards density.

---

When Unbundling Creates More Value

The music industry's transition from albums to singles is the most dramatic unbundling event in economic history. And it proves that the optimal bundle size can be one.

In 2002, the average consumer bought 2.1 CDs per year at roughly $15 each — about$31.50 in annual music spending. Albums were bundles: 12-15 tracks, of which the typical buyer wanted 2-3. The value density of an album was terrible. Consumers were paying $15 for a bundle they valued at$3-4.

iTunes launched in 2003 with $0.99 singles. By 2006, digital single sales had exploded. The industry panicked about falling revenue, but something more interesting was happening at the unit economics level: per-track revenue was increasing. Consumers who previously bought zero albums because the bundle price exceeded their willingness to pay were now buying individual tracks. The addressable market expanded.

The industry's total revenue did fall — from $14.6 billion in 1999 to$6.9 billion in 2014. But this decline was driven primarily by piracy and format transition friction, not by unbundling per se. The subsequent recovery, driven by streaming (a rebundling, which we address next), reached $17.1 billion by 2023.

The lesson is not that unbundling always wins. The lesson is that unbundling wins when the existing bundle has low value density — when consumers are forced to buy items they do not want to get items they do. Physical distribution costs forced the album format. When digital distribution removed that constraint, the market instantly repriced to reflect actual per-item valuations.

This principle generalizes. Cable television is undergoing the same unbundling. The average cable bundle contains 200+ channels; the average household watches 17. Value density is approximately 8.5%. Streaming services are unbundled cable channels — and despite widespread complaints about "too many subscriptions," consumers are spending more on video entertainment in total than they did on cable. They are paying more because each subscription has higher value density than the cable bundle did.

The pattern: industries with value density below ~20% are unstable bundles. They will unbundle as soon as distribution costs fall far enough to make per-item sales viable. Industries between 20-40% are contested — some consumers prefer the bundle, others prefer components. Above 40%, bundles are stable.

---

Micro-Bundling: Spotify's Third Path

Spotify has discovered something that neither pure bundling nor pure unbundling captures: micro-bundling. And it may be the dominant strategy for digital goods going forward.

Spotify's catalog contains over 100 million tracks. No consumer wants all of them. Most consumers want a few hundred to a few thousand. The full catalog, as a bundle, has extremely low value density — you are "paying" for 100 million tracks and listening to 500.

But Spotify does not sell the catalog as a monolithic bundle. It sells access to the catalog, and then micro-bundles within it through playlists. A playlist is a curated sub-bundle: 30-80 tracks selected for a specific mood, activity, or taste profile. The value density of a well-curated playlist is extremely high — most tracks are relevant to the listener.

This is the insight: the unit of bundling does not have to be the unit of pricing. Spotify prices at the catalog level ($10.99/mo for everything) but bundles at the playlist level. The consumer's daily experience is not "I have access to 100 million songs" — it is "I have a playlist of 50 songs perfect for my morning run." The value density of the experienced bundle is high, even though the value density of the priced bundle is low.

This decoupling of pricing unit and experience unit is Spotify's structural advantage. It allows the company to offer a massive catalog (important for long-tail retention) without suffering the dilution paradox (because consumers never evaluate the full catalog as a bundle).

Apple Music and YouTube Music have copied the playlist mechanic, but Spotify's head start in algorithmic curation — Discover Weekly, Daily Mix, Release Radar — gives it a structural edge. The quality of micro-bundling depends on the quality of curation. And curation quality depends on data volume, which depends on user base, which depends on curation quality. This is the flywheel that makes micro-bundling defensible.

The micro-bundling framework extends beyond music. Netflix's "Top 10" and "Because You Watched" rows are micro-bundles — curated subsets of the full catalog designed to present high value density to each individual user. TikTok's For You page is the most aggressive micro-bundling system ever built: every scroll presents a single-item micro-bundle, curated in real-time. The value density of one algorithmically selected video is as high as it can possibly be.

---

Revenue Simulation: Finding the Optimal Bundle Size

We built a Monte Carlo simulation to model the interaction between classical bundling theory (the Bakos-Brynjolfsson surplus extraction effect) and the dilution paradox (the Brough-Chernev averaging effect). The simulation assumes:

• 10,000 simulated consumers with log-normally distributed valuations
• Individual item valuations drawn from LogNormal(mean=$10, sd=$5)
• A dilution discount factor that increases with bundle size: $d(n) = 1 - 0.03 \times (n - n^*)$ for $n > n^*$, where $n^*$ is the dilution threshold
• Mixed bundling with individual purchase option at full price
• Price set at the 40th percentile of the WTP distribution (a common heuristic)

We ran the simulation across bundle sizes from 1 to 30 items, with dilution thresholds ($n^*$) varying from 2 to 10.

Three findings emerge from the simulation:

First, optimal bundle size is always finite. Even with a generous dilution threshold of 8, revenue peaks at 8 items and declines thereafter. There is no scenario in which adding products indefinitely increases revenue.

Second, the dilution threshold determines optimal bundle size almost exactly. When the threshold is 3, optimal bundle size is 3. When the threshold is 5, it is 5. When the threshold is 8, it is 8. This makes intuitive sense: you should add items until you hit the dilution boundary, then stop.

Third, the revenue loss from over-bundling is larger than the revenue loss from under-bundling. A bundle of 20 items with a dilution threshold of 5 captures only 25% of peak revenue. A bundle of 3 items captures 79% of peak revenue. When in doubt, bundle less.

---

Implementation Guide: Optimal Bundle Size by Category

The dilution threshold varies by product category, consumer sophistication, and evaluation context. Based on our synthesis of the academic literature and simulation results, we propose the following guidelines:

Five principles for implementation:

Principle 1: Measure value density, not just WTP. Track perceived per-item value through conjoint analysis or Van Westendorp pricing surveys. If adding an item to the bundle reduces per-item perceived value, do not add it regardless of its standalone merit.

Principle 2: Name the bundle after its strongest item. Brough and Chernev's research shows that the bundle's perceived value is heavily anchored to the most prominent item. "Microsoft Word and friends" commands higher WTP than "Microsoft Office Suite" even though they describe the same product. The anchor item should be the one with the highest standalone brand recognition.

Principle 3: Use tiers to create multiple value density peaks. Instead of one large bundle, create 2-3 tiers where each tier maintains high value density. Adobe does this well: Photography Plan (Lightroom + Photoshop), Single App plan, and All Apps plan. The Photography Plan has extraordinary value density — two flagship products at $9.99/month. It is Adobe's most popular subscription tier precisely because it feels like a steal.

Principle 4: Hide the denominator. Consumers can only apply the averaging heuristic when they know how many items are in the bundle. Spotify does not tell you there are 100 million tracks. It shows you a playlist. Apple does not list all 6 apps on the Microsoft 365 landing page — it leads with Word document creation, the use case with the highest perceived value.

Principle 5: Unbundle before you rebundle. If your current bundle has low value density, do not try to fix it by adding a high-value item. Instead, remove the low-value items first. Subtraction improves value density faster than addition. A five-item bundle that drops to three items increases value density by 67% even if total WTP decreases by only 10%.

---

The Zero Marginal Cost Trap

We began with a provocation: adding a product that costs nothing to produce can destroy revenue. We now have the framework to understand why, when, and by how much.

The zero marginal cost of digital goods is a trap for product teams trained in physical-goods logic. In physical markets, cost disciplines bundling. You cannot add infinite items to a box because each item has weight, material cost, and shipping implications. The marginal cost curve does the work of curation for you.

In digital markets, there is no such discipline. The marginal cost of adding one more feature, one more show, one more track is zero. The only constraint is attention — and attention is not on the balance sheet. It does not appear in the P&L. No CFO has ever flagged "excessive attention cost imposed on consumers" as a budget line item.

But the market prices it in. Consumers with finite attention and averaging heuristics will pay less for a bloated bundle than a curated one. The revenue impact is real, measurable, and in most cases, larger than the revenue from the marginal products being added.

The companies that understand this — Apple with its deliberate product count discipline, Adobe with its Photography Plan, Spotify with its micro-bundling architecture — are not leaving money on the table by offering less. They are capturing more money by maintaining value density.

The question is not "can we add this to the bundle?" In digital markets, you can always add more. The question is "should we?" And the answer, more often than our instincts suggest, is no.

---

---

Further Reading

• Bundling on Wikipedia — Pricing strategy overview
• Netflix — The content bundling paradox
• Spotify — Micro-bundling through playlists

References

1. Adams, W. J., & Yellen, J. L. (1976). Commodity Bundling and the Burden of Monopoly. The Quarterly Journal of Economics, 90(3), 475-498.

2. Bakos, Y., & Brynjolfsson, E. (1999). Bundling Information Goods: Pricing, Profits, and Efficiency. Management Science, 45(12), 1613-1630.

3. Brough, A. R., & Chernev, A. (2012). When Opposites Detract: Categorical Reasoning and Subtractive Valuations of Product Combinations. Journal of Consumer Research, 39(2), 399-414.

4. Chernev, A. (2011). The Dieter's Paradox. Journal of Consumer Psychology, 21(2), 178-183.

5. Rifkin, J. (2014). The Zero Marginal Cost Society: The Internet of Things, the Collaborative Commons, and the Eclipse of Capitalism. Palgrave Macmillan.

6. Schwartz, B. (2004). The Paradox of Choice: Why More Is Less. Ecco.

7. Schmalensee, R. (1984). Gaussian Demand and Commodity Bundling. The Journal of Business, 57(1), S211-S230.

8. Stigler, G. J. (1963). United States v. Loew's Inc.: A Note on Block-Booking. The Supreme Court Review, 1963, 152-157.

9. Stremersch, S., & Tellis, G. J. (2002). Strategic Bundling of Products and Prices: A New Synthesis for Marketing. Journal of Marketing, 66(1), 55-72.

10. Venkatesh, R., & Mahajan, V. (2009). The Design and Pricing of Bundles: A Review of Normative Guidelines and Practical Approaches. In Handbook of Pricing Research in Marketing, 232-257.

11. McAfee, R. P., McMillan, J., & Whinston, M. D. (1989). Multiproduct Monopoly, Commodity Bundling, and Correlation of Values. The Quarterly Journal of Economics, 104(2), 371-383.