How bundling benefits sellers and buyers

There is a widespread belief in technology circles that bundling of cable TV, newspaper, magazine and other information goods will go away now that those products can be distributed à la carte on the internet. The assumption seems to be that bundling is an artifact of another era when distribution was physical. But this reasoning misses the economic logic behind bundling: under assumptions that apply to most information-based businesses, bundling benefits buyers and sellers.

Consider the following simple model for the willingness-to-pay of two cable buyers, the “sports lover” and the “history lover”:

What price should the cable companies charge to maximize revenues? Note that optimal prices are always somewhere below the buyers’ willingness-to-pay. Otherwise the buyer wouldn’t benefit from the purchase. For simplicity, assume prices are set 10% lower than willingness-to-pay. If ESPN and the History Channel were sold individually, the revenue maximizing price would be \$9 (\$10 with a 10% discount). Sports lovers would buy ESPN and history lovers would buy the History Channel. The cable company would get \$18 in revenue.

By bundling channels, the cable company can charge each customer \$11.70 (\$13 discounted 10%) for the bundle, yielding combined revenue of \$23.40. The consumer surplus would be \$2 in the non-bundle and \$2.60 in the bundle. Thus both buyers and sellers benefit from bundling.

This model is obviously dramatically oversimplified. In real life, bundling tends to flatten the demand curve (here is some background on demand curves, and here is academic paper that presents this argument in rigorous mathematical terms). Suppose the demand curves for ESPN and the History Channel look like this:

The green boxes represent revenue for the seller. The deadweight loss areas to the right of the green boxes are transactions that would have benefited buyers and sellers but are not occurring because the revenue-maximizing prices are set too high.

Now consider what happens when you bundle channels. The key assumption is that individual buyers lie on different x-axis points of the demand curves of different channels. Sports lovers lie on the left of the ESPN demand curve but on the right side of the History Channel curve. To aggregate demand curves, you don’t stack one on top of the other. You add consumers’ willingness-to-pay separately for each channel.

Using the above simplified model, the two demand curves that go from \$10 to \$3 become one curve that stays flat at \$13. In general, adding the individual demand curves creates a flatter demand curve:

A flatter demand curve lets sellers charge prices that capture larger areas under the curve and pass more surplus back to consumers. The only  loser is the deadweight loss area.

Some things to note about bundled pricing:

1. Bundled pricing is one reason why subscription models like Spotify should ultimately win out over à la carte models like iTunes. Subscription commerce can also be thought of as a form of bundling.

2. There are other ways to get some of the benefits of bundled pricing – for example versioning goods, and offering bulk discounts.

3. The benefits of bundled pricing are proportionate the buyers’ variance of preferences for the goods. Hence bundled pricing works best in highly “taste-based” goods like media, and wouldn’t have any benefit for fully commoditized goods (e.g. a bundle of stocks)

4. Bundled pricing can also hurt consumers if it is used by incumbents to exploit their broader catalog to “deter entry” by new competitors. This was a common complaint against Microsoft in the 90′s when they bundled applications like Internet Explorer with Windows.

Pricing to the demand curve

Many college microeconomics courses include the following exercise. The teacher offers the students an imaginary trip to Hawaii, and asks them to write down on notecards how much they are willing to pay for the trip. The teacher takes the notecards and graphs the bids. Here’s how the graph might look:

The y-axis is the students’ “willingness-to-pay” and the x-axis is the students sorted from highest to lowest bids. The line is known as the demand curve.

Now imagine you’re the company selling these trips. For simplicity, suppose you’ve already bought the trips, so your marginal cost is zero. What’s the optimal price you should charge? If you set the price at, say, \$500, then the students who are willing to pay above \$500 would buy the trip, and the rest wouldn’t:

Your total revenue and (assuming zero marginal cost) profit will be the area of the green square (revenue times quantity).

Notice the sections under the curve to the right and above the green box. To the right are students who were willing to pay but were priced out. Those are missed sales opportunities. Above the green box are students who were willing to pay more than you charged. That is lost revenue. (Since the underpricing benefits customers, the area above the green box is called the consumer surplus).

After you have chosen the right price, the only way to make the area under the curve greener is to charge different customers different prices. The theoretically optimal way to do this is to look at each notecard and offer to charge each student, say, 10% less than the prices he or she bid. In real life you can’t do this (although Priceline has gotten close by asking customers to enter their willingness to pay). Some companies – most famously Amazon – have attempted outright price discrimination, but this tends to anger customers and can even run afoul of the law.

So the goal of pricing is to capture as much area under the demand curve as possible. In practice, the best way to do this is to find proxies for willingness-to-pay that are easy to observe and that customers will accept.

For example, airlines know that business customers will pay more than vacation travelers. They therefore look for acceptable proxies to segment business and vacation travelers and capture more of the area under the demand curve.

This is why flights are cheaper when you book early, stay over on weekends etc. The airline pricing models assume you are a vacation traveller.

Book publishers would like to price their books according to customer enthusiasm. Hardcore fans will pay more for books when they are first published, and casual readers will wait. If publishers offer the same book at different prices at different times, their price discrimination will be too obvious (interestingly, time windowing for movies doesn’t provoke much outrage). So book publishers offer modestly better goods – hardcovers – to early buyers.

Enterprise software companies price using proxies for the customer’s budget. Oracle databases are priced by the number of processors. Salesforce is priced by the number of end users (“seats”). Many enterprise software companies obfuscate the highest tier of pricing, telling sales prospects at that level to “call us.” What this really means is: “Call us, so our sales people can attempt to estimate your budget and price discriminate accordingly.”

Sometimes, the search for pricing proxies can lead to absurdity. I once heard someone from a prominent hardware company tell a story about how his company had offered two versions of a printer. The cheaper model was identical to the more expensive one, except the cheaper one printed fewer pages per minute. To accomplish this, the cheaper printer had the same hardware as the expensive one, except the cheaper one had an additional chip that forced it to slow down. This made the cheaper printer more expensive to produce. Situations where cost and price have zero or negative correlation are far more common than most people assume.

Equity value

Warren Buffet once said:

Buy into a business that’s doing so well an idiot could run it, because sooner or later, one will.

This is a useful way to understand the meaning of “equity value”. You learn in finance that equity value is the overall value of a the stock (i.e. equity) of a business, which in turn is the present value of all future profits. Of course with startups the future is extremely uncertain, leading to a huge variance in valuations.

In perfectly competitive markets, all profit margins tend toward zero. So equity value is a function of the degree to which you can make your market inefficient by making your business hard to copy (so called “defensibility”). If your defensibility depends solely on having superior people, you have what VCs call a “service business.” In a competitve labor market, service businesses tend to have low margins and therefore low equity value. A popular saying about service businesses is “the equity value walks out of the building every night.”

Different types of tech businesses exhibit different relationships between capital, revenue, profits, and equity value. Enterprise software companies tend to require lots of capital to get to scale but command high equity values once they do, partly because enterprises are risk averse and like to adopt the most popular technology, leading to winner-take-all dynamics. Adtech companies tend to be quick to revenue but slower to equity value, and sometimes risk becoming service businesses. The equity value of consumer internet companies vary widely, depending on their defensibility (usually networks effects and brand) and business models (e.g. transactional vs ad supported). Biotech companies require boatloads of capital for R&D and regulatory approval but then can generate lots of equity value, with the defensibility coming primarily from patents. (Patents introduce market innefficiencies, but, proponents argue, are necessary to create sufficient incentives for entrepreneurs and investors). E-commerce companies generally require a lot of capital as well, since their defensibility comes mostly through brand and economies of scale.

Some thoughts on when to raise money, and the current financing environment

A key question for founders is when they should try to raise money. More specifically, they often wonder whether to raise money now or wait, say, 6 months when their startup has made more progress. Here are some thoughts on this question generally along with some thoughts on today’s venture financing market.

– In the private markets, macro tends to dominate micro. Venture valuations have swung by roughly a factor of 4 over the last decade. In finance speak, venture tends to be high beta, moving as a multiple of the public markets, which themselves tend to move more dramatically than economic fundamentals. Hence, it is easy to imagine scenarios where the same private company will command 1/2 the valuation in 6 months due to macro events, but it’s rare for a company to increase their valuation 2x through operations alone in 6 months.

– Therefore, when it seems to be the top of a venture cycle, it’s almost always better to raise money sooner rather than later, unless you have a plausible story about how waiting will dramatically improve your company’s fundamentals.

– Prior to the Facebook IPO, the consensus seemed to be that private valuations were near the top of the cycle. Today, FB is valued at up to 50% below what private investors expected. Moreover, the financial crisis in Europe seems to have worsened, and unemployment numbers in the US suggest the possibility of a double dip recession.

– It takes many months to understand how macroeconomic and public market shifts affect private company valuations since (with the exception of secondary markets) private transactions happen slowly. So we don’t know yet what these recent events mean for private markets. According to a basic rule of finance, however, it is safe to assume that companies “comparable” to Facebook are worth up to 50% less than private investors thought they were worth a few weeks ago.

– The question then is what companies are comparable to Facebook. Clearly, other social media companies with business models that rely on display or feed based advertising are comparables. Internet companies that have other business models (freemium, marketplaces, commerce, hardware, enterprise software, direct response advertising, etc) are probably not comparables. The public markets seems to agree with this. Defensible companies with non-display-ad business models have maintained healthy public market valuations.

– One counterargument to the “all social media companies are now worth less” argument is the discrepency between how the smart Wall Street money and smart internet money views Facebook and social media companies generally. The smart Wall Street money thinks like Mary Meeker’s charts. They draw lines through dots and extrapoloate. This method would have worked very poorly in the past for trying to value tech companies at key inflection points (and tech investors know that what matters are exactly those inflection points). In Facebook’s case, Wall Street types look at revenue and margin growth and the trend toward mobile where monetization is considerably worse (for now). Smart internet investors, by contrast, look at Facebook in terms of its power and capabilities. They see a company that is rivaled only by Google and Apple in terms of their control of where users go and what they do on the internet. Smart internet investors are far more bullish than smart Wall Street investors on Facebook. Thus if you believe the internet perspective over the Wall Street perspective, you’d likely believe that Facebook and social media in general is undervalued by the public markets.

Technology and job creation

In response to my recent post “Making industries ‘garage ready’ for startups“, venture capitalist Jordan Elpern-Waxman made an interesting comment:

If I understand correctly, “garage-ready” essentially means separating design from manufacturing, i.e. “creativity-intensive” processes from capital-intensive ones. This may be an inevitable result of industry maturation and specialization, but there is a downside to it, at least for the so called “developed” nations. The result of differential costs for commodity labor, the fungibility and liquidity of capital, and the ease of transmitting both human and machine-readable information across arbitrary distances, means that capital-intensive processes – i.e. making things – migrate to locations with lower total cost of operations (which, Germany excepted, tend to be locations with lower labor costs). Another way of saying this is that nothing is fabless; the foundry is merely outsourced and moved to a cheaper location. This reality is great for the creative class and for the lower cost locations, but it’s less happy for the residents of the higher class locations that are not so lucky to be part of the creative class.

I’m not ready to draw the conclusion that this is the cause of the economic inequality in the US and malaise across Europe and Japan, but there definitely appears to be some correlation. Again, I don’t know if these results of the “garagification” of an industry can be reversed or mitigated in the name of societal stability, but if anyone can find a way to do it it would be the creative class. Unfortunately, because techies and entrepreneurs are solidly part of the creative class and perhaps even *the* primary beneficiaries of the separation of design and manufacturing, we generally avoid acknowledging or discussing the negative aspects of this trend.

Note that I said “reversed or mitigated.” Trying to reverse or stop these trends is probably a quixotic goal, but perhaps mitigation is in fact possible. For example, is it possible to create a country in which the entire labor force is “creative”? I myself have trouble seeing how such a possibility could be made real, but I’d like to see more intellectuals and entrepreneurs spend some brainpower on the question.

It is true that new technologies often lead, in the short term, to lower wages and fewer jobs. Craigslist, for example, has about 30 employees yet, by replacing the classified ad industry, eliminated many thousands of jobs (local newspaper reporters, classified ad salespeople, etc). The same could be said for almost every popular website.

On the flip side, new technologies have driven down prices (Walmart and Amazon), led to massive increases in information productivity (Google and Wikipedia), and created new income sources (eBay and Craigslist). Greater productivity and lower prices at least partly compensate for part-time jobs and lower wages.

Jordan is right that these are questions we – the technology community – should spend more time discussing.