31 Comments

Summary:

The marginal cost of delivering web-based content is approaching zero, so while some publishers continue to charge for it, they are the exception and not the rule. This has led to the reliance on advertising as a revenue model by most consumer web publishers. But if […]

supply-demand

The marginal cost of delivering web-based content is approaching zero, so while some publishers continue to charge for it, they are the exception and not the rule. This has led to the reliance on advertising as a revenue model by most consumer web publishers. But if the marginal cost of a page of content is zero, isn’t the marginal cost of a page of ads zero, too? This assumption, coupled with the financial crisis, has many people questioning the viability of advertising as a long-term revenue strategy. So is advertising dead?

Most web publishers have optimized their products to produce page views because that’s what they think they’re paid to provide. Given that the online advertising pricing model is still largely based on the cost per thousand page views (CPM), such a focus is easy to understand. But since the number of people on the planet and their consumption is finite, advertising should work as the primary revenue model on the consumer web. The key is to focus on the right objective — driving measurable revenue for your advertisers.

Economics 101: Allocative Efficiency

Figure 1 shows the model of perfect competition — the point at which the supply and demand curves intersect and yield maximum aggregate surplus (consumer + producer surplus). Economists define a consumer surplus as the value consumers would pay above actual market price (the blue area under the demand curve). Producer surplus is the value above the marginal cost of production (the pink area above the supply curve). The point at which these two curves intersect is the equilibrium point, which is also known as “perfect competition” and is generally held up — by economists, anyway — as the goal for all markets.

Figure 1: Traditional Supply and Demand Curve

img_6441

Moore’s Law and “The Audience” Supply Curve

The canonical example of a supply curve illustrated in Figure 1 assumes that the marginal cost per unit of supply increases with respect to quantity.

What if marginal cost approaches zero? That is, what if the supply curve approaches zero, as I have depicted in Figure 2? While the variable cost of software distribution is clearly near zero, people are often quick to point out the economic maxim of “In the long run, all costs are variable.” In other words, fixed costs (like buying and running servers and storage, hiring people to write code and owning buildings where employees to do their work) should really be amortized and added to variable costs, in which case marginal cost is far from zero. But Moore’s Law, open-source software, and globalization are driving the fixed costs of operating and capital expenses (on a unit-amortized basis) to zero, too. In the race between mankind’s ability to consume online content and Moore’s Law, Moore’s Law wins hands-down.

Figure 2: The Audience Supply and Demand Curve

img_64451

In Figure 2, the supply curve favors gravity over traditional economic logic and approaches zero as quantity approaches infinity. This explains not only why it’s so hard to charge a consumer for web services, but the dominance of advertising-supported services for information-based web properties — and why it appears as if 100 percent of the surplus accrues to consumers (blue area under the demand curve).

What About the Advertising Supply Curve?

Isn’t the logical conclusion, then, that advertising pricing will also approach zero? If advertisers pay based on page views and the marginal cost of an ad view is near zero, how can companies charge $10, $20 or $50 for 1,000 views?

Advertisers ultimately seek profit and use advertising to accomplish the following objectives:

1. Drive near-term purchases.

2. Drive “considered” purchases, like the purchase of a car, by influencing the consumer’s perception with repeated messages over long periods of time.

3. Increase a consumer’s willingness to pay a premium above marginal cost to improve profit margins.

It’s extremely difficult to measure goals two and three. Brand advertisers often aim to achieve all three goals and have various (unscientific) ways to measure performance. The majority of firms that can afford the luxury of spending without actually knowing whether there is a positive ROI are very large firms with large discretionary marketing budgets, a very small percentage of the roughly 40 million businesses out there.

Direct advertisers wouldn’t mind achieving goals two and three, but focus the vast majority of their measurement and optimization on the first objective. These firms can reinvest profits from marketing in more marketing until acquisition costs equal profit because their spend is measurable. Consequently, performance-based advertising is self-funding, which helps to explain the rise of search marketing (and Google).

Until there is a better way to measure the ROI of all three objectives articulated above, the big winners in online advertising will be those that have the capacity to generate a massive quantity of high-quality leads at a low cost.

Display vs. Search Advertising

In an effort to compare the efficiency of lead generation, let’s compare Google’s price and cost per lead to a very rough guesstimate of Yahoo’s non-search business. Please note, however, that I can’t find the data I need to do this well, so this is a back-of-the-envelope analysis that is surely way off. It should only be used for illustrative purposes — nothing more.

According to a recent Morgan Stanley research report, Google generated 11.3 billion paid leads in the first quarter of 2009; according to a recent blog post from Yahoo, that company generates 180 billion page views per month. Let’s assume that all of these pages have one display ad (excludes the very efficient Yahoo Search), that the average click-through rate for Yahoo display is 0.10 percent, and that all costs go against display.

In Table 1, I’ve estimated Google’s cost per lead vs. Yahoo’s non-search cost per lead. Note that Google earns revenue of only 49 cents per lead vs. my estimate of $2.93 for Yahoo’s display advertising. More importantly, the fully amortized cost to Google of providing a lead is 30 cents vs. roughly $2.17 for Yahoo. As the supply of leads increases, prices will likely experience a dramatic aggregate decline. When that happens, anyone who can’t produce supply for a cost under market price will obviously find himself in a bad place. The low-cost provider will expand market share as volume increases and high-cost providers fade away.

Table 1: Back-of-the-Envelope Comparison of Direct vs. Display Price and Cost Position

ad_supply2

One of these days someone will figure out how to measure the impact online advertising has on offline purchases, considered purchases, and a consumer’s willingness to pay. When that happens, some publishers will earn an order of magnitude more from advertising, while others will experience massive price deterioration. In the meantime, my money is on lead generation as the best way to build a consumer web content business.

Mike Speiser is a Managing Director at Sutter Hill Ventures. His thoughts on technology, economics and entrepreneurship will appear at this time every week.

  1. Korion Morris Sunday, June 14, 2009

    As a technologist and student who happens to be studying economics, this was a very refreshing article. Thanks so much!

  2. “But if the marginal cost of a page of content is zero, isn’t the marginal cost of a page of ads zero, too? This assumption, coupled with the financial crisis, has many people questioning the viability of advertising as a long-term revenue strategy. So is advertising dead?”

    I don’t understand this reasoning, you immediately lost me there. If the marginal cost of a page of ads is zero, that would be a good thing right? That would mean they could lower their CPM.

    And from the perspective of the ad host, marginal cost of an ad is never zero, because of website mutilation (opportunity costs).

    Also, why would the slope of the supply curve in figure 2 ever become negative? I couldn’t make that out from your explanation. Horizontal, I would understand.

    I probably need to review my economics, or study this later in the day :)

    1. “I don’t understand this reasoning, you immediately lost me there. If the marginal cost of a page of ads is zero, that would be a good thing right? That would mean they could lower their CPM.”

      You are right that you could lower your CPM. And so could the competition, which would ultimately lead to no market for advertising as price approached zero. This is exactly why it’s so hard to charge for content directly. That’s one of my points. But the bigger point is that you can share duplicate content with the world at no marginal cost because there isn’t a cost to doing so (other than power, etc.). Whereas, once you convince someone in the market for one car to buy one car, that consumer is no longer in the market. So content “supply” and “ad” supply are not alike, despite the implicit assumption that they are.

      “Also, why would the slope of the supply curve in figure 2 ever become negative? I couldn’t make that out from your explanation. Horizontal, I would understand.”

      It doesn’t go negative. It approaches zero.

      1. “It doesn’t go negative. It approaches zero.”

        First of all, on your graph, the supply curve goes down. That means the slope is negative, per definition.

        Second, you say they will end up lowering their CPMs until it’s 0, but really you’re ignoring the value that the ad brings. I mean, this value may be hard to measure, but it’s still very real. It’s definitely above 0. This means that the equilibrium in perfect competition will not be 0, but above that.

        Perfect competition theory says that you cannot make profits in the long run. Marginal benefit=marginal costs. But this does not have to mean that it ends up being zero! It merely means that you will be paying exactly what you will get in return. You will be paying an amount above 0, because the value of the ad is more than 0..

  3. Big hole in argument :
    Your logic is purely based on costs on of delivery. Which might be near the net cost for a content aggregator like Giga OM. but very different for a content owner/author like WSJ.

    1. Niraj,

      You are right if the content owner: (1) has content that has no good substitute, and (2) the content owner controls distribution. I agree with you that #2 is easy, but for the vast majority of content out there there are great substitutes. I read the WSJ for years, I’m a business guy, and investor, and I no longer have a subscription. There is very little at the WSJ that I cannot find in the New York Times, GigaOM, Twitter, and so on.

      For the early part of the demand curve there are often a small group of loyalists that believe that no good substitute exists. You can often charge these people, either like the WSJ or you can try the Freemium model like GigaOM (some content free, some for charge).

      If you want a nice business, that might work. If you want the next Google, it’s hard to see how such a model will work. Because as you move to the right on the demand curve (as the number of people approaches infinity), the view that said content has no good substitute drops dramatically.

      1. “There is very little at the WSJ that I cannot find in the New York Times, GigaOM, Twitter, and so on.

        This is because the model is not mature and there is significant subsidy built into the system. Subsidies don’t last for ever. At some point NYT and WSJ will charge for content production when they cannot monetize using their own captive distribution mechanisms.

        The ground rules of economics of hyperlinks will evolve and the costs of GigaOM type aggregators will eventually increase , increasing the net-distribution costs also(i.e if you view GigaOM as a distribution channel)

  4. Tom Eisenmann Sunday, June 14, 2009

    Agree with Niraj that analysis omits content production cost. Likewise, it ignores marketing costs required to boost supply of impressions. Consider how much money MSFT will spend to boost paid search share with Bing. At the macro level, size of audience and the time they spend on Web is limited, so increasing supply of impressions entails stealing share from other media, which in turn requires spending on content and marketing. So, seems unlikely that supply curve slopes ever downward.

    That said, analysis of cost per lead for Google vs Yahoo is clever and revealing. Thanks!

    1. Professor Eisenmann,

      Haven’t seen you in years since you taught a class for two of mine! First, see my argument about substitutes made to Niraj.

      Your point about marketing costs rings true for certain players. But Microsoft is a late entrant in a positive feedback market where the only way into the game is to buy distribution. Instead look at Google’s or Twitter’s marketing costs? And while it’s true that Google spent an enormous sum on MySpace, they later came to regret that move as a bad economic decision. Even after amortizing all marketing costs (which I did along with all OPEX in my quick and dirty analysis) the marginal impact to marginal costs are, well, marginal.

      Your point on the macro level is unquestionably true (with the exception that you needn’t necessarily steal from media, but from any other activity that occupies time). Just as consumers are limited on time and spend, so too are consumers limited with respect to time. That was a key question I had as I was drawing my second figure — I came to the point of view that consumer budgets are dramatically more fixed than audience time. And that spending on content and marketing have less to do with gaining attention than do having the right product, which is a relatively small cost and a fixed one at that (YouTube and Twitter both produce enormous sums of content and did so with tiny teams and minimal marketing / content budgets).

      Thanks for reading and commenting. I’ll think through the macro point some more.

      -Mike

  5. “Isn’t the logical conclusion, then, that advertising pricing will also approach zero? If advertisers pay based on page views and the marginal cost of an ad view is near zero, how can companies charge $10, $20 or $50 for 1,000 views?”

    Yes, because ad revenue doesn’t only serve to cover for production/hosting/… costs, but also for opportunity costs.

    Opportunity cost = the economic value of the lost space the ad occupies on the website. *Including* the profit that could have been made had the space been sold to a competing advertiser…

    With all due respect, and coming from a mere last year student of applied economics, I think the article is quite flawed. Especially the reasoning behind figure2 is not well documented. Usually, marginal costs != slope of curve.

    1. I’m making a contrarian point, so you’re unlikely to find Figure 2 in your textbook/s. You are also unlikely to find a good analysis of the impact of vast improvements in technology on the true cost of supply for web based content in those books — at least the ones I’ve read lack such an appreciation.

      Be careful to accept what you read as an article of faith — the efficient market hypothesis which is also likely in those books does not seem to be supported be recent empirical evidence in the markets ;–)

      If I were on the other side of this argument, I think Professor Eisenmann’s question above with respect to the macro question of the finite minutes available to a human is where I would focus.

  6. How can the supply curve go downward? This would mean you will get more supply at a lower price than you would for the same good priced higher?

    With all due respect, I believe you are not making much sense here….

    1. Mike Speiser Sam Sunday, June 14, 2009

      I’m trying to convey many things at once, so perhaps text is better than a graph.

      Since we’re talking about marginal cost of the entire industry the thing that would drive “marginal” costs up would be scarcity of some asset. If there is no scarcity, then the curve is basically horizontal. Or, if we overlay time as well and marginal cost decreases with respect to time, then the curve might even decrease.

      So, it’s clear to me that over *time*, the marginal cost of production of online content and services is going down. This is a result of Moore’s law in CPU and storage, open source software, cumulative knowledge in the engineering community about how to build these services, off-shoring, and so on.

      The question is on the scarcity bit — my basic argument implies (1) that we’re far away from having to steal time from one service in order to grow a new service, and (2) that buying content or eyeballs to achieve #1 isn’t a viable strategy.

      Many folks here are arguing against both of the scarcity assumptions. I’ll think about it some more and appreciate your comment.

  7. It has nothing to do with cost, but with value provided. You link to Wall Street Journal, but a better example would be GigaOM Pro. It’ll do well because regardless of its costs to produce it will generate its audience far more money than it costs them.

    1. If you believe that no good substitutes exist, I agree. I don’t believe that is often the case…

  8. Matt — 100% on the money..who cares about the marginal cost …same with wireless networks- marginal cost is almost zero, doesnt mean that its of 0 value…..

    1. Mike Speiser SS Sunday, June 14, 2009

      You guys are right if no good substitutes exist. If the NY Times charged tomorrow, do you think 100% of their customers would pay rather than switching to the LA Times? The Washington Post? 50%? 5%? How about Google — if they charged everyone for usage, what percent of consumers would pay?

  9. Funny, I was taught about Variable Cost, not Marginal Cost, but it’s the same thing. The thing with this logic is it depends on how you’re looking at the business model. Like Niraj and Tom were mentioning, this variable cost is ignoring the cost of producing the content. One could argue that the cost of paying a content author to produce an article is a fixed cost since you pay a certain fee (or if they’re a staff writer, the time spent on the article X their salary rate = the fixed cost of that content). But that assumes your business model is about paying out content authors upon delivery and not maintaining a staff of writers, artists, photographers, technology/IT, and the like. So, the other side of the coin is the argument that you need to pay all these people and they’re contributing to the cost of producing an article or whatever content you’re providing. If that’s the case, you’ll never hit a cost of zero.

    Technically, even if you ignore those costs as fixed costs, you’re still never going to really hit zero. No matter how many decimal places you’re talking about, each person hitting your site costs you in terms of bandwidth and electricity. Unless you’re getting both for free, your cost isn’t zero. If you’ve got millions of people hitting your site, you’ll need more resources (even if you’re using a cloud) so there’s still increased (and thus variable) cost per visitor…even per visit.

    I hate these sorts of articles. You’re ignoring a basic fact: supply and demand has less to do with cost as it has to do with profit. And it doesn’t have much basis here. If you really want to talk economics, you would be focusing more on what price you can set based on your audience size. And, that’s not based on supply and demand. Just because you have a larger audience doesn’t mean there’s less demand. It’s the opposite.

    Advertisers are looking for larger volumes of eyeballs so that they can increase the number of clickthroughs and impressions because they’re trying to take that dismal 1-5% conversion rate and multiply it by a larger number to get a larger conversion rate. It’s the assembly line mentality. Produce more in less time so that you sell more and make that margin of yours worth more in the end.

    Instead, advertisers should be looking at how to improve their margins, so to speak. How do you improve the quality of the ad or at least the fit of the ad (better targeted, narrow focus) so that you take that conversion rate and make it 40-50% instead? If you can do that on a site that gets 10k visitors a month versus 10million visitors a month, you can reduce your costs because that smaller site (at least given the current model) won’t necessarily be able to argue for a higher advertising rate. There’s no justification. Of course, if the operator of that 10k site is more savvy and understands that they have a core demographic that’s well focused on the advertiser’s target demographic, they can negotiate for a higher rate to get the money that would’ve been left on the table (and in the advertiser’s pocket). But that’s a different economic principle… :)

    1. Mike Speiser Ken Sunday, June 14, 2009

      Marginal cost and variable cost are not the same thing (http://en.wikipedia.org/wiki/Marginal_cost). The back-of-the-envelope analysis included variable + fixed costs.

      To summarize the criticisms and my responses here, before I power down:

      + My argument assumes that there are good substitutes. If there aren’t good substitutes and people really want the content, you can charge for it. But that’s rarely the case — if it were, everyone would be paying for content.

      + My argument ignores the growing marginal cost of content production. I think this criticism is rooted in old line media thinking. Twitter’s content grows as its content grows. How do their costs of content acquisition grow over time? Google’s? Facebook’s? This is true for mainline media if you are paying reporters — even there, bloggers are keeping a cap on the cost of production…

      + My argument ignores the growing marginal cost of marketing. Same fundamental argument as above. What are Twitter’s marketing costs? Facebook’s? Eventually there is truth to thus, but to be safe I included ALL fixed costs in my analysis (marketing, content, G&A).

      + My argument assumes that we are far away from having to trade attention between online services. This is the fundamental macro flaw in the argument, according to Professor Eisenmann — that’s a fair argument. I will think about that one some more. It’s clear that the delta between this for consumers and advertisers is so far apart that it’s not an issue yet, but I can certainly see how looking at the long-term that this is a fundamental limitation…

      Thanks for all the great feedback.

      1. Hi, Mike,

        Thanks for the responses…you started a great conversation even if we’re all focusing on the details rather than the big picture. :)

        I actually did read that Wiki and if you notice, it talks about the change in FC and VC over the change in quantity. Fixed costs by their very definition are not going to change from unit to unit so you’re talking about the change in variable costs per unit which would be the variable cost of the unit. Splitting hairs because I know you can argue that “variable cost” is usually rolled up over the entire production but still.

        Back to the main gist of your post, again, I might be splitting hairs but like I said before, you’re never going to NOT have a variable or marginal cost that is greater than absolute zero. Sure, it’ll end up being 0.0001% of a penny at some point but it’s still a cost due to electricity and bandwidth costs.

        Electricity isn’t getting cheaper and though Intel and other technology companies have sort of veered towards low heat/more energy efficient products there will always need to be an electrical source pumping electrons into that equipment.

        Bandwidth has been getting a bit cheaper but we still run into roadblocks (whether it’s dark fiber or the geographical disparities of Internet reach). Again, like electricity, newer technologies that are expanding our ability to connect to the net from wherever, including GigaOm favorites like LTE and WiFi-Max. However, these still will have to run the gamut of Moore’s law-like efficiencies to bring costs down so that they are negligible.

        And, while you want to ignore marketing and production costs (I understand, we’re talking about consumer Web production which is really “free” but I’ll get to that), you can’t really. At least not completely. The reason Facebook and Twitter are whether they are (on everyone’s radar and “producing” so much content is because of the network effects that have finally reached a critical mass. If you looked at either three or four years ago, people were barely using them. How do you value that? You can argue fixed costs there which is why I’m not saying you HAVE to include any sort of marketing/sales budgets or whatnot into your calculations.

        I feel, though, you still have to either find a cost that makes sense or at least asterisk the line item because outside of determining the advertising price/rate for an existing business, this sort of analysis would be what you’d need for a business looking to dip their toe into social applications like Facebook or Twitter (or to quote Dennis Leary: “Facyspacies and your TweetyPages”). The cost of a new business trying to achieve that critical mass in a shorter time frame than the existing players did would add (potentially) to the variable and marginal costs.

        Again, splitting hairs. Overall, you’re on the right track when it comes to the low cost of content acquisition for an existing enterprise and I am surprised that advertising costs haven’t become a commodity in the Websphere. Obviously, there’s still some perceived value in the sites that have the most eyeballs that is in complete opposition to what you’d expect in the “real world” model of supply and demand. I don’t know if you can really escape that when it comes to virtual commerce.

        Look at virtual goods (I’m talking microtransaction virtual goods, not gold farming and other labor intensive virtual goods). They have the same thing going for them. Outside of bandwidth to download and electricity to run the sites where you download them (or transfer them), their production costs are fixed (you build the item once and distribute many). Still, they can be priced pretty high considering the negligible cost to manufature them.

        Always great to participate in a stimulating conversation. :)

        kn

  10. Mike Speiser Sunday, June 14, 2009

    Agree Ken, that marginal cost will never reach zero, but rather just approach zero. Or perhaps it won’t even do that, as many others have noted. You raise some very good and well thought-out points. And I truly appreciate the time and thought you have put into this post . -Mike

Comments have been disabled for this post