Stardate 20030502.1706

(Captain's log): I've written many times about economy of scale and about network effect. They often both apply but they are unrelated as a practical matter, unless there are complicating factors (about which more later).

When it comes to commercial products, you begin with the fact that cost, value, and price are different things. The cost of an item is the total amount of money that must be spent to create it and deliver it to the potential customer. The value, on the other hand, is the perception the customer has of how useful and desirable that item will be. And those two are almost entirely unrelated to one another.

The price, in turn, is how much money the customer must fork over to get the item. If everything goes well, the value will be higher than the cost, and the price can be set in the sweet spot between them. It will then be higher than the cost (so that the producer makes a profit) and lower than the value (so that it's perceived as a bargain by the customer).

In general, the price must be higher than the cost, or else the business won't be viable. And in general, customers won't buy if the price is higher than the value. But customers won't all have the same calculation of the value, and as the price rises what you get is a smaller and smaller number of people willing to buy. But in many cases the situation is more complicated than that.

As volume increases and more of the units are sold, you can get network effect, which means that the value of the item rises as a direct result of increased usage. The classic example of network effect is telephones: the value of the telephone in my home is a function of how many people I'm capable of calling using it, so the more people who have phones, the more valuable my phone is to me. Computer operating systems are also strongly subject to network effect, because when there are more users there will be more applications developed that can run on that OS, making the OS itself more valuable because it grants the customer access to a larger app base.

This is not invariant. For instance, pharmaceuticals are usually subject to economy of scale, but in general are only weakly susceptible to network effect, if at all. If I have a disease then the value to me of the drug that treats it isn't a function of how many other people are also being helped by that drug. My calculation of the value of the drug is a function of how serious the disease is and how effective the drug is at helping me.

In some cases you can get negative economy of scale, where increased production raises the unit cost. (Usually that's the result of component shortage.)

And you can get negative network effect, where the value of an item drops as it becomes more common. The most extreme case of that is "collectibles", but it can happen in marketplaces, too. The reason for this is that rarity can drive value.

Take, for example, the Rolls Royce. They're costly to build because so few of them are produced. The high cost explains the high price, but doesn't explain why the value is so high, and why anyone is actually willing to pay that price. A Rolls may well be a better car than my Chevrolet, but it's not clear that it's 15 times better on a purely technical evaluation of the car's features. Nonetheless, the selling price of a Rolls is more than 15 times what I paid for my Chevrolet, and Rolls has never had any trouble finding buyers for the cars they produce.

So where is the value coming from? Interestingly, it's precisely the fact that so few of them are made. To a great extent, it is the high price itself which actually makes the value high. It's because a Rolls is exclusive, rare, out of reach of most people. If Rolls Royce were to adopt largescale manufacturing techniques used elsewhere in the auto industry, they could produce the same product they make now, at the same quality level, at 100 times the volume and at a lot lower cost per unit. But if they did so, the perceived value would also drop because it would no longer be seen as being anything like as distinctive to own and drive a Rolls. The cost would drop, but so would the value, and it's not at all clear which would drop faster. If the value dropped faster than the cost, at some point there would be no sweet spot for the price where the product would sell and also be profitable.

Manufacturers such as Ferrari are in a similar position. Ferrari appeals to a different group of customers, but it gains nearly as much from the perception of rarity and exclusivity.

In the car industry, different companies tune their costs and values for different points on this scale. To some extent cars do get a positive network effect from the fact that a car brand which is broadly sold will also be broadly supported and you won't have trouble finding people to work on it. But in general, car brands are subject to negative network effect, with the perceived value dropping with volume primarily due to factors of prestige; to a great extent because the car companies have deliberately been using such intangible factors as a way of differentiating themselves and building brand loyalty. The Porsche business unit of Volkswagen appeals to much the same customer base as Ferrari but has tuned its business model for a lower point, with higher volume, a lower effective value, and a lower price than Ferrari, though a great deal higher point than Chevrolet did with the Cavalier that I own.

Now clearly it's the case that a Porsche 911 is a much better car than my Cavalier. It certainly has a higher value on objective grounds, though less than you might think because much of the difference is useless. For instance, it has a much higher top speed than my car, but it isn't legal to drive that fast on any road in the US. However, some people like the idea of owning a car which is capable of that, even if they never use it. Despite that, most of its increased value is due to prestige and to rarity, which in turn is actually supported by the high price that Porsche charges. It's exactly the fact that it's expensive that makes it worth the price to many of its customers because it's out of reach of the masses.

Almost any kind of award is subject to negative network effect: military decorations, for example, or such honors as knighthood, or visits to the Presidential Ranch. (The only American military decoration I know of which is not subject to negative network effect is the Purple Heart, for reasons which should be obvious. The perceived value of most other decorations has fallen over time as a result of the military equivalent of "grade inflation".)

A few days ago I personally ran into another manifestation of inverse network effect, when I changed the links in my sidebar.

The value of any given link in any given site's blogroll is proportional to the overall volume of traffic that site gets, and inversely proportional to the total number of links in that blogroll. So membership in any given site's blogroll is subject to negative network effect.

Not absolutely, of course. Like all such things, it's more complicated than that. I get a large number of refers from Glenn Reynolds (hundreds per day) even though his blogroll is immensely long. But that's because a lot of people use Instapundit as their browser home pages, and a lot of them use his sidebar to visit favorite sites. Most of those refers from Instapundit are regular readers of mine, who get to me through his blogroll instead of using a direct shortcut stored locally. They actively seek out the link to me and use it, and as a result I have no doubt I get a lot more regular refers from Glenn than do Nikita Demosthenes or Nick Denton, who are adjacent to me in his list. But since they're mostly regular readers of mine, most of them would keep visiting me even if Glenn's site went down, because they'd switch to using local shortcuts or someone else's blogroll.

Because his list is so long, it's unlikely that very many people have found my site for the first time by clicking that link in his blogroll. (That said, I'm still happy to be on it.) Most of the people who have discovered me through Glenn's page (and many have) followed links to specific articles I've written that Glenn was kind enough to link to from the main body of his page.

I started my blog in March of 2001, and for months my traffic level was stalled at about 300 page loads per day. I was, I thought, producing good material but there didn't seem to be any way for people to discover that my page existed, so as to even get a chance of seeing it, and maybe becoming intrigued and becoming regular readers.

I once referred to this as stickiness, the chance that any given first-time visitor will be impressed and decide to keep the link and visit again. Some proportion of first-time visitors will become long time readers, and any given blogger can affect that by the quantity and quality of material produced for the site. To that extent, stickiness is a function of merit and hard work, as judged in the marketplace of the blogosphere. But in order for someone to stick, they have to visit. What made me feel helpless was that there didn't seem to be any way for me to bring people to visit in the first place. Until that happened, they wouldn't even have the opportunity to stick.

Eventually a few people who had more volume than I did became semiregular readers and linked to specific articles of mine, especially in the aftermath of the attack in September of 2001, because I was providing some insight into military affairs that was unique and which some thought was valuable.

Looking back on my own experience, what I found was that after I reached a certain sustained traffic level, (somewhat above a thousand page loads per day) it kept growing on its own. That was because a certain critical mass of traffic meant that owners of other sites who specialized more in linking (as opposed to my own primary concentration on writing) were beginning to watch me, and I began to get more and more refers from them when they linked to specific things I'd posted, and thus gained a moderately sustained rate of first-time visitors who might stick. That became self-sustaining, with more regular readers leading to more links leading to yet more first-time visitors. To some extent it proves yet again that them as has, gets. The problem is one of bootstrapping; if you don't have any exposure, no one can learn that you exist. Once you have some exposure, you'll naturally get more exposure.

During that first summer of discontent, what I really wished for was a favor from someone higher up in the food chain. So having become relatively successful now, I've been attempting to do for others what I wished had been done for me. And there lies the problem: it isn't possible for me to help everyone who wants to be helped. It's not that I'm a prick and won't do it; it can't be done.

I don't have blogroll links to people like Reynolds; Glenn doesn't need the traffic. Instead of keeping a hugely long and ever-growing blogroll which includes all the top-traffic sites along with a swarm of others, I've tried to keep mine short so that the value of each link is high to the owner of that site. And what I've been trying to do is to select sites (all of which have been writers rather than linkers) which have been consistently producing quality material on an ongoing basis but who don't seem to have gotten a great deal of exposure. As one of my ongoing background projects, whenever I encounter a site which seems to be doing a lot of writing, I'll keep a link to it in a special "candidates" folder and will revisit it from time to time to see if they've kept it up, with regular posts and consistently high quality. And every few months I harvest the contents of that folder and you see the result.

What I'm hoping I can feed a lot of first-time visitors to them on an ongoing basis for a few months. If they keep producing, then at the end of that time what I hope is that they'll have built up a large enough base of regular readers so that they'll be over that same threshold and will be able to keep building their readership thereafter. And if they haven't managed to do so after such a period, they probably never will. Thus every few months the short list will be changed, with the old sites being removed and a new list being put in their place, to try to do the same for them.

The reason for preferring writers is that I'm a writer, and my readers are probably more interested in finding writers. It's mostly a personal choice; I don't claim that writers are somehow better than linkers. But my refer traffic is probably going to be more valuable to other writers.

Of course, if anything this makes a place on the list even more valuable and I get a lot of letters from people asking to be put on the list. There are problems with this.

If, for example, I linked to everyone who has linked to me, that list would be preposterously long. If I put everyone on it who asked, the list would become extremely long and the value of each link on the list would be negligible. The cruel logic of the situation is that part of the reason it's seen as valuable is because I don't respond to the several letters per week I receive that say, "I've added you to my blogroll, will you add me to yours?"

Basically, no, I won't. None of the sites I've put on the list have expected it. I don't send them mail, for instance, when I do it. (It's always seemed to me that it would be pretentious to do that, almost like I was somehow expecting them to send me some sort of genuflecting letter of gratitude. Some have been grateful, but none of them owe me anything.)

Sadly, another reason I don't respond to such requests is that Sturgeon's Law is in full force in the blogosphere: 90% of blogs are crap, if not an even higher proportion than that. (Not yours, of course.) There's a benefit to being on my blogroll for the site owners, but it's also there for my readers, and I don't want to route them to sites which are a waste of their time.

Now it might be noticed that there is a part of my blogroll which changes and a part which doesn't. The top section is a small group of links to sites run by personal friends; they get a permanent spot, because it's my site and I'll do that if I want to. It would probably be more understandable if I labeled the sections, but I haven't managed to come up with labels that satisfy me. (As I was writing this, it occurred to me that I could give them different background colors as a way of making the sections distinct, and I'll give that a try.)

As to the rest, for a long time when I found a site and thought it was really good and wanted to add it, I feared that the list was becoming too long. In order to add a new one I'd need to remove one already there. And if I did that for only one site, the owner of that site would justifiably wonder why it was them and not someone else. So each time it would gnaw at me and I'd feel guilty even if they didn't bitch about it. (I don't need that kind of anguish.)

Ultimately the only way I could manage this so as to give the greatest benefit while hurting the fewest number of people's feelings was to replace them all once. I did it last fall and I just did so, and in a few months I'll do it again. I suspect I'm going to start doing this more often, likely every three months or so. I can't solve everyone's problems; I can't drive traffic to every deserving site (let alone to the other 90%). But I can at least help a few people out and help them get the exposure I think they deserve. (And I will make the decision on who I put there, because it is my site.)

It isn't possible for me to please everyone. People I don't include hate me; people who were on the list and get removed resent it and make elaborate plans for revenge. But if it were long, it would be of negligible value. That's one of the problems with reverse network effect: it is precisely the exclusivity that makes it valuable, and therefore it is unavoidable that the majority be disappointed.

If everyone who wanted a Rolls Royce had one, most of them wouldn't want a Rolls Royce any more.