Internet Economy of Scale
The original internet bubble failed because people forgot the main point: The reason you automate things is to make them automatic. Automation reduces the manpower requirements and thus reduces costs.
Why, then, did Critical Path have over a thousand employees?
In the earlier days, ISPs had to provide many services: connectivity (of course), email, netnews, ftp, shell accounts, etc. Each of these services requires a certain amount of manpower to maintain. As the size of each service grows, the manpower grows too, but sublinearly. It takes a whole person (because you can't have 1/2 person) to maintain an email service for 10 accounts, 100 accounts, or 1,000 accounts. You need a small team to handle email 1,000,000 accounts. You might need a slightly bigger team to handle email for 100,000,000 accounts. The key though is that that you don't need 100 times as many employees to manage email for 100 times as many accounts.
That's economy of scale. The idea behind Critical Path was economy of scale. If you consolidate the email requirements of dozens of ISPs into a single outsourced shop, then you can dramatically reduce the number of employees. Critical Path should have been able to handle the email for 1,000 ISPs with maybe dozens of employees. But they had more than 1,000 employees at their peak. It's easy to see why they didn't make any money.
(Please disregard the fact that, at one point, the stock market valued Critical Path at billions of dollars.)
Since Critical Path failed, many companies have stepped up to fill their shoes. There are zillions of low cost providers of email who do a competent job. In fact, there are many providers of free email that do a better job than the for-fee companies.
That's the power of economy of scale. With enough users, the opportunity gets ahead of the costs. Companies can offer free email at large enough scale that the marginal cost of an account is less than the ad revenue each account generates.
Today, the wheel of reincarnation is at work, bringing us managed hosting at reasonable costs.
What do you have to do to deploy an internet server? You need to have a data center (power, air conditioning, connectivity), the physical server, and software, software, and more software. You need to install an operating system. You need to update the operating system. You need to configure IP addresses, gateways, RAID, etc. You need backup procedures and restore procedures. And lots of other things.
The same economy of scale applies. I can rent a managed server from many companies today at very reasonable prices. I then need *no* administrators.
That's the theory. In practice, the managed server companies fall into two camps: too expensive or too incompetent.
In the too expensive category, the consideration is how much it costs for managed hosting versus doing it all yourself. If a server costs $1,000 per month, and an administrator costs $100k per year, then the break even is at about 8 servers. If you need fewer than 8 servers, then clearly the managed hosting makes sense.
Be careful though: if you think you might need more than 8 servers someday, and you start with managed hosting, there is a trap. When you get to the point that you need more than 8 servers, you are likely in a situation where you are growing fast and don't have time to rethink your hosting situation. You might end up renting 20 servers or 100 servers at a price that will seem very expensive. Of course, that's a high quality problem.
What about the fact that a managed hosting provider will be able to apply economy of scale to the administration and do a better job than your one employee could do?
Which brings us to the other category: too incompetent. When theory meets practice, you have an opportunity for discrepancy. It's true that the managed hosting companies should be able to do better than employees. The problem is that, in many cases, they aren't doing so yet.
I've been testing the waters for inexpensive managed hosting, and it's not pretty. It has to get better because economy of scale and the free market will make it happen. It's just not there yet.
Why, then, did Critical Path have over a thousand employees?
In the earlier days, ISPs had to provide many services: connectivity (of course), email, netnews, ftp, shell accounts, etc. Each of these services requires a certain amount of manpower to maintain. As the size of each service grows, the manpower grows too, but sublinearly. It takes a whole person (because you can't have 1/2 person) to maintain an email service for 10 accounts, 100 accounts, or 1,000 accounts. You need a small team to handle email 1,000,000 accounts. You might need a slightly bigger team to handle email for 100,000,000 accounts. The key though is that that you don't need 100 times as many employees to manage email for 100 times as many accounts.
That's economy of scale. The idea behind Critical Path was economy of scale. If you consolidate the email requirements of dozens of ISPs into a single outsourced shop, then you can dramatically reduce the number of employees. Critical Path should have been able to handle the email for 1,000 ISPs with maybe dozens of employees. But they had more than 1,000 employees at their peak. It's easy to see why they didn't make any money.
(Please disregard the fact that, at one point, the stock market valued Critical Path at billions of dollars.)
Since Critical Path failed, many companies have stepped up to fill their shoes. There are zillions of low cost providers of email who do a competent job. In fact, there are many providers of free email that do a better job than the for-fee companies.
That's the power of economy of scale. With enough users, the opportunity gets ahead of the costs. Companies can offer free email at large enough scale that the marginal cost of an account is less than the ad revenue each account generates.
Today, the wheel of reincarnation is at work, bringing us managed hosting at reasonable costs.
What do you have to do to deploy an internet server? You need to have a data center (power, air conditioning, connectivity), the physical server, and software, software, and more software. You need to install an operating system. You need to update the operating system. You need to configure IP addresses, gateways, RAID, etc. You need backup procedures and restore procedures. And lots of other things.
The same economy of scale applies. I can rent a managed server from many companies today at very reasonable prices. I then need *no* administrators.
That's the theory. In practice, the managed server companies fall into two camps: too expensive or too incompetent.
In the too expensive category, the consideration is how much it costs for managed hosting versus doing it all yourself. If a server costs $1,000 per month, and an administrator costs $100k per year, then the break even is at about 8 servers. If you need fewer than 8 servers, then clearly the managed hosting makes sense.
Be careful though: if you think you might need more than 8 servers someday, and you start with managed hosting, there is a trap. When you get to the point that you need more than 8 servers, you are likely in a situation where you are growing fast and don't have time to rethink your hosting situation. You might end up renting 20 servers or 100 servers at a price that will seem very expensive. Of course, that's a high quality problem.
What about the fact that a managed hosting provider will be able to apply economy of scale to the administration and do a better job than your one employee could do?
Which brings us to the other category: too incompetent. When theory meets practice, you have an opportunity for discrepancy. It's true that the managed hosting companies should be able to do better than employees. The problem is that, in many cases, they aren't doing so yet.
I've been testing the waters for inexpensive managed hosting, and it's not pretty. It has to get better because economy of scale and the free market will make it happen. It's just not there yet.
posted by Tim at 9:40 AM
0 comments
