Part 1 - Dial Access Basics
To set the right price on your service offerings, you've gotta understand your cost structure?not necessarily an easy thing for ISPs to get a handle on. Here's one way to figure it.
I 'm not an accountant but here's my carefully considered outlook on how to price. Calculating costs and determining pricing can be difficult for ISPs, yet clearly it's crucial for the longevity and profitability of your company.
Ideally you would make profit on everything that you sell. In the real world, profit is not always the direct goal of every business transaction. You may price one item at or below cost because you need to acquire a critical mass of sales of that product. In the Internet business, it's a popular practice to sell dial-up at break-even?or even at a loss?to make money when the company is sold.
One trap you certainly don't want to fall into is mindless price wars with your competitors. Your competitors may have a different cost structure. They may also be selling at or below cost in order to acquire as many customers as they can.
Who are your real competitors?
First, you need to make sure that the people you are pricing against are truly your competitors. In other words, you choose who your competitors are. If you are an 'A' level company, with 'A' level service, price it at 'A' levels. If you are a 'B' level company, with 'B' level service, price at 'B' levels. If your competitor is truly a 'A' level service, but is charging 'B' level prices, either their costs are artificially low, or they are not making sound business decisions. In either case, they're risking failure. If they're pricing artificially low, they'll have to struggle when they go up .If they are making bad business decisions, they are going to hang themselves.
Tip:
Determine your costs and pricing before you look at your competitor's pricing. If after you've done this you find your competitor's pricing is lower then yours, you can adjust yours slightly. Remember. In multiple polls of what business users really want, price ranked below both quality of the network, and quality of service. For most ISPs, there are three main categories of service that need to be priced:
Dial-up access (56k or ISDN),
Web hosting (both virtual and not) and
Dedicated access (point-to-point, and frame 56k and up.)
The remainder of this column will address pricing for dial-up access. We'll look at the other categories in succeeding weeks.
[DISCLAIMER: All numbers from here on are chosen for the sake of illustration. It is not the numbers that are important, it is the equations. -JZ]
Pricing Dial-up
The first step in determining costs for dial-up access is determining the PPP access cost.
If you don't outsource, you need to figure out your own port access cost before you can determine your PPP access cost. Best choice here is to hire an accountant. The rest of this column is for those of you who choose to go the outsource route.
If you are outsourcing dial-up, you either pay per user or per port. If you are paying per user, you automatically have your per user cost for PPP access. If you're buying per port, then you need to figure out your optimum and operational user-to-modem ratios.
The optimum ratio is the user-to-modem ratio you'd have if you were not growing and had the exact number of users per modem you want to operate at. There is a trap here. Unless you have capped the number of users you serve, replacing the ones you lose as you go, you will NEVER be exactly at your optimum rate.
Pricing Your Services ?? Part 1 ?-?continued
So, you want to figure out your projected operational rate. The reason for this is growth. If you have 1,000 users, and a 10:1 ratio, you would have 100 modems. This is great for optimum situations, but lets say you add 100 users that month. Then you would have 1,100 users for 100 ports, therefore a 11:1 ratio. If your optimum rate planned for few if any busy signals (which it should), then your operational rate of 11:1 will cause busy signals?and cause you to loose customers. In turn, your growth rate will drop and your cost of acquisition for new users will go up. This is a nasty cycle you'll be wise to avoid. If you massively exceed your optimum rate, you can end up loosing as many or more than the number of users you added. I've seen it happen.
Easy does it
In order to ease your growth and defray your marketing costs, you actually end up having existing users pay you for adding new users. Going to my last example, you plan on adding 100 users this month, for a total of 1,100, and your optimum rate is 10:1, you should have 110 ports going into the month,?a 1000-user-to-110 modem ratio (roughly 9:1). So if your port cost is $50, each user's share is roughly $5.55?a cost of $5.55/user for PPP access.
This number can of course vary from month to month. I would suggest that smaller ISPs try to plan to have at least 20 percent excess capacity at all times. So if your optimum user-to-modem ratio is 10:1, you should plan your costs on 8:1. After tracking your growth pattern for six months or more, you'll start to learn what modem-to-user ratio is right for your operations.
Tip:
If you use this cost determination method, you don't necessarily have to give the users the break if you find out your costs are lower. I have met some ISPs who automatically drop their prices when their costs go down. This doesn't mean you might not want to drop your prices, it just means think first.
What's the right ratio?
I can't in good faith tell you at what user-to-modem ratio you should operate. It's a sliding scale that factors in size and the service you wish to provide to your users. But here's how to calculate it:
Start off with your RADIUS logs. You need to know how long the average user stays on, how many times s/he logs on in a given day, and when your peak usage is. Then you adjust your ratio, keeping it low enough that users won't get busy signals at peak times.
With some creative setup you can get MRTG to graph how many modems are in use at any time. If you find that your peak usage hours are, say, between 6 PM and 10 PM, you need to look at that 4 hour period in your logs.
Note that I say 'average user.' The line campers who are on throughout your peak hours really mess up your user-to-modem ratio. You need to be sure your pricing and AUP/TOS address this.
If the average user (I am making up numbers, it varies widely) stays on for 24 minutes of peak time, every day, (12 hours/month.) then you have to set aside 24 mins out of the peak time for each user which would give you a 10:1 modem to user ratio.
Here's the math:
6 PM to 10 PM = 4 hours of peak time.
4 hours of peak time = 240 minutes of peak time
240 minutes of peak time /24 minutes per user = 10 user/peak period
Therefore, 1 modem can handle, on average, 10 users (10:1 modem:user ratio)
These user-to-modem ratio calculations assume that not all of your users call in at one time. The formula above is based on the assumption that all of your ports were used completely throughout the peak period. Since peak time tends to look more like a bell curve, if you want to make sure you have as few busy signals as possible, you need to have a slightly lower user-to-modem ratio.
Those of you with better math skills than mine could probably refine the calculation to correct for the bell curve, which would of course produce more accurate (and useful) results. But this will get you into the ballpark.
Part 1: Dial Access Basics
Part 2: Basic Operations Costs
Part 3: Web Hosting
Part 4: Dedicated Access
Part 2 - Basic Operations Costing
Having detailed outsourced PPP costs and user-to-modem ratios in our first pricing column, today, we'll examine a number of other basic ISP operational costs.
Equipment costs
Depreciation schedules are an important factor in determining monthly equipment costs per user. Unfortunately, unlike some unnamed telcos, ISPs can not depreciate their equipment over a 30 year schedule. Check with your accountant about the finer points of depreciation schedules and their effect on taxes and the like. For purposes of this discussion, we'll assume a moderately conservative schedule of two years (24 months).
Using a 2 year depreciation schedule, and assuming a per-port cost of $240 and a 10:1 modem-to-user ratio, your basic equipment cost per user would be $1 per month.
Here's the math:
2 years * 12 months = 24 months.
$240 per port /24 months = $10/port/month
$10/port /month /10 users/port = $1/user/month
Access line
Next question is on the access line, CT1, or PRI. If your cost per CT1 is $480/month., and there are 24 lines (ports) in a CT1, and a user to modem ratio of 10:1, your cost per user would be $2 per month.
Here's the math:
$480/mo/CT1 /24 lines/CT1 = $20/line/month
$20/line/mo /10 users/port = $2/user/month
If you are paying rent or co-location cost on your equipment, you would use the same equation as above.
Access costs
When figuring the cost for internet access (from your provider) per user, the math is interesting. A lot of people have tried to figure out exactly how many dial-up users you can have per T-1, but it is a gamble and changes depending on what speed the users connect at, what speed they sustain, and how much traffic they sustain. I'll show you the equation, but your best bet is to look at your MRTG graphs and decide for yourself.
A T-1 can transmit data at 1.536 Megabits per second. This doesn't take in account overhead or packet loss, however. For absolute accuracy, you'll need to calculate the exact packet loss and overhead. I'll leave this to you techs; for purposes of this discussion, I'm going to pretend they don't exist.
If you are only doing dial-up (i.e. no frame customers), you need to figure out how many ports/users it takes to saturate a T-1 during peak times. If you figure on users actually using their full bandwidth (let's assume average throughput of 33Kbps) then you would get 46 users/ T-1/ second.
The problem with the equation is that they do not actually sustain that bandwidth. If the users average using the full 33Kbps 10 minutes out of an hour, you get a 6:1 bandwidth ratio. That gives you 6 users per hour of full usage. 46 streams times 6 users per hour gives you 276 users using full streams per hour. Since there are 4 peak hours per day, and 276 users using full streams per hour, 1104 users can have full streams in every peak period. Since there are 30 peak periods per month, and 30 days in the month, the cost for the T-1 per month is split among 1104 users. Figuring a cost of $2,000/month for a T-1 (including local loop) then each user is responsible for $1.82 of that T-1 per month.
Here' the math:
1,536 Mbps * 1,000 = 1,536 Kbps.
1,536 Kbps / 33Kbps = 46 concurrent 33Kbps streams/second.
60 minutes/hour / 10 minutes of full usage/user/hour = 6 users/hour of full usage.
46 streams * 6 users/hour = 276 users/hour/T-1
4 peak hours * 276 users/hour/T-1 = 1104 users/peak period
1104 users/peak period * 30 peak periods/month / 30 days/month = 1104 users
$2000/month/T-1 /1104 users = $1.82/month/user
So, using our assumptions, your cost for PPP access is $1.82 per user per month.(This does not include setup costs or maintenance.)
Pricing Your Services ?? Part 2 ?-?continued
Ancillary service costs
E-mail accounts and personal web page hosting make up a very small part of the cost of a dial-up user. The amount of bandwidth per user, per e-mail account/web page per month is so negligible that you can consider it paid for in the Internet access cost. (If you want to figure out the exact bandwidth cost, look in the web hosting pricing section of this column.) What is important is the machine the e-mail boxes/web pages are hosted on. If a box costs $3,000, and you consider a 12 month depreciation schedule, and you can put 1000 users on each box, then your cost per month, per user for e-mail and web hosting is
Here is the math:
$3000/server / 12 months = $250/month
$250/month / 1000 users = $0.25/month/user
Once you have the cost for PPP access, and e-mail and web page hosting cost, you need to figure out your support and overhead costs. The cost for support is very hard to calculate because you need to know what percentage of your overhead is attributed to dial-up access.
Support costs
Technical support and customer service are the easiest to figure out. The ratio of users to technical support or customer service staff is a basic business decision, as it affects the quality of your overall service. It also varies depending on what type of service your business is built around (higher ratio for ISP that sell primarily dial access) and, to a lesser extent, your service area.
In order to figure out the cost, you add up the number of man-ours, times the cost per man-hour, plus your overhead per person. In order to find your overhead cost per person (or per desk), add up all your costs for lights, phones, rent, etc. and divide it by the number of employees/desks. This number will be slightly skewed, as some areas (sales and the like) may make more phone calls then others, but will give you an average cost.
Example:
One technical support/customer service representative is taking care of 750 users. The rep is being paid $10/hour and is working 40 hours/week plus benefits. Your total operations cost is $10,000/month, and you have 20 people working.
Here is the math:
$10/hour * 40 hours = $400/week
4.33 weeks/month * $400/week = $1,732/mo.
33% (benefits, taxes, etc.) * $1,732/mo. = $571.56/mo in benefits, taxes, etc.
$571.56/mo in benefits, taxes + $1,732/mo. in salary = $2,303.56/mo in total compensation
$10,000/mo / 20 people = $500/person/month in overhead
$500/mo in overhead + $2,303.56 in total comp = $2,803.56 /mo in total cost for that rep $2,803.56 /mo. /750 users = $3.74/user in support cost
The tricky part comes in figuring out how much of the total management/maintenance cost is allocated to each user. If you have a PPP access cost of $1.82/user/mo, an e-mail and web hosting cost of $0.25/user and a support cost of $3.74/user, your total cost per user, before management costs, would be $5.81. When you add in your average management/maintenance costs, and your 'oh shit' costs, you total cost per user would probably be about $7.
Customer acquision costs
The cost of acquiring new customers?via marketing or purchase?is one you'll probably be acutely aware of. But how to treat these costs in terms of ongoing operating expenses is less obvious.
One simple solution is to set up a depreciation schedule, just as with equipment, and consider it a cost over that period of time. I.e. if you had an average customer acquisition cost of $60, and had a depreciation schedule of 12 months, you would add $5 in cost per user per month.
There is a another school of thought on customer acquisition schedules: AOL at one time had a customer acquisition cost of $160/user. The only way to write off that cost without doing a depreciation schedule?which would make the cost per user much too high?is to treat it as an investment and assume you'll make back the acquisition cost when you sell your ISP. I don't recommend this type of 'creative' accounting, but it has been done, and will probably be done many times again.
Part 1: Dial Access Basics
Part 2: Basic Operations Costs
Part 3: Web Hosting
Part 4: Dedicated Access
Part 3 - Web hosting
Having looked at costing out dial access services and basic operational costs, today, we'll tackle special issue related to Web hosting. The principles are the same, but the details differ.
In costing out Web hosting services, you must factor in both the equipment dedicated to the hosting service (mainly servers), and the cost of the bandwidth consumed by your cuctomers?plus the cost of IP addressed used. [DISCLAIMER: All numbers are chosen for the sake of illustrations, not to accurately reflect current market prices. It is not the numbers that are important, it is the equations. -JZ]
Equipment costs
If a server costs $3,000, and can host 50 virtual domains, the cost per domain/per month is $5?working on a 12 month depreciation schedule.
Here's the math:
$3,000 /12 months = $250/month
$250/month/50 domains = $5/month/domain
This assumes that all domains get equal shares of hard drive space. If you need to charge more for hard drive space, take the cost of the hard drive, divide it by your depreciation schedule, and then divide it by the number of megabytes in the hard drive. You will see the cost is low.
IP address and Bandwidth costs
Actually, IP addresses are so cheap that the cost for a single IP per month is usually in the penny range, so it's really a negligible cost. Not so for bandwidth.
There are two common approaches to calculating bandwidth costs:
In smaller situations, bandwidth is typically figured per GB (GigaByte) of data transferred. In larger situations, especially with dedicated servers, bandwidth is typically charged per Mbps (Megabits per second) sustained?the way you are billed. Web data, by nature, is a bursty medium. Figuring in GB of data transferred makes it much harder to account for bursts. The Mbps method makes it easier to charge for bursts.
GigaBytes-transferred costing
The first thing you need to figure out is the total GBs of data a single T1 can transfer in a given time period?assuming no overhead or packet loss. (If you have stats for your packet loss or overhead, you can subtract the overhead from the amount of data that can be transferred and continue with the equation.)
A T1 transmits data at 1.536 Mbps (Megabits per second). To translate that to Bytes per second you divide by 8 (since there are 8 bits in every byte), and you find that a T1 can transfer data at 192 KBps. Multiply that by 60 seconds in a minute, times 60 minutes in an hour, and you will find that a T1 can transfer 691.2 MB of data per hour, or 6.589 GB per day. Assuming that your web traffic is evenly spread over a day(which in most hosting situations it isn't, of course) your cost per GB of data transferred?given T1 costs of $2000/month?is just over $4.
Here's the math:
1.536 Mbps /8 = 192KBps
192 KBps * 60 seconds/minute = 11.52 MB/minute
11.52 MBps * 60 minutes/hour = 691.2 MB/hour
691.2 MB/hour * 24 hours/day = 16.589 GB/day
16.589 GB/day * 30 days/month =497.67 GB/month
$2000/month/T1 / 497.67 GB/month = $4.02/GB of data transferred.
Pricing Your Services ?? Part 3 ?-?continued
(GigaBytes-transferred costing, continued)
In most cases, you will have peak times of web page viewing. If your curve is very steep, you need to keep this in mind and adjust your cost calculations appropriately. If, for example, most of your traffic were concentrated in a peak period of 4 hours, then your cost would be a much higher $24/GB of data transferred.
Here's the math:
691.2 MB/hour x 4 peak hours/day =2.765 GB/peak period/day
2.765 GB/peak period x 30 days = 82.944 GB/Month
$2000/month/T1 / 82.944 GB/Month = $24/GB of data transferred.
Mbps costing
Mbps is usually calculated by the 95th percentile rule. Using a tool, such as MRTG (Multi Router Traffic Grapher), readings of bandwidth utilization are polled every 5 minutes. Polling is easier to do with dedicated servers/ports, but can be done by IP address. All of the samples that were pulled throughout the month are then ordered from highest to lowest. The top 5 percent of readings are then discarded, and the customer is billed based on the next highest reading.
A word of caution::
There are 8,640 5-minute samples in an average month. The 95th percentile rule allows users to peak (at whatever the limitation you set is) for 432 5-minute periods, or 2160 minutes (36 hours) per month. You need to have a network design and pricing plan to accommodate these peaks. It is also possible to do 90th percentile or lower, but if you make your percentile too low, and don't put a cap on the peaks, users will get enough time and free bandwidth to cover all of their peaks.
Mbps costing is usually done with a committed rate and a burst rate. The burst rate is usually a $20-25 increase over the committed rate. If your T1 costs $2000/month including local loop (always include loop charges), then your Mbps cost is $1,302.08/month. So, how do you make money? By oversubscribing your pipes. With oversubscription, you are betting that even though someone is committing to 1 Mbps, they will not use the full 1 Mbps continuously. If they do, you'll loose money. That is why some companies charge a resell rate for people they think will actually use the bandwidth. The typical oversubscription rate is 4 times, which?using our arbitrary numbers?makes your cost per Mbps $325.50/month.
Here's the math:
$2,000/month/T1 / 1.536 Mbps/T1 = $1,302.08/month/Mbps
$1,302.08/month/Mbps / 4x oversubscription rate = $325.50/month/Mbps (oversubscribed)
In the end, remember, let your costs determine your pricing?before you let your competition determine it.
Part 1: Dial Access Basics
Part 2: Basic Operations Costs
Part 3: Web Hosting
Part 4: Dedicated Access
Part 4 - Dedicated Access
Last week we discussed costing issues for Web hosting services ?a major component of many ISP's business. Today, we'll tackle the last of our three service categories: dedicated access.
In order to determine the cost of dedicated leased lines,you first have to make a distinction between point-to-point (PTP) and frame (or SMDS, ATM, or whatever non-PTP line you may offer).
The key in making money in leased line connectivity is oversubscription. The principle of oversubscription is simple: It's a gamble that not all of the leased-line customers want full connectivity at the same time. At what rate you choose to oversubscribe your bandwidth at is your choice; this may change, depending on whether the connectivity is PTP or frame.
Customers who purchase PTP connectivity tend to require more bandwidth. This may force you to operate at an oversubscription rate as low as a 2:1, therefore raising your costs?and your price.
Note: If you sell to ISPs, web hosting, or other resellers of bandwidth, they are likely to fall into this category, and it is normal to charge them a higher rate. Frame relay customers tend to use less bandwidth. With most frame connections, the customer has a committed information rate (CIR) and burst rate. In most cases the CIR is one-half of the burst rate (for example, 128k CIR on a 256k line). This constitutes a built-in oversubscription model for you, allowing you to oversubscribe at least 4:1.
Assuming a 4:1 oversubscription rate, and a cost of $2,000 (including local loop, as always), then your bandwidth cost per T-1 would be $500.
Here's the math:
$2,000 / 4:1 over subscription rate = $500/T-1 in bandwidth cost.
Equipment cost.
If you have a router dedicated to lease lines, then determining the cost is relatively easy. (I will assume this. If it is shared, then you need to determine what proportion of the router goes towards leased line, and what part to your own capacity.)
If you use a router that cost $2,500 and has 2 serial ports (one for connectivity in, one for outbound frame connectivity to your customers) and your 2 CSU/DSUs (one for each serial port) cost $500 each, then your total equipment cost would be $3,500. Assuming a moderately conservative 24 month depreciation schedule, your monthly cost for that router is around $145. If you serve 4 T-1 connections off of that router, your monthly equipment cost per T-1 would be $36.25.
Here's the math:
$2,500 (for the router) + $1,000 (for the CSU/DSUs) = $3,500 (total equipment cost)
$3,500 / 24 months = ~$145/month in equipment cost
$145/month / 4 T-1s = $36.25 (equipment cost per T-1 per month)
More time consuming is determining the cost of maintenance, power and A/C for the router. To accurately determine the cost, you need to track the time (in man-hours) devoted to maintaining the router and multiply it by the cost per man hour. (Found by adding up the salary, benefits, taxes, etc of the employee, and dividing by the number of hours they work.) For power, you can actually find out what the amperage usage of the router is, and multiply it by the kwh you pay per month, and you will have your cost.
Part 1: Dial Access Basics
Part 2: Basic Operations Costs
Part 3: Web Hosting
Part 4: Dedicated Access
No comments:
Post a Comment