Philadelphia wireless project — I can't make the numbers work

Here's my daily dose of rants and puzzlements. Today's revolves around the recently-announced wireless project in Philadelphia. Like lots of folks, I'm hoping they'll make it go. Like lots of folks, I can't quite figure out how they're going to make the numbers work. Here's why.

Get an envelope out, we gonna do some figuring on the back of it. Let's see here, the Philly folks say they're going to light up the whole town for free Internet access. Ok, great. Presume all the geek problems away and let's say they get it lit. Now, if they really finally run with “free” they've basically replaced all the ISPs in town. Well, maybe that's ok. After all maybe this is an amenity that cities should do, like roads and sewers and stuff. Here's where the numbers get hard…

If residential Internet access is free, then most everybody will switch to it. Some significant percentage of 1,000,000 households. First “numbers” problem — how much is upstream access for all those folks going to cost??

Get your envelope out. Let's say we give everybody up to 3 mBits (kinda like cable). And let's assume that people aren't all going to use it at the same time. That's called “oversubscription” and everybody does it – ISPs, phone companies, you name it. What oversubscription ratio? For the sake of the envelope, say 30-times. So for every megabit of upstream access, we can sell 30 mBits of “downstream” or customer access. Or 10 customers (remember? 3 mBits/customer). So for every 10 customers, we need another megabit of upstream access for those peak times (after supper).

If most of our 1,000,000 households switch to our free service, we gonna need a *lot* of upstream access. Let's say 20% of the households sign up. 200,000 customers means we have to buy 20,000 MBits of upstream access. 20 GigaBits! Wow! A T1 line is 1.5 MBits, so that would be 13,000 T1's. That a lot. Ok, Internet access is overbuilt and getting really cheap, so maybe they've got a really good deal. Maybe $10/month per MBit (that's cheaper than anything I've heard of, but hey, benefit of the doubt that's my motto…). $200,000/month for upstream. $2,400,000 a year of taxpayer money…

Once the customers have signed up for Internet, any time anything goes wrong with their computer that they can't understand, who they gonna call? The ISP, that's who. So now Philly's got to provide help-desk support for 200,000 customers. Let's say each customer calls once a year with a hard question that takes an hour to answer. 200,000 hours a year. People work 2000 hours a year, so that's 100 people answering help-desk calls at, say $50,000/year. Hmm, $5,000,000/year…

See where this is heading? I could run through all the rest of the stuff that an ISP does — network maintenance, upgrades, fixing people's connections, paying the bills, etc. Roll all that up in a ball and it looks to me like you've got at least $10-15 million/year to keep it running.

On the one hand, that's a lot of money. On the other hand, compared to losing money on ballparks and stadiums and roads and sewers and economic incentives for people to move in, maybe it's not. I'd love to see the numbers for that project.