Hacker News Reader

Comment by ApolloFortyNine | Hacker News Reader

A WISP, at scale (50 users or more), should have a cost per user of about $10/m. You can oversell bandwidth considerably, more than most people realize. I've specifically seen an 80 customer wisp reporting their peak time is only 230mbps. This largely makes sense as you people streaming video is going to be your biggest usage, and that really only average 15mbps at the worse. A $1/mbit for ISP bandwidth should be doable in most areas, and since it's wireless your costs per user are actually rather low, given a good tower.

But WISPs do make the most sense in areas where there is no good alternative (only DSL or Dish).

View on HN

Yeah, the big problem is labor. if you are only making a few bucks a user, making enough money to make up for your opportunity cost (vs. working as a tech for a large company) requires a lot of users.

I've got a lot of experience serving the low-margin market, both for ISPs, WISPs, but mostly on the hosting side, (and hosting is actually easier; the less you have the customer on the equipment, the less time you spend fixing stuff they broke.) and let me tell you, low margin does not at all line up with low support expectations.

Even if you can charge your customer twice what it's costing you and making ten bucks a month per user, one tech support issue can end up eating all the margin that customer would have given you for years.

I'm certainly not saying to price it so you only make $10/m per user. I'm simply stating what your base cost is. $50/m should be expected with a WISP, as I tend to see rates of 50mbps+ from well run WISPs. I was just pointing out that it's hard to price someone out of market without dropping prices to ridiculous rates.

I'd love to know where you can get a fiber circuit for $500/m. (And even if you can, then you probably can't compete with whoever's selling you that circuit anyway.) When we started we haggled considerably.. (had to sign a 20 year deal) and managed to get a 2G/10G circuit for $1,500/m. (A 1G circuit would have been $1200/m)

Some places are cheaper than others, looking at this thread especially outside US.

In Poland you can get 1Gbps connection for 200USD/month and 5Gbps for ~600USD/m

These are not fiber circuits, ie dedicated bandwidth, they are shared residential services.

Do you think there's a company advertising 5gbps at $600/m for residential services? Those are dedicated fiber prices, all be it very good ones (I'd question the peering outside of Europe).

There's free.fr that sells 10G for 50EUR/month.

Those are not shared residential services, but normal business/ISP grade connection with SLA and so on.

I guarantee it's not dedicated. it may go over a dedicated line at some point, but I'm in the layer2 business running GMPLS and shared switched traffic and you're either on GPON or some Metro Ethernet gear. you are sharing the neighborhood bandwidth on a metro switch, or you're sharing both light spectrum (GPON) AND the neighborhood bandwidth. you could have 100g but it's still shared.

2G/10G? Normally I see it considerably cheaper for download bandwidth vs upload bandwidth, as your provider will just sell/reserve the upload bandwidth for servers, since servers use vastly more upload.

Of course, the rate you gave holds with the $1/mb I was assuming. You oversell bandwidth obviously.

It's probably 2G sustained, 10G burst of some sort.

>> You can oversell bandwidth considerably

Not as much as Comcast/$ESTABLISHED company can. The people most likely to switch to a WISP or alternative ISP are the people most likely to crush your bandwidth.

> A WISP, at scale (50 users or more), should have a cost per user of about $10/m. You can oversell bandwidth considerably, more than most people realize.

I run a small non-profit WISP (with price per user about $15/m), and the cost of bandwidth is not a significant part of the budget (about 15 %). Biggest part of the budget (about 60 %) is hardware and work for development and repairs of the network.

> since it's wireless your costs per user are actually rather low, given a good tower.

I would say that wireless has pretty high cost per user. To have good coverage and not overloaded APs, you target ~5 users per AP and 1-3 APs per node.

Reply on news.ycombinator.com