Also just a note for those curious: this is how I like to measure what's needed for a WISP, and track how it changes over time.
Set up a 1gbps connection and start adding customers. Watch the usage on the circuit as your customer base grows. You want to make sure that any time a customer comes online, they can get their full speed according to the package you're selling. So if you're selling 100mbps packages, you always want to have at least 100mbps available on that circuit, even at peak times of day. As you add customers, you can start to estimate how much traffic, on average, each customer adds during the peak time of day (it's not as much as you'd expect.) Then you can estimate how many customers you'll have before you approach your limit (1gbps minus max speed package.)
You have to monitor peak utilization of the network on the order of seconds to minutes. The rule of thumb from queuing theory is that network performance tends to tank when capacity utilization hits about 80% => a 1 Gbps link is only good if your max utilization is 800 Mbps. But if your monitoring only looks at 1 hour averages over the course of a month, you're going to miss the periods of seconds where everyone's getting ping times of >100ms and having their VoIP calls drop out.
There are a lot of things you'll learn along the way if you try to run an ISP. =-)
Set up a 1gbps connection and start adding customers. Watch the usage on the circuit as your customer base grows. You want to make sure that any time a customer comes online, they can get their full speed according to the package you're selling. So if you're selling 100mbps packages, you always want to have at least 100mbps available on that circuit, even at peak times of day. As you add customers, you can start to estimate how much traffic, on average, each customer adds during the peak time of day (it's not as much as you'd expect.) Then you can estimate how many customers you'll have before you approach your limit (1gbps minus max speed package.)