ISPs and communities are always asking me for metrics to help them estimate how much grant funding they might need to subsidize building a new broadband network. Unfortunately, there is no such metric because broadband costs are always unique to a given community.

One of the statistics they want to give me is the population density – the number of homes and businesses per square mile. They are often surprised when I tell them that number alone tells me almost nothing about broadband costs. It’s far more important to know where homes are located in relation to roads.

Let me give two easily contrasting examples. One is a rural area made up of rather large family farms engaged in growing row crops. Generally, the farm and a few buildings sit somewhere in the center of the farm, and there may be a few more homes added over the years along roads. From the perspective of building fiber, this is about the worst situation you can find. The cost of fiber per customer is extremely high when neighbors can’t see each other. This situation gets particularly expensive in a county with a square grid road system where there are at least a few homes on every road – that adds even more cost.

Contrast this with places that look on paper to have a lower population density. For example, some of the rural counties in New Mexico are sparely populated compared to a farming community in the Midwest when looking at people per square mile. But in the rugged terrain of New Mexico, the houses and business test to be relatively close to the handful of roads that traverse a county. While there aren’t a lot of homes, there also aren’t a lot of miles of roads. The number of homes per mile of fiber construction might be much higher in New Mexico than a Midwest farm community, even with far fewer people.

But there are still nuances that matter. Another issue is how far the homes are from the roads. In some of the mountain valleys of Pennsylvania, homes tend to be close to the country roads since the hills rise up immediately behind the homes. I looked at the cost of building a county in West Virginia and another in Minnesota where a large percentage of homes are far back long lanes off the handful of main roads. The amount of fiber per customer needed in the Pennsylvania example is far less than the other two cases.

While the relationship between homes to roads is a major factor, it’s not the only one. There are some counties where the cost to build is higher than expected for other reasons. I was looking at a county in Colorado where the existing utility poles were in dreadful shape, with more than half needing replacement. In the same county, there is hard rock right up to the surface of the soil. In this particular county there is no affordable way to build fiber. When people talk about places that need an alternative to fiber, they are talking about situations like this. This contrasts significantly from places with deep topsoil, like the farmlands in Minnesota and Iowa or the farming valleys of California. I’ve always joked that in some of these places, you can bury fiber with a tablespoon.

Most places in the country are not as extreme as the above examples. A more typical circumstance is where a county has a mix of different construction situations. For example, there might be a lot of farmland, but also many homes located along rivers where it’s rocky and much more expensive to build. It’s not unusual to find pockets of good and bad poles in most counties, meaning that any estimate of cost means looking everywhere first to understand the places that will be more expensive to build. I worked in one county that seemingly had water everywhere, with numerous small streams, lakes and ponds, and wetlands. Building a mile of fiber at any one place didn’t seem too expensive, but the cost of crossing the bridges and of avoiding the wetlands makes this whole county a challenge for building an extensive fiber network.

I know folks want quick and easy answers, but I refuse to give back-of-the-envelope construction cost estimates until I know an area well. I’ve seen too many places where a quick estimate would have been way off from the actual cost. There are consultants around who have generic models that will provide a quick cost estimate – but until an engineer has put eyes on the local situation, such quick estimates might not be the paper they are written on.