I have only thoroughly digested the proposed BEAD grant rules for about a third of the states, but just in the sample I’ve analyzed, two states are going to use a cost model as part of the grant review process – Arizona and Missouri. I have to imagine there are others.

A cost model is just what it sounds like. There are consultants in the country who have built complex models that are supposed to predict the cost of building broadband anywhere in the country. The models have to be loaded with specific inputs for any given location, and the models are then supposed to calculate what it will cost to build a broadband network.

As somebody who has helped to do the hard work of estimating the cost of constructing broadband networks for several decades, I can tell you unequivocally that there are no cost models that work – absolutely none. The cost of building a broadband network invariably relies on local factors and nuances that a cost model can never capture. A model might do a decent job guessing the cost of building broadband in the open prairies of the Midwest where farms are far apart, but even there the models are going to whiff sometimes.

Consider some of the reasons why cost models will never be accurate:

• For aerial fiber construction, a model can never predict the percentage of poles that must be replaced or that need major make-ready work. I’ve worked on projects just in the last few years where the percentage of bad poles varied between 10% and 90%. The models can also never model the ease or difficulty of working with a given pole owner. An uncooperative pole owner can add a mountain of time to a construction project – and time is money.
• A cost model is going to miss the unpredictable issues that can add big delays and significant cost. This might be rights-of-ways that are hard to obtain, especially for bridges, railroads, interstate underpasses, wetlands, and federal or state land. The cost model is not going to predict that the local folks in some counties are land-right advocates, and getting easements will be an uphill battle.
• A cost model is not going to predict the need for an environmental study – or even if it does, it won’t know if the study will be easy and cheap or complicated and extravagantly expensive.
• Cost models are not going to properly predict the prevailing wages that are needed for the BEAD grants. Most state prevailing wage tables do not specifically list all of the various job functions needed to build a fiber network. I’ve spent several days on a single project figuring this out.
• Cost models are not going to know that existing poles are now fifteen feet inside the woods that have grown since the poles were built. A cost model is not going to know that a County Highway Department has an edict against putting fiber in drainage ditches. A cost model is not going to understand that some mountain and rural roads have zero shoulders on both sides of the road.
• A cost model probably does not include the most current costs of both materials and labor for building fiber. Engineers who estimate the cost of fiber projects for a living have been updating their costs almost weekly for the last several years due to supply chain issues and inflation.
• No cost model is going to understand the extra cost of buying materials that comply with the Build American rule – because nobody knows that yet.
• I could easily list a dozen more such issues, but I don’t want the blog to get too long.

These two states are using the cost models in the worst possible way because they are using the costs suggested by the cost models to help pick grant winners. That means they are accepting the results from the cost models as a real metric and will reject grant proposals that are above or not within a certain range of the cost model estimate.

Cost models have a useful function in a grant office, which is to double-check construction estimates made by ISPs. I’ve known grant offices that have done this for years as a way to perform a sanity check on proposed grant costs. But I don’t see that process working for BEAD. Most of the States have said that the grant review process is going to go very quickly, meaning there won’t be time for a real analysis of an ISP’s proposal. That alone is very disturbing. State Broadband Offices are clearly trying to push the grant awards out the door with as little analysis as possible – and using a cost model can save a lot of time and effort. Real analysis requires industry experts with long-time experience who really knows if an ISP is exaggerating costs – something that many SBOs are unfortunately lacking.

I have no idea how these two states or others came to the decision to use a cost model. I have to assume some consultant sold them on the idea that they have a highly accurate cost model. But there is no such thing as an accurate cost model, and there never will be.

I fear that states that use cost models are not going to get the results they are hoping for. In places where the cost model predicts a higher cost than actual for a project, ISPs will gladly go along with the higher cost model number and ask for more money. ISPs simply won’t seek grants in areas where cost models predict a lower cost than what can be constructed. If the cost model suggested costs are too low – nobody might ask for a grant.

I give ISPs in states with cost models the same advice I am giving everywhere. Ask the the money you need to make a grant work, and nothing lower. At the end of the day, the States must award the grant to somebody, and your grant offer might be high and still be the best offer on the table. I have a hard time advising anybody to accept less than you need, because that is a path to financial disaster.