The great BEAD monitoring challenge

Washington, DC (January 27, 2025) - Over the next 12 to 18 months, the vast majority of the $40 billion or in broadband deployment funding that U.S. Congress enacted in the 2021 Infrastructure Investment and Jobs Act will have been allocated to specific internet service providers (ISPs) for specific projects. Unless something goes wrong, all the roughly 6 million to 7 million BSLs targeted by the Broadband, Equity, Access, and Deployment (BEAD) program will be included in some BEAD project.

Early results confirm what has been long anticipated — that most of the subsidized new coverage will consist of end-to-end fiber projects. But in most states, some fixed wireless and/or LEO satellites will be in the mix as well. State broadband offices will be able to take pride in their achievements as their Final Proposals get submitted and approved.

But the hardest part of the work will still lie ahead. For ISPs, that's the actual construction and operation of the broadband networks that will deliver universal broadband coverage throughout America. For state broadband offices, that's paying for and monitoring the coverage.

State broadband offices need to distribute the BEAD funds, not all at once but as reimbursements or milestone-based fixed award subamounts. Then, they need to make sure the promised networks actually get built on time and to specs. Or, if not — and you can’t expect 100% success on the first try — states need to apply appropriate accountability and skillfully correct course.

How monitoring and accountability make subsidies work

A government agency shouldn't take for granted that just because it gives a company money to build something, the something gets built. There needs to be accountability. As the saying goes: “Trust but verify.” There are a wide variety of scenarios, ranging from the completely excusable to the criminally fraudulent, in which actions and accomplishments ex post won't align perfectly with plans and promises ex ante. Good grants administration means finding out about them and handling them with skill, fairness, and transparency.

It's important to think about incentives. Sometimes it’s just not feasible for a company to fulfill its promises, maybe for reasons that couldn’t have been foreseen, or maybe for reasons that should have been weren’t. In other cases, it's feasible, but it's not in the company’s interests to do so.

At an earlier stage, it can be in a company's interest to overpromise, if the anticipated accountability for keeping promises is weak. In the worst case, pathological situations can arise in which everyone overpromises because that's the only way to compete, and then everyone deceives to obfuscate the unfulfillment of their overpromising. Then again, there are cases where changes are in the public interest, as when a grant recipient alters plans to integrate new technology that improves performance.

Robust accountability for grant performance doesn't mean that all promises are kept, or even that promises are always kept when it's feasible to do so. Rather, it means that the majority of promises are mostly kept, that intentional deceit is kept to a minimum, it’s detected and punished when it occurs, and that when plans change, the changes are negotiated, the grantmaker has a seat at the table to represent the public interest, and good data is available to inform the negotiations.

States should plan ahead for robust BEAD accountability. Accountability should be tough enough to deter overpromising without being too burdensome for subgrantee ISPs. If an ISP declines an award because it's afraid of the paperwork and red tape of compliance, that's a failure. Worse yet would be a case where ISPs that competed for BEAD grants decide, upon seeing the state’s monitoring plan, to decline awards precisely because they think the monitoring method is invalid and fear that it will result in adverse determinations and clawbacks even when they have, in fact, fully executed on their obligations.

Accountability should be tough but also fair, and it's probably better to let a few subpar networks pass than to punish, based on erroneous data, an ISP that did the job right. But the point of a good monitoring plan is not to have to make that choice, because you really know. If your monitoring plans are robust, and an ISP declines an award precisely because they were bluffing and never really planned to build the network to the specifications they had declared, and the state’s BEAD monitoring plans convince them they can’t get away with that, then you've nipped a problem in the bud.

Now for a key point. As part of BEAD monitoring best practice, Connected Nation recommends that all states should roll trucks and lay eyes on the stuff that gets built. Not all of it. Probably not even most. Field validation is expensive. The temptation will be to buy fancy tech solutions that get sold as alternatives to boots on the ground and skip field validation. That’s asking for trouble.

Innovative data collection may well have an important role to play in BEAD monitoring. But internet infrastructure is physical, material stuff, and ultimately, physical verification that that material stuff is present where it needs to be is data for which there is no substitute. A knowledgeable field validator with binoculars in hand can learn things that a remote speed test engineer can never know. Good field validation is an irreplaceable component of the epistemic journey to justified confidence that the BEAD-funded broadband networks are really getting built.

Bear in mind, too, that innovation cuts both ways. Ever-improving AI will make it easier to produce fake electronic records that seem to prove the deployments got done. The seemingly low-tech art of looking at stuff in the field and figuring out what it is, is a good safeguard against new forms of tech-empowered fraud.

Fortunately, field validation isn't that expensive. NTIA guidance has clarified that monitoring subgrantees is not one of the expenses that are subject to the 2% administrative cap. States can spend what they need to roll trucks and lay eyes on the infrastructure they build. Skillful use of a variety of data, including provider-reported and public-facing, should be used to decide where to roll the trucks to. But don't omit getting out there and looking at the stuff.

BEAD Final Proposals will need to Include a BEAD Program Monitoring Plan

The BEAD program has a federalist design, with states operating under oversight by the National Telecommunications and Information Administration (NTIA) but still enjoying substantial leeway to run the program in a way that suits each state’s particular circumstances and each broadband office’s capacities. While enjoying some autonomy, states must operate within the framework of NTIA guidance, starting with the May 13, 2022, BEAD NOFO, the first and most authoritative component of the NTIA’s BEAD rulemaking.

The BEAD NOFO specifies that “each Eligible Entity [state or territory] must, at a minimum, employ … (4) robust subgrantee monitoring practices” (p. 51), and adds later that “Eligible Entities shall … develop monitoring plans, subject to the approval of the Assistant Secretary [of the NTIA], which may include site visits or desk reviews, technical assistance, and random sampling of compliance requirements” (p. 96). Published well before state BEAD planning began, the BEAD NOFO left it to the future to define what “robust” monitoring meant, and how NTIA approval would take place. But even the NOFO makes clear the expectation of random sampling.

Two and a half years after the NOFO, in November 2024, the NTIA released BEAD Final Proposal Guidance. This guidance indicates that states will be required to submit, as an attachment to their Final Proposals, a BEAD Program Monitoring Plan. However, the guidance indicates that:

The Eligible Entity should refer to forthcoming NTIA guidance on completing its BEAD Program Monitoring Plan. (p. 27) 

The early bird states of Louisiana, Delaware, and Nevada, which have already submitted BEAD Final Proposals and got them approved, had to develop BEAD Program Monitoring Plans without the benefit of this forthcoming guidance. Other states with imminent Final Proposal deadlines should consider following suit and at least beginning to develop BEAD monitoring plans based on the guidance now available rather than waiting, since the guidance may come after their Final Proposals or too late to be effectively integrated into their BEAD monitoring plans. Of course, states may need to adjust their monitoring plans to conform to the guidance if it gets released before their Final Proposals get approved.

The Final Proposal Guidance gives some further direction on what subgrantee monitoring should accomplish:

Examples of details in a BEAD Program Monitoring Plan include, but are not limited to:

• • Detailing how subgrantees will comply with the terms and conditions of the award
  •  Detailing how the Eligible Entity will ensure subgrantees implement projects on schedule and make adequate progress toward achieving identified metrics, milestones, goals, objectives, and planned outcomes;
  • Detailing how the Eligible Entity will ensure subgrantees meet financial and programmatic reporting requirements
  • Detailing how the Eligible Entity will ensure subgrantees expend federal funds
  • Detailing how the Eligible Entity will monitor subgrantees to ensure the project is completed on time, and limit potential waste, fraud, and abuse of federal funding

This high-level language relates more to what subgrantee monitoring needs to achieve than how it can be done. State broadband offices’ subgrantee monitoring activities will need to suffice to ensure on-time implementation and continuing progress, achievement of identified metrics, required financial and programmatic reporting, and enforcement of special award conditions (e.g., about speeds available, or pricing). But the guidance does little to suggest activities by which state broadband offices can satisfy these requirements or to indicate how the NTIA will judge whether state plans are sufficient.

Among states with approved Final Proposals, Louisiana’s 34-page monitoring plan is the most detailed, and describes how the state will assess risk subgrantee by subgrantee and project by project, both ex ante and on an ongoing basis during construction, then review invoices and other documentation and conduct virtual meetings and site visits, culminating in an exit conference when construction is complete.

The monitoring plans from Delaware (six pages) and Nevada (three pages) are more concise but include similar elements. All the plans leave much to be worked out in the execution. In particular, none of them detail the site-visit process to the point of explaining how a first-hand observer will verify whether the claimed broadband infrastructure physically exists, including proactive or reactive verification of the existence of the locations covered by the awards. These Final Proposals may be helpful for other states to consult as a source of ideas, but they may also be superseded by forthcoming guidance, as appears to have happened already with respect to alternative technology.

At the heart of any robust yet cost-effective monitoring approach is, as the NOFO originally suggested, random sampling. While BEAD monitoring best practice will be further defined through additional guidance, Final Proposal postings and approvals, and states’ activities and lessons learned, it can be expected to revolve around this concept.

Risk-weighted random sampling is the key to efficient monitoring

Risk-weighted random sampling sounds intimidating, so let's start with a simple example.

Suppose you're in the kitchen, and you want to find out whether the food you’re cooking is done. Maybe you're making soup, and you want to see if it's hot enough. You don't need to try it all. Just take a spoonful. If you've been stirring, the temperature should be consistent throughout. If the spoonful is nice and hot, dinner is served!

Now suppose you're making a stir fry, with lots of different kinds of vegetables. This time, one bite might not be enough. You might want to try an onion and a piece of broccoli and a red pepper. Just because one kind of vegetable is cooked to perfection doesn't mean they all are. Cook times differ.

Finally, suppose your stir fry has meat in it. You'll probably be sure to sample the meat before you serve it. You might put it on a cutting board and slice through the middle or use a meat thermometer. After all, no one gets sick from eating undercooked broccoli, but undercooked meat can be dangerous.

These are all examples of risk-weighted random sampling. With the soup, one random spoonful sufficed. With the vegetarian stir fry, there were stratified risks by vegetable type, so the sampling method of trying a piece of each vegetable was adapted to the structure of the risk. And with meat in the stir fry, a new kind of risk was introduced, so a more costly but robust data collection methodology was applied.

As state broadband offices cook up their BEAD monitoring solutions, they should adopt their own strategies of risk-weighted random sampling. For example, different technologies will require different kinds of data and different methods of drawing inferences from it. Or again, different companies will merit different levels and timing of scrutiny, with those that might escape clawbacks through bankruptcy if deemed noncompliant perhaps meriting earlier and closer scrutiny than those that have strong track records and robust financials, with deep pockets to pay clawbacks ex post if they are found to underperform.

A good chef is also savvy to the difference between what statisticians call “false negatives” and “false positives” and their implications for the balance of risks. In the kitchen, you might think the chicken in your stir fry is cooked when it turns white and firm on the outside, when really, it’s still pink and raw in the middle. False positive. Or you might think your green beans are still raw because they look vibrant and tight, when really, they’re soft and ready to eat. False negative.

A good cook needs to know the false positives and the false negatives, and to use a meat thermometer to make sure the chicken is cooked through, while using a fork to check that the green beans are soft rather than waiting for them to overcook and turn visibly wilted.

State broadband offices, too, will need to be savvy to both false positives and false negatives. Invoices can be a false positive. An invoice might show that an ISP purchased a lot of fiber-optic cable. Good! But that doesn’t mean the fiber has been installed, and it might never get installed in the right place.

Speed tests can be a false negative. Just because there are no speed tests from an address doesn't mean that broadband hasn't been deployed there, and slow speed tests might occur even when a top-notch network is in place due to weak customer side equipment, too many devices, being too far from the Wi-Fi, or subscribing to a discount package that doesn't offer the maximum speeds the infrastructure is capable of.

Data sources that are prone to false positives and false negatives should not be discarded but supplemented. They can often provide good service as part of a larger monitoring strategy. A good monitoring strategy can derive informational value from a wide variety of evidence, while working around the limitations of each evidence type. In particular, provider-reported, and/or public-facing evidence, which is cheap to acquire, can be used to create and update risk ratings, which can be used to determine the probability with which locations will be targeted for costlier data collection activities.

A good strategy of risk-weighted random sampling can limit the costliest monitoring activities to a fraction of the affected locations, while at the same time ensuring that every subgrantee knows that every claimed location might be checked, so they’d better build to specs and report accurately.

Desktop research

Quite a bit of BEAD monitoring can be accomplished from a computer.

First, states will require reporting from subgrantees. ISPs that receive BEAD grants will produce a lot of electronic documentation for state broadband offices, which will contain a lot of information about the progress of construction.

What subgrantees will report may vary by state. For example, states that pay on a reimbursement basis will need to be especially thorough in collecting and scrutinizing invoices as a basis for making payments, whereas states that pay on a basis of fixed amount subawards can go lighter on invoices but need robust monitoring when milestones are achieved.

States could require photos and/or shapefiles, or accept and use photos and/or shapefiles as evidence without requiring them. They can develop “proofs of passing” modules for subgrantees to perform as part of project closure. These are all forms of data that subgrantee ISPs will provide, but again, it shouldn’t simply be accepted at face value. Trust, but verify.

Second, public-facing data might be used in BEAD monitoring. Provider websites need to have some information about coverage, so that customers who might want to order service can find out whether it’s available. Depending on their level of sophistication, the personnel of state broadband offices and their contractors may be able to query websites about broadband availability at specific addresses in the same way that customers themselves do.

An advantage of public website information is that providers have a commercial stake in its accuracy. They run a reputational risk that erroneous coverage claims might alienate potential customers. Another resource is the FCC National Broadband Map, which semiannually collects coverage data from providers and compiles it into a national map of coverage. If BEAD subgrantees build the promised networks and offer the promised coverage, that coverage should in due course appear in the FCC maps.

Third, communities and residents know something about local broadband deployment and can be tapped for information by state broadband offices. A consumer complaints portal, for example, can enable citizens who are supposed to get broadband coverage through the BEAD program to report on whether deployment has taken place and complain if it doesn’t. Mayors and county officials can be asked to monitor BEAD deployments and given opportunities to alert state broadband offices if promised deployment fails to occur. Local focus groups or surveys can be organized.

Fourth, speed tests can play an important role in BEAD monitoring if used carefully. A favorable speed test result is a pretty reliable indicator that broadband is available at a given address.

All these desk-accessible data sources have their limitations. Data that is self-reported by providers is only as good as the providers are capable and honest. Accountability must mean more than just taking subgrantees at their word.

Not all ISPs’ websites display address-level coverage information to unidentified visitors, and mass data collection from provider websites by state broadband offices might look like a hack attack and trigger security alerts. It may also be morally inappropriate for state broadband offices personnel or contractors to impersonate potential customers.

The FCC National Broadband Map is fundamentally self-reported and is also subject to long lags between when coverage is first available and when it appears on the FCC map. Communities and residents are hard to mobilize and may be ill-informed. As mentioned above, speed test data is prone to false negatives.

For all these reasons, state broadband offices’ monitoring strategies should not be limited to the desktop but should involve collecting data in the field as well.

Paying promptly but with confidence

Throughout the BEAD monitoring process, the goal should not only be accurate accountability but also prompt payment. Making sure that subgrantees are really building shouldn't get in the way of paying them promptly when payments are due.

A good chef knows that sometimes timing is critical. If you’re making a French soufflé, you need to put it in the oven immediately after you fold the whipped egg whites into the soufflé base, or else the air in the egg whites will start to deflate, and the soufflé won’t rise properly.

In the same way, when subgrantees submit invoices quarterly, or submit evidence of milestone completion, they will expect to be paid. A certain due diligence is needed in checking the documentation, but long delays could cause cash-flow problems for some subgrantees. In the worst cases, subgrantees might slow down or stop work, either because they want to limit their exposure to nonpayment risk, or simply because they lack the cash to keep paying their crews. Projects could even fail because of long payment delays, with difficult legal ramifications in allocating responsibility between the non-building ISP and the slow-paying broadband offices.

The soufflé chef should solve his timing problem by preheating the oven. Similarly, state broadband offices should position themselves to pay promptly with a monitoring plan that can build confidence on an ongoing basis.

A system of random inspections might give a broadband office reassurance that shortcomings will be detected, so timely payments are an acceptable risk. Strong local coordination might yield knowledge about permits and construction activities so that the broadband office is ready when the time comes to pay promptly and is confident the project is on track, even if the documentation provided as justification for payment is lengthy and takes time to fully review.

Field validation

Last but not least, good BEAD monitoring will require some boots on the ground.

The vast majority of telecommunications infrastructure is deployed not on private land but in the public right-of-way. For their own staff in the field, ISPs must be able to identify the assets that they own and maintain. It follows that it’s possible for knowledgeable observers to find out what’s there just by looking.

Usually, this does not even require the cooperation or knowledge of the ISP, and that has certain advantages. It can save the trouble of coordinating, and inspection can even be stealthy, depriving an ISP of the opportunity to manipulate what gets seen. In other cases, though, an ISP’s cooperation in the inspection process can give the field validators access to things not ordinarily visible.

The process of physical inspection of networks differs somewhat for aerial vs. underground fiber, for cable and DSL, and especially for fixed wireless vs. wireline. Field validation can be prone to false positives – for example, when fiber conduit gets installed in advance of the actual fiber-optic cable being blown through it. Dark fiber and lit fiber look the same in an outside plant audit. A favorable speed test result may constitute more certain positive evidence of broadband availability than an outside plant audit can provide, but the outside plant audit provides more certain negative evidence that a broadband service is not available. There can’t be fiber-optic service at an address if the fiber just isn’t there.

Field validation evidence for fixed wireless can be prone to false negatives. There can be special reasons why a fixed wireless signal that’s generally available isn’t there on the specific day when a validator is in the field. Sometimes a signal exists but its owner isn’t known, especially in the case of unlicensed fixed wireless coverage.

A well-designed field validation strategy will make use of risk-weighted random sampling. Desktop research can indicate where the risks are highest that the promised coverage is falling short, and where default risks are highest. When evidence from the field shows that a BEAD subgrantee isn’t getting the job done, the state broadband office should start asking some hard questions.

Alternative technology

While BEAD prioritizes end-to-end fiber, and approved Initial Proposals to date have secured fiber commitments for the vast majority of BEAD-funded BSLs have been so far, other technologies are also BEAD eligible. Fixed wireless and LEO satellite, in particular, will be critically important in reaching many of the hardest-to-serve locations and achieving nationwide universal coverage.

BEAD makes a distinction between the “reliable” technologies of cable (coax) and licensed fixed wireless (DSL is also listed as a “reliable” technology but typically cannot meet the speed requirements and so is not expected to receive BEAD funding) and the “alternative” technologies of LEO satellite and unlicensed fixed wireless. NTIA guidance on how BEAD should handle alternative technology projects was released on January 2, 2025.

LEO satellite presents special challenges for BEAD monitoring, since it’s the nature of the technology that its equipment is not ground-based, is constantly in motion, and is not dedicated to serving any particular area. Satellites in low Earth orbit move much faster than the Earth rotates and maintain service to any given BSL through regular handoffs from one satellite to another.

So, a whole fleet of satellites constantly relieving each other is needed to deliver a continuous signal to any given location, but that fleet of satellites serves many other areas as well, extending to the other side of the world. It’s not really possible to track a BEAD subgrant to specific equipment that’s dedicated to serving specific targeted locations.

To deal with this, NTIA guidance for LEO Capacity Subgrants gives “flexibility” to design grants that cover “estimated cost for reserving capacity in [a target] project area” during a 10-year period of performance, adding that “given the nascent nature of LEO technology, NTIA will provide additional technical assistance to Eligible Entities on how to evaluate the reasonableness of costs.” It’s too soon to forecast the ramifications of the new alternative technology guidance for BEAD monitoring, but state broadband offices will need to track how well alternative technology subgrantees fulfill their promises.

The epistemic journey to justified confidence that the BEAD networks are getting built

 BEAD monitoring is an epistemic problem, a problem pertaining to knowledge. Usually people say they “know” or “don’t know” things, without feeling the need for more nuance than that. Sometimes more refinement is needed, and we have to think in terms of degrees of confidence and probability distributions, and sources of uncertainty and evidence processing. For states administering the BEAD program with an interest in knowing, first, whether BEAD construction is on track, and later, whether BEAD networks are complete and operational, a lot of epistemic nuance will be needed.

They’ll need to avoid both overconfidence and credulity, on the one hand, which might create an opening for BEAD-funded ISPs to underdeliver, and undue caution and skepticism, on the other hand, which could lead to tardy payments and risks of disruption, or even to unfairly penalizing subgrantees that are fully compliant.

Skillful solutions to the epistemic challenge of BEAD monitoring will involve gathering and analyzing a variety of evidence, updating confidence based on evidence in an appropriate way, and allocating scrutiny in proportion to risk. Risk-weighted random sampling will be critical to containing costs while still maintaining accountability. But too much economy in BEAD monitoring will be penny-wise and pound foolish, since it will leave state broadband offices without adequate information to track progress, pay promptly, and identify and address delays and shortcomings.

States should spend what it takes for robust monitoring of the BEAD rollout that will deliver universal internet coverage.

Connected Nation stands ready to help states implement BEAD monitoring, including robust field validation powered by risk-weighted random sampling.

See our new BEADwise Data and Verification Services suite for more information.