Robust and Extensible Measurement of Broadband Plans with BQT+

Independent, street address-level broadband data is essential for evaluating Internet infrastructure investments, such as the $42B Broadband Equity, Access, and Deployment (BEAD) program. Evaluating these investments requires longitudinal visibility into broadband availability, quality, and affordability, including data on pre-disbursement baselines and changes in providers’ advertised plans. While such data can be obtained through Internet Service Provider (ISP) web interfaces, these workloads impose three fundamental system requirements: robustness to frequent interface evolution, extensibility across hundreds of providers, and low technical overhead for non-expert users. Existing systems fail to meet these three essential requirements. We present BQT+, a broadband plan measurement framework that replaces monolithic workflows with declarative state/action specifications. BQT+ models querying intent as an interaction state space, formalized as an abstract nondeterministic finite automaton (NFA), and selects execution paths at runtime to accommodate alternative interaction flows and localized interface changes. We show that BQT+ sustains longitudinal monitoring of 64 ISPs, supporting querying for over 100 ISPs. We apply it to two policy studies: constructing a BEAD pre-disbursement baseline and benchmarking broadband affordability across over 124,000 addresses in four states.

💡 Research Summary

The paper addresses a pressing need in U.S. broadband policy: the ability to collect address‑level data on availability, advertised speed tiers, and pricing directly from Internet Service Providers’ consumer‑facing websites (Broadband Availability Tools, or B‑ATs). Existing regulatory datasets such as FCC Form 477, the Broadband Data Collection, and the National Broadband Map rely on provider self‑reporting, are coarse in both space and time, and lack price or eligibility information. Consequently, policymakers overseeing the $42 billion BEAD (Broadband Equity, Access, and Deployment) program cannot evaluate the effectiveness of past investments or monitor ongoing changes without an independent, longitudinal source of data.

Prior work, exemplified by the Broadband‑plan Querying Tool (BQT), tackled this problem by writing a deterministic, end‑to‑end script for each ISP. Each script encodes both the measurement intent (what fields to extract) and the exact navigation sequence required to reach those fields. While this approach works for short‑term studies on a handful of large ISPs, it fails to meet three system requirements that arise at policy scale: (1) extensibility across hundreds of heterogeneous ISPs, (2) robustness to frequent, unsynchronized UI changes, and (3) low technical overhead so that non‑engineers (policy analysts, NGOs, legislators) can author and maintain the queries.

BQT+ is introduced as a fundamentally different architecture that separates intent from execution by modeling each ISP’s web interaction as an abstract nondeterministic finite automaton (NFA). In this model:

• States are detector‑defined predicates over observable UI cues (e.g., “‘Enter address’ textbox is visible”, “cookie consent popup appears”).
• Actions are the set of permissible user interactions (click, type, select).
• Transitions map a (state, action) pair to a set of possible successor states, allowing multiple admissible paths through the interface.

The NFA captures the structure of the interaction without prescribing a single linear trace. At runtime, a traversal engine observes the current UI, selects an enabled action, and follows one of the possible transitions. If the website changes—adding a pop‑up, reordering steps, or inserting an extra click—the engine simply needs an updated state or transition definition; the rest of the automaton remains valid. This yields three concrete benefits:

1. Extensibility – Adding a new ISP requires only a declarative specification of its NFA, not custom imperative code. The authors demonstrate support for over 100 ISPs after initially handling 64.
2. Robustness – Across an eight‑month longitudinal study of 64 ISPs, 56 UI changes were observed (35 ISPs changed at least once). BQT+ absorbed all changes by editing or adding a handful of states/transitions, whereas the original BQT would have required dispersed code rewrites.
3. Low Technical Overhead – The state/action specifications are written in a simple DSL; non‑programmers can modify them without deep knowledge of Selenium or Python control flow. Execution logic (retry, error handling) is encapsulated in separate modules, further reducing the maintenance burden.

Implementation details include a Selenium‑based driver for page rendering, image/text matching for state detection, and a heuristic policy for choosing among multiple possible transitions. Errors trigger a generic recovery module that is independent of the NFA, preserving the declarative core.

The system is validated through two policy‑relevant case studies:

• BEAD pre‑disbursement baseline – For four states (Virginia, Maryland, North Carolina, South Carolina), BQT+ generated address‑level baselines of plan availability, average advertised speeds, and median monthly costs, providing the granular data needed to assess where BEAD funds should be allocated.
• Affordability benchmarking – Over 124 000 addresses across the same states, the authors measured plan prices and speed tiers, identified gaps in low‑cost options, and supplied the results to Virginia legislators for a statewide affordability bill.

Both studies illustrate that BQT+ can produce high‑resolution, up‑to‑date datasets that are impossible to obtain from self‑reported regulatory sources, and that the system can include small, regional ISPs whose coverage is critical for equity analyses.

In conclusion, the paper argues that a declarative interaction‑state abstraction is essential for sustainable, policy‑grade broadband measurement. By treating the web interface as an NFA, BQT+ transforms what was previously a brittle, code‑heavy pipeline into a modular, maintainable platform. The authors suggest that this approach can generalize beyond broadband—e.g., to e‑commerce price monitoring or public‑service portal audits—wherever large numbers of heterogeneous web interfaces must be queried over long periods. Future work includes integrating machine‑learning‑based UI element detection, automated synthesis of NFA specifications from observed traces, and extending the framework to other domains requiring systematic web‑based data collection.