Google PageSpeed Insights runs a live performance audit of bangordailynews.com using real-world Chrome user data (Core Web Vitals) and a simulated Lighthouse lab test. It measures how fast pages load and how responsive they feel for readers — slow scores directly affect search ranking and reader retention.

Core Web Vitals

Lighthouse Scores lab data

Lab Metrics Lighthouse simulation

Trends last day

A static audit of every script tag, iframe, and stylesheet declared in the HTML source of each tracked page. This is distinct from the Network Load Analysis (which captures what the browser actually loads at runtime, including dynamically injected scripts). Page source shows the declared starting point — what is written into the HTML before JavaScript runs. Scripts present in source but not in the network log may be conditional or suppressed; scripts in the network log but not in source were injected dynamically by other scripts.

Chrome UX Report (CrUX) is Google's database of real-user performance data collected anonymously from Chrome browsers worldwide. Unlike Lighthouse lab simulations, these numbers reflect what actual BDN readers experienced — on their real devices, connections, and conditions. CrUX data is a 28-day rolling window, updated weekly, giving us a stable trend line that smooths out day-to-day noise. The API returns the 25 most recent weekly snapshots (~6 months of history), letting us see whether performance has improved or degraded over time. Note: CrUX requires ~10,000+ real Chrome sessions per URL per month — only the homepage and site-wide origin qualify; individual article URLs do not have enough traffic.

Google Analytics 4 tracks how readers use the site — how long they stay, how many pages they read, whether they come back, and what devices they use. This source is used primarily to look for behavioral signals consistent with the browser stability issues documented in the Network Load Analysis: specifically, whether mobile sessions are anomalously shorter than desktop sessions, and whether older mobile devices show suppressed engagement relative to current hardware.

A HAR (HTTP Archive) file is a raw network log captured from a real article page load — every request the browser made, how long each one took, and how large each resource was. Our analysis script processes that log to surface what's actually slowing pages down: third-party scripts, oversized images, render-blocking resources. Think of it as a receipt for everything that happened during one page load.

Mather Economics is a subscription analytics firm that tracks reader behavior on BDN. Their data arrives via a nightly file on Mather's FTP server. All metric names, calculations, and thresholds come directly from Mather — the primary KPI is their Conversion Ratio, which measures how many anonymous readers convert to paying subscribers out of every million unique visitors.

Trends weekly snapshots

A cross-source analysis produced by running all latest dashboard data through Claude CLI. The prompt asks Claude to surface anomalies, trends, cross-source patterns, open questions, and risks — citing specific numbers for every claim. This is an experimental feature. Output is raw, unverified AI text. It reflects patterns in the data as Claude reads them and should be treated as a starting point for further inquiry, not a source of record. An admin runs npm run build:prompt to generate response.