Analytics should respect the page you are on

A public product page and a private local workspace may live under the same brand, but they don't have the same measurement needs. A visitor reading a feature description is doing something different from an author reviewing a screenshot or opening an offline lesson.

October's ProgreTech web-app work added consent-aware analytics to hosted public surfaces while preserving local and offline boundaries. The useful developer lesson is to decide where measurement belongs before adding a tag.

The quickest way to check the setup is to watch the browser's network panel from a fresh visit. We'll get to that after deciding which pages should collect anything at all.

Make a surface map

List the pages and modes the application serves. Include public marketing pages, article pages, signed-in views, authoring tools, exported HTML, and offline mode.

For each one, write the intended network behavior. A public article may offer optional analytics after consent. A portable learning file may need to work with no network requests at all. An authentication callback should avoid sending sensitive URL contents into a measurement system.

This map is more useful than a single global “analytics enabled” flag. It lets reviewers understand why a script appears on one route and is absent from another.

Decide before loading the optional code

A consent control should affect actual behavior. If the page fetches the optional analytics script before the person makes a choice, the initial request has already happened.

Keep the consent UI and its necessary code available independently of the external tag. Store the choice in a way that the application can read on the next visit. Give the person a visible way to change it later.

The initial state deserves special attention. A fresh browser, cleared storage, and an offline revisit should each produce a sensible result. Avoid treating a storage read failure as permission to start optional requests.

Send a small, deliberate page description

Automatic collection can include more context than the site needs. Design a sanitized metadata contract for the public pages you measure.

A stable article title and canonical public path are useful examples. A draft body, a private document name, a search query containing personal details, or an authentication URL should stay outside that contract.

Keep the allowed values in one small place. For a modest site, a map of public page names is easy to read and maintain. It also avoids accidentally passing a private URL through a general-purpose tracking helper.

Inspect the actual requests

Use a clean browser profile and observe the network from the first page load. Before consent, check that optional measurement requests are absent. Accept, navigate, and inspect the requests that appear. Decline or withdraw the choice and verify the intended behavior afterward.

Then repeat the review on excluded surfaces. Open a portable file, use a local authoring route, and try the offline path. Save a sanitized receipt of the test without retaining private payloads.

A useful test matrix is compact:

fresh visit / no choice
accepted / public article navigation
declined / reload
withdrawn / next navigation
offline / saved reading
excluded local or portable surface

Each row should have an observed result. A screenshot of a consent banner covers the visual control, while a network inspection covers collection behavior.

Keep consent usable offline

The preference controls are part of the application's own interface. If they disappear offline because their stylesheet or script was never cached, the product has an avoidable inconsistency.

Cache the local assets needed to display and change the preference. Keep external analytics optional and separate. Browser storage and service-worker updates should preserve the person's intended choice according to a documented policy.

A practical release boundary

Measurement can help a publisher understand whether useful material is being found. The implementation should stay proportional to that purpose.

Start with a fresh browser, make each consent choice, and inspect the requests. Repeat that check on the local and offline pages. Save the small test recipe with the site so the next tag change gets the same attention.