Most tracking compliance failures are not banner failures: they are pixels, tags, SDKs and link decorations that someone added to a page or app, which send data to third parties before consent or from pages nobody thought to check. Regulators now treat tracking technologies as a security and governance problem as much as a consent one: the ICO's finalised storage and access guidance covers pixels, fingerprinting and scripts, the FTC has used the Health Breach Notification Rule against health data shared through advertising tools, and HHS guidance treats tracking on patient portals as a HIPAA disclosure. This guide covers the governance layer: inventory, classification against the exceptions, technical enforcement of choices, sensitive pages, testing and evidence. It is for privacy engineers, web and app security teams, marketing technology owners and DPOs. Banner design, CMP set-up and consent signals are covered in the Consent Management Guide: Cookies, CMPs and GPC.
Key takeaways
- The rule covers storage and access, not just cookies. Pixels, web storage, fingerprinting, link decoration and scripts all fall under the ePrivacy Article 5(3) and UK PECR rules on storing or accessing information on a device.
- Exceptions are narrow and purpose-bound. Under the ICO's April 2026 guidance, any advertising use, cross-site tracking or second purpose takes a technology outside the strictly necessary and statistical exceptions.
- Sensitive pages change the stakes. Trackers on health, finance, children's and logged-in pages can turn an analytics tag into a disclosure of special category data or PHI.
- Enforce consent in code, then prove it. A tag manager rule is not enough; block by default, restrict destinations with Content Security Policy, and test with automated crawls before and after consent.
- Treat tags as third-party code. Every tracker is a supplier with access to your pages, so it needs an owner, a contract, a review and change control.
What the rules cover now
In the EU, Article 5(3) of the ePrivacy Directive requires consent for storing information, or gaining access to information already stored, on a user's device, unless the storage or access is for transmitting a communication or is strictly necessary for a service the user requested. In the UK the equivalent rule is in PECR, as amended by the Data (Use and Access) Act 2025. The GDPR then governs any personal data processed afterwards. For the legal background, see the ePrivacy Directive Guide.
- UK status: the ICO finalised its guidance on the use of storage and access technologies on 29 April 2026, after a draft in December 2024 and an update in July 2025 for the Data (Use and Access) Act. It names cookies, tracking pixels, link decoration and navigational tracking, device fingerprinting, web storage and scripts or tags.
- UK exceptions: the guidance sets out communication, strictly necessary, statistical purposes, appearance and emergency assistance exceptions, and explains the "simple means of objecting" required for the statistical and appearance exceptions.
- Penalties: the Data (Use and Access) Act brings maximum PECR fines into line with UK GDPR levels, which makes tracking failures a material financial risk in the UK, not only a reputational one.
- EU enforcement: the CNIL fined Google a total of 325 million euros (decision of 1 September 2025) partly because, at Google account creation, refusing advertising cookies was harder than accepting them and users were not told that access depended on them, affecting more than 74 million accounts.
- US position: there is no federal cookie consent rule, but tracking that discloses health or other sensitive data is pursued through the FTC Act, the FTC Health Breach Notification Rule, HIPAA and state privacy laws. See the US State Privacy Laws Guide.
Tracking technologies and how they collect data
| Technology | How it works | What it can leak | Where to find it |
|---|---|---|---|
| Cookies (first and third party) | Values stored by the browser and sent with requests | Persistent identifiers, session state, ad IDs | Browser storage, Set-Cookie headers, tag manager containers |
| Tracking pixels and beacons | Tiny image or script requests to a third party, often with URL parameters | Page URL and title, referrer, IP address, event data, hashed emails | Network requests to ad and analytics domains, email templates |
| Web storage (localStorage, sessionStorage, IndexedDB) | Script-controlled storage on the device | Identifiers that survive cookie clearing | Browser developer tools, automated crawlers |
| Device fingerprinting | Combining device and browser attributes into an identifier | A stable identifier without any stored value | Script analysis, calls to canvas, audio and font APIs |
| Link decoration and navigational tracking | Identifiers appended to URLs (click IDs) or bounce redirects | Cross-site identity carried in the URL | Outbound links, campaign URLs, redirect chains |
| Scripts and tags (session replay, chat, heatmaps) | Third-party JavaScript running in your page | Keystrokes, form fields, page content, screen recordings | Tag manager, page source, Content Security Policy reports |
| Mobile SDKs | Third-party libraries compiled into apps | Advertising IDs, location, app events, contacts | App dependency lists, network traffic analysis |
| Server-side tagging and conversion APIs | Your server forwards events to third parties | Anything your server chooses to send, invisible to browser tools | Server code and configuration, data flow maps |
Server-side tagging deserves a specific warning. It can reduce what third parties collect directly from browsers, but it also hides the data flow from browser-based scanners and from users. If the purpose needs consent, moving the call to the server does not remove that need.
Classifying each tracker against the exceptions
The ICO's exceptions are tests of purpose. Classify each tracker by what it is used for, not by what the vendor calls it.
| Exception (ICO, April 2026) | Likely to qualify | Does not qualify |
|---|---|---|
| Communication | Session cookies for load balancing; fingerprinting solely for network management | Anything not needed to transmit the communication |
| Strictly necessary | Basket contents, keeping a logged-in user identified, security for a requested activity, link decoration for authentication, embedded video that sets nothing until play | Advertising of any kind, cross-site or cross-device tracking, social tracking of non-members or logged-out users |
| Statistical purposes | Page visit totals, scroll depth, device and browser types, referrers, A/B testing, coarse location, load times, with aggregate output | Recordings of individual visitors, ad performance measurement, linking visitor IDs to purchases, profiling or segmentation |
| Appearance | Screen size detection, language choice, dark mode, external fonts used only to serve the font | Content chosen from inferred interests, any advertising |
| Emergency assistance | Device location after a request for help | Location for any other purpose |
- Dual purpose ends the exception: the ICO says that if a technology is used for a second purpose, such as security plus analytics, the strictly necessary exception no longer applies.
- Analytics conditions: the statistical exception requires the sole purpose to be improving your own service, any provider acting only as your processor without combining the data with its own, aggregate output (daily aggregation is suggested as a starting point), clear information, and a simple, free way to object. A provider that is a joint controller cannot rely on it.
- Objection mechanics: toggles on by default that users can switch off are one option, but you must not rely solely on browser settings to determine whether someone has objected.
- EU difference: the statistical and appearance exceptions are UK rules introduced through the Data (Use and Access) Act. For EU audiences, Article 5(3) has only the transmission and strictly necessary exemptions, so check how each national regulator you deal with treats analytics before relying on an exemption.
Sensitive pages and sensitive data
The largest enforcement actions involve trackers that sent sensitive data to advertisers. The legal route differs by country, but the engineering failure is the same: a generic tag deployed site-wide, including on pages whose URL, title or form fields reveal something sensitive.
- Health data in the US: in February 2023 the FTC announced its first enforcement action under the Health Breach Notification Rule against GoodRx, alleging it shared users' medications and health conditions with Facebook, Google, Criteo, Branch and Twilio despite promising not to, with a 1.5 million dollar civil penalty and a proposed permanent ban on sharing health data for advertising.
- HIPAA regulated entities: the HHS bulletin on online tracking technologies says tracking on user-authenticated pages such as patient portals and telehealth generally has access to PHI, that vendors meeting the business associate definition need a business associate agreement, and that cookie banners are not HIPAA authorisations. A federal court in American Hospital Association v. Becerra (20 June 2024) vacated the part treating an IP address plus a visit to a public, unauthenticated health page as PHI; the rest remains posted while HHS evaluates next steps. See the Health Data Privacy Beyond HIPAA Guide (My Health My Data).
- Special category data under the GDPR: page URLs and events that reveal health conditions, sexual orientation, religion or political views can be special category data, which needs an Article 9 condition (usually explicit consent) on top of the ePrivacy consent.
- Logged-in and transactional pages: account, checkout, application and support pages expose names, emails, order contents and free text. Session replay and form analytics tools are the usual culprits.
- Children's services: trackers on services likely to be accessed by children face stricter rules. See the COPPA and Children's Privacy Guide.
Page tiers for tag policy
| Tier | Examples | Tag policy |
|---|---|---|
| Tier 1: public, non-sensitive | Home page, product marketing, blog | Consented analytics and advertising tags allowed under the CMP |
| Tier 2: sensitive topic | Condition pages, symptom checkers, financial hardship, recruitment | No advertising pixels; analytics only if aggregate and without URL or title detail |
| Tier 3: authenticated or transactional | Portals, account settings, checkout, forms with free text | No third-party trackers except processors under contract; no session replay on input fields |
Technical enforcement of choices
- Block by default: non-exempt tags must not load until the CMP records consent. Configure the tag manager with a consent condition on every tag, and default new tags to blocked.
- Restrict destinations: use Content Security Policy directives (script-src, connect-src, img-src, frame-src) to allow only approved tracker domains, with report-uri or report-to so new destinations show up as violations. See the Content Security Policy Guide.
- Control tag manager access: limit publish rights in the tag manager to a small group, require review before publishing a container version, and keep the version history. The tag manager is a code deployment channel into production pages.
- Strip sensitive parameters: remove query strings, page titles and form values from events on Tier 2 and Tier 3 pages, and disable features that automatically collect form fields or hash emails unless they are explicitly approved.
- Respect withdrawal: when a user withdraws consent, stop further loading and delete or expire the cookies and storage you set where you can. Withdrawal must be as easy as giving consent.
- Mobile apps: initialise SDKs only after consent, check SDK default settings (many start collecting at launch), and include SDKs in software composition analysis.
- Server-side flows: forward events only when the consent state attached to the event allows it, and log that consent state with each forwarded event.
Where identity fits
Tracking and identity meet in two places. First, trackers on login, account and federated sign-in pages can capture identifiers, emails and tokens in URLs: keep third-party scripts off authentication flows entirely. Second, the tag manager, CMP and ad platform accounts are privileged SaaS identities that can inject code into every page, so they need MFA, named owners, least privilege and offboarding like any production access. A compromised tag manager account is a direct route to web skimming. See the Client-Side Security Guide and the Web Skimming Incident Playbook.
How to implement: a phased approach
- Inventory: crawl all domains and key user journeys with a headless browser in three states (no interaction, reject all, accept all), and list every cookie, storage key and third-party request. Add app SDKs from dependency manifests and server-side destinations from code.
- Assign owners and purposes: give each tracker a business owner, a purpose, the exception relied on or the consent category, the vendor's role (processor or controller) and the contract in place.
- Remove what nobody owns: anything without an owner or purpose after 30 days is removed. Old pixels from past campaigns are common.
- Tier the pages: label pages by the three tiers above, starting with authenticated and sensitive-topic pages, and apply the tag policy to each tier.
- Enforce in report-only first: deploy Content Security Policy in report-only mode to learn real destinations, then enforce. Pilot tag manager consent conditions on one site before rolling out.
- Test continuously: run the three-state crawl on every release and at least weekly, and fail the release if a non-exempt tracker fires before consent or a blocked destination appears.
- Roll back safely: keep previous tag manager container versions and CSP policies so a breaking change can be reverted in minutes; never roll back by disabling the CMP.
- Govern change: require a privacy review ticket for any new tag, SDK or server-side destination, with the purpose, data sent and page tiers.
Decision guidance
- Client-side or server-side tagging: server-side gives you control over what is forwarded and hides less from you, but more from users and regulators. Choose it for control and minimisation, never as a way around consent.
- Analytics under the UK statistical exception or with consent: rely on the exception only if the provider is your processor, output is aggregate and you offer an objection route. If you want individual journeys, replay or ad attribution, ask for consent.
- Small organisations: fewer tags is the cheapest control. Use one analytics tool, no advertising pixels on sensitive pages and a CMP with automatic blocking.
- Large organisations: central tag governance with a tracker register, CSP enforcement, release gating and a quarterly attestation by each tracker owner.
Detection, evidence and metrics
- Evidence to keep: crawl results per release, the tracker register with owners, tag manager version history, CSP violation reports, consent logs and vendor contracts.
- Monitoring: alert on new third-party domains in CSP reports, on tags published outside change control, and on trackers firing on Tier 3 pages.
| Metric | Target |
|---|---|
| Non-exempt trackers firing before consent in crawls | Zero |
| Trackers in the register with an owner, purpose and contract | 100% |
| Third-party trackers on authenticated and sensitive pages | Zero, except approved processors |
| New third-party destinations detected outside change control | Falling, each investigated within 5 working days |
| Time to remove a tracker after a decision to remove it | Under one release cycle |
Framework mapping
| Framework | Reference | Relevance |
|---|---|---|
| ePrivacy Directive and UK PECR | Article 5(3); PECR storage and access rules | Consent or exception for storage and access on devices |
| GDPR and UK GDPR | Articles 6, 7, 9 and 28 | Lawful basis, consent conditions, special category data, processor contracts |
| NIST SP 800-53 Rev. 5 | PT-3, PT-4, PT-5 | Processing purposes, consent and privacy notices |
| NIST SP 800-53 Rev. 5 | CM-7 Least Functionality; SA-9 External System Services; SR-6 Supplier Assessments and Reviews | Removing unneeded scripts, governing third-party code and vendors |
| NIST CSF 2.0 | GV.OC-03, ID.AM-04, ID.AM-07, DE.CM-09 | Legal requirements, supplier service and data inventories, monitoring runtime environments |
| ISO/IEC 27001:2022 | Annex A 5.34 Privacy and protection of PII | Meeting privacy obligations for PII |
Common failure patterns
- Trackers that fire before the banner: tags hard-coded in page templates or loaded by another tag, outside the tag manager's consent conditions.
- Site-wide pixels: an advertising pixel deployed to every page, including portals, condition pages and checkout, sending URLs and titles that reveal health or financial details, the pattern behind the GoodRx action.
- Reject that does not reject: cookies or storage still set after "reject all", or refusing made harder than accepting, the issue the CNIL found at Google account creation.
- Analytics stretched beyond the exception: relying on the statistical exception while using the same tool for ad attribution, session replay or audience building.
- Invisible server-side flows: conversion APIs forwarding emails and purchase data regardless of consent state, missed by browser scans.
- Ownerless tags: pixels left over from agencies and campaigns, with tag manager publish rights still held by former suppliers.
Practitioner checklist
- Crawl every domain and key journey in no-interaction, reject-all and accept-all states, and add app SDKs and server-side destinations to the inventory.
- Keep a tracker register with owner, purpose, exception or consent category, vendor role and contract for each entry.
- Classify each tracker against the ICO exceptions by purpose, and treat any second purpose as needing consent.
- Tier pages and keep advertising pixels and session replay off sensitive and authenticated pages.
- Block non-exempt tags by default and enforce destinations with Content Security Policy.
- Restrict tag manager publish rights, require review, and remove former agencies and suppliers.
- Gate releases on automated pre-consent crawl results.
- For US health services, map tracker use against the HHS bulletin and the FTC Health Breach Notification Rule.
- Keep crawl evidence, consent logs and tag manager history for audits and complaints.
Standards and references
- ICO: Guidance on the use of storage and access technologies
- ICO: What are the exceptions?
- EUR-Lex: Directive 2002/58/EC (ePrivacy Directive)
- HHS: Use of Online Tracking Technologies by HIPAA Covered Entities and Business Associates
- FTC: Enforcement action to bar GoodRx from sharing consumers' sensitive health info for advertising
- CNIL: Cookies and advertisements inserted between emails: Google fined 325 million euros
Related NHI Mgmt Group resources: Consent Management Guide: Cookies, CMPs and GPC · ePrivacy Directive Guide · IAB TCF and Global Privacy Control Signals Guide · Client-Side Security Guide. This guide summarises tracking technology rules and enforcement and is not legal advice.