Strategy
Part of A privacy first measurement strategy that starts with the decision, then the metric
Website, email, paid platform and app measurement, compared on blind spots
Compare website, email, paid-platform and app measurement through one method covering decisions, data access, permissions, attribution and blind spots.
A privacy first measurement channel strategy should preserve each channel's observation boundary. Website events, email responses, advertising-platform conversions and app activity are produced under different technical and permission conditions. They are not interchangeable evidence of the same person or outcome.
What to take away
- Website, email, paid platform and app data are produced under different conditions and are not interchangeable evidence.
- Each channel has a material blind spot, so no single channel should answer a question it cannot observe.
- Platform-attributed totals need separate labelling before any merge.
- Email delivery and clicks do not prove later impact, so keep each measure separate.
- Choose channels only after the outcome and comparison method are fixed.
Comparison method
The source check closed on 5 September 2026. Four common channel contexts were included because each can contribute to an English organisation's campaign or service decision. Criteria were identical: decision served, unit observed, dependency, permission question, attribution limit and useful verification.
Channel measurement blind spots
Website
- Observed
- page/server events
- Dependency
- consent design
- Blind spot
- omits people
- Check
- refusal-state test
Email/CRM
- Observed
- sends and responses
- Dependency
- recipient status
- Blind spot
- no later impact
- Check
- cohort reconcile
Paid ads
- Observed
- platform conversions
- Dependency
- tags and attribution
- Blind spot
- not incrementality
- Check
- split modelled values
App
- Observed
- in-app events
- Dependency
- permission and SDKs
- Blind spot
- no downstream joins
- Check
- no-permission path
No account, tag, email, app or consent flow was tested, and prices, performance scores and market rankings were excluded. ICO sources apply across the UK. Google and Apple records describe global product ecosystems and do not establish English availability, legality or results.
Directly observed
- Website or browser
- implemented page or server events
- Email and CRM
- sends, responses and recorded customer actions
- Paid advertising platform
- conversions recognised under platform settings
- App measurement
- in-app events and permitted platform signals
Main dependency
- Website or browser
- collection design and consent or exception analysis
- Email and CRM
- recipient status, message rules and record matching
- Paid advertising platform
- tags, imports, attribution choices and vendor processing
- App measurement
- operating-system policy, SDK behaviour and user choice
Material blind spot
- Website or browser
- events may omit people or misstate intent
- Email and CRM
- delivery and clicks do not prove later impact
- Paid advertising platform
- modelled or attributed credit is not incrementality
- App measurement
- cross-app visibility may be restricted
Sensible check
- Website or browser
- synthetic event and refusal-state test
- Email and CRM
- reconcile a defined cohort to operational data
- Paid advertising platform
- separate observed, imported and modelled values
- App measurement
- inspect SDK data flow and denied-permission state
Website or browser measurement
Use this channel when the decision depends on behaviour within a controlled site journey. Specify each event, denominator and failure state. The Information Commissioner's Office (ICO) storage and access technologies guidance covers cookies, pixels, scripts and related techniques. Classification depends on the actual purpose and operation, not on a label such as "analytics" or "first party".
Email and CRM evidence
Known-customer data can connect communication with later operational events, provided the join and permission basis are properly reviewed. The ICO's electronic mail marketing guidance was updated on 28 April 2026. It explains PECR rules for electronic mail; it does not make an open or click a causal outcome.
Keep send eligibility, delivery, interaction and transaction as separate measures. State how shared inboxes, forwarding, repeat purchases and opt-outs affect the cohort.
A practical cohort match is to freeze the eligible, non-suppressed recipient list at send time, then join customer ID to later orders in the CRM and exclude orders outside the agreed window.
Paid-platform conversions
Google Ads conversion documentation describes distinct website, app, phone and offline conversion sources, with separate actions for different events. The settings are a vendor-defined reporting system. An advertiser still needs its own conversion definition, deduplication and reconciliation.
Do not merge a platform-attributed total with site events without showing overlap. Export platform conversion records and site order records to a warehouse such as BigQuery or a spreadsheet. Normalise the order ID, then join on that ID or a salted customer hash. Label every row as site, platform matched or platform only.
Illustrative figures: the platform reports 120 conversions and the site records 100 sign-ups. If 20 site sign-ups match a platform conversion within the same 30-day window, count 100 site sign-ups plus 100 platform-only conversions, not 220. A forced total of 220 counts the 20 matched outcomes twice.
Data flow: platform export, normalise ID, join to site orders, label matched and unmatched rows.
App measurement
Apple's user privacy and data use page explains AppTrackingTransparency requirements for covered tracking across apps and websites owned by other companies. This is Apple policy. UK privacy and electronic-communications analysis remains separate.
Test the no-permission path and inventory every SDK. A successful event in a developer console says little about downstream joins unless the data route is documented.
Before you act
Choose channels after the outcome and comparison are fixed. Government guidance on measuring service success recommends a range of sources rather than digital analytics alone. For private teams, that is a useful method boundary.
- Specify each event, denominator and failure state.
- Keep send eligibility, delivery, interaction and transaction separate.
- Separate observed, imported and modelled conversion values.
- Test the no-permission and refusal-state paths.
- Report channels beside one another when units and windows differ.
Common questions
Why can platform conversion totals not be merged with site events?
Platform settings are a vendor-defined reporting system. Merging a platform-attributed total with site events without showing overlap hides duplication. Advertisers still need their own conversion definition, deduplication and reconciliation. Label any modelled component and keep the observed part where available.
What does the ICO guidance on electronic mail marketing actually cover?
The ICO electronic mail marketing guidance was updated on 28 April 2026 and explains PECR rules for electronic mail. It does not make an open or click a causal outcome. Send eligibility, delivery, interaction and transaction should stay separate measures, and the cohort should state how shared inboxes, forwarding, repeat purchases and opt-outs affect it.
What should be checked before trusting app measurement data?
Apple's user privacy and data use page explains AppTrackingTransparency requirements for covered tracking across apps and websites owned by other companies, but this is Apple policy and UK privacy analysis remains separate. Test the no-permission path, inventory every SDK, and document the data route, because a successful console event says little about downstream joins.
