
Open the visual model ↗
Name the claim before choosing the metric
Separate four questions: what the publisher delivered, what traffic was observed, what conversions a system attributed, and what outcomes were caused by the campaign. An impression or newsletter placement supports delivery. A tagged visit supports observed referral under the analytics method. A platform-attributed conversion assigns credit under its rules. Incremental effect requires a credible counterfactual, often a controlled design. These claims should not share one unlabeled results column.
Write the campaign brief with contracted units, destinations, approved creative, reporting systems, conversion definition, attribution window, model, privacy constraints, and owners. If the sponsor cannot share downstream results, promise only the publisher-side measures that can be produced. Do not backfill an unsupported sales claim from clicks or brand-search movement.
| Question | Evidence | Safe label |
|---|---|---|
| Was placement delivered? | Publisher ad/email logs | Delivered units |
| Was referred traffic observed? | Tagged sessions under analytics rules | Observed referred sessions |
| Which touchpoint got credit? | Named attribution model/window | Attributed conversions |
| Did campaign cause lift? | Qualified treatment/control design | Estimated incremental effect |
Source notes: Attribution overview · About Conversion Lift
Build a campaign taxonomy that survives redirects
Assign a publisher campaign ID and creative or placement ID. Agree on lower-case, controlled values for source, medium, campaign, content, and any sponsor-side ID. Google Analytics documents standard campaign parameters and warns through its behavior that naming consistency matters; these parameters organize referred traffic, not causality. Keep a registry with the final destination and exact tagged URL.
Test the full chain: ad or newsletter link, publisher redirect, sponsor redirect, landing page, consent state, and analytics receipt. Confirm that parameters are not stripped, duplicated, or exposing personal data. Where the sponsor requires a click tracker, document redirect order and latency. Use test identifiers and remove test events from the agreed report when the system supports it.
| Field | Illustrative value | Rule |
|---|---|---|
| publisher_campaign_id | sp-2026-014 | Immutable internal key |
| utm_source | publisher-name | Controlled lower-case vocabulary |
| utm_medium | newsletter | Channel, not placement name |
| utm_campaign | sponsor_launch | Agreed campaign label |
| utm_content | issue42_top | Creative/placement variant |
Illustrative values; never place personal data in campaign parameters.
Source notes: URL builders: collect campaign data with custom URLs
Reconcile systems without hiding unmatched traffic
Create a match table from publisher campaign and click IDs to the sponsor's campaign and conversion IDs. Record clock zones, attribution windows, consent effects, cross-device limitations, duplicate handling, and model. The publisher may count 5,000 outbound redirects while the sponsor sees 4,200 sessions because their systems observe different checkpoints and units. Report both; do not describe the numerical difference as 800 missing events unless a reconciled event-level mapping proves that interpretation.
Hypothetically, 1,000,000 delivered impressions lead to 5,000 publisher-recorded outbound clicks, 4,200 sponsor-observed sessions, and 120 last-click-attributed purchases. The click-through rate is 0.5% using delivered impressions. The sponsor's attributed conversion rate is about 2.86% using its 4,200 sessions. Neither figure establishes that 120 purchases were incremental. Present the denominator, model, and window next to each rate.
| Checkpoint | Count | Owner/definition |
|---|---|---|
| Delivered impressions | 1,000,000 | Publisher ad server |
| Outbound clicks | 5,000 | Publisher redirect |
| Landing sessions | 4,200 | Sponsor analytics |
| Attributed purchases | 120 | Sponsor model/window |
| Incremental purchases | Not measured | Requires separate design |
Source notes: Attribution overview
Put boundaries in the sponsor report
The report should restate promised units and delivery first, then observed traffic and any sponsor-supplied outcomes in separate sections. Include definitions, period, time zone, data source, attribution model and window, unmatched share, known incidents, and whether results are preliminary or final. Label sponsor-reported data as such and preserve the received file or signed-off total.
If an incrementality study is proposed, define eligibility, assignment, contamination risks, minimum detectable effect, privacy review, and decision rule with a qualified analyst before launch. Some ad platforms offer controlled lift products, but product availability and methodology do not turn an ordinary tagged campaign into an experiment. A strong report is useful precisely because it limits every claim to the evidence that produced it.
- Keep delivery, observed referral, attributed outcome, and incrementality separate.
- Freeze campaign names and identifiers before launch.
- Test every redirect and landing destination.
- Report unmatched events and consent limits.
- Do not infer causal lift from attributed conversions.
Source notes: About Conversion Lift
Continue the work
Sources & limits
Attribution depends on configured systems, identity, consent, windows, and models. Causal claims require a suitable study and qualified analysis.
- URL builders: collect campaign data with custom URLs
Google Analytics documentation · Publication date not stated · Primary source checked · 19 September 2026 - Attribution overview
Google Analytics documentation · Publication date not stated · Primary source checked · 19 September 2026 - About Conversion Lift
Google Ads documentation · Publication date not stated · Primary source checked · 19 September 2026
Source claims and editorial judgments remain separate. Send a correction with the passage and supporting evidence.

