Measuring SignalLift performance

Purpose

Operational guide for analysts, AdOps and technical teams.

Enables user to quantify the impact of PPID and PPS signals on revenue, fill, and eCPM, using GAM360 reporting and Google sheets.

Our Approach

Segmentation

• SignalLift traffic is split into signal-lift group (users with PPID / PPS exposure) and hold-out group (no signal).

• Segmentation is applied using GAM key-value (AnonymisedSignalLift) to ensure Revenue and Ad Opportunity alignment.

• Traffic split is:

  • 98% SignalLift traffic (treatment group)

  • 2% Hold-out group

• This mimics Google Impact Estimate behaviour.

Metrics and Formulas

Denominator – AdX Opportunities

• Compound metric:

TOTAL_IMPRESSIONS + UNFILLED_IMPRESSIONS + AD_SERVER_IMPRESSIONS

• Rationale:

  • Captures every auction where AdX could participate:

    • AdX wins (TOTAL_IMPRESSIONS)

    • AdX losses (AD_SERVER_IMPRESSIONS)

    • No-fill opportunities (UNFILLED_IMPRESSIONS)

  • Fully key-value filterable in GAM

  • Measures revenue per opportunity, capturing both fill rate and CPM effects

• Decision justification:

  • Using only AD_EXCHANGE_REQUESTS or TOTAL_IMPRESSIONS underestimates opportunities

  • Summing these three metrics provides a conservative, accurate view of supply-side performance, consider both fill rate and CPM effects

Numerator – Revenue

• Metric: AD_EXCHANGE_REVENUE

• Filtered by the same key-values as the denominator

• Rationale:

  • Isolates revenue generated by AdX for the signals

  • Excludes GAM line-item revenue to avoid inflating eCPM

• Decision justification:

  • Ensures numerator and denominator represent the same opportunity set, providing true per-opportunity yield

Yield Calculation

• Formula:

  CODE

  eCPM = AD_EXCHANGE_REVENUE / (TOTAL_IMPRESSIONS + UNFILLED_IMPRESSIONS + AD_SERVER_IMPRESSIONS) * 1000

• Captures the combined effect of:

  • CPM changes (higher bids won)

  • Fill rate changes (more auctions successfully monetised)

• Enables straightforward comparison between signal-lift and the hold-out group

Uplift Calculation

• Formula:

  CODE

  Uplift (%) = ((eCPM_signal - eCPM_holdout) / eCPM_holdout) * 100

• Measures the relative increase in revenue per opportunity due to the PPID/PPS signal

• Can be applied across inventory types, devices, geographies, or publisher segments

• Provides a clear commercial view of incremental performance attributable to the signal

Process

Follow these simple steps to reveal the SignalLift performance:

1. Open the SignalLift report. See Setting up your Anonymised SignalLift report for details on how to create it.

2. Export the data to a CSV file or a Google Sheet

3. Take a copy of the Anonymised SignalLift Template (simply click on the link, you will be requested to Make a copy). Note: The file does contain a functional script which will also be imported.

4. Open your copy of the Anonymised SignalLift Template.

5. On the first tab of the script there are some further instructions. The choice is to:

   1. Import your CSV file from GAM into the sheet, or

   2. Copy & paste the data from your exported Google sheet to the template.

6. Run the functional script.

7. The revenue uplift data will be presented on a second tab labelled Summary.

Methodology Notes / FAQs

Why do TOTAL_IMPRESSIONS + UNFILLED_IMPRESSIONS + AD_SERVER_IMPRESSIONS sometimes exceed expected served impressions?

• Answer: This sum captures all monetisation opportunities, not just ads actually served. Differences occur due to:

  • Multi-slot pages (each slot counts as an opportunity)

  • Guaranteed line items that serve instead of AdX

  • Unfilled inventory where no ad ultimately served

Why do we exclude GAM line-item revenue from the numerator?

• Answer: Only AD_EXCHANGE_REVENUE is used, because the denominator includes all opportunities AdX could participate in. Including line-item revenue would distort eCPM and signal-lift comparisons.

How does this eCPM metric capture both CPM and fill rate effects?

• Answer:

  • CPM effect: Higher bids won by AdX increase AD_EXCHANGE_REVENUE (numerator)

  • Fill effect: More unfilled or line-item-served impressions increase the denominator

• Together, the formula measures revenue per opportunity, reflecting both price and fill changes.

Why is this approach sufficient to report signal-lift?

• Answer:

  • Fully key-value–filterable, ensuring signal vs hold-out groups are aligned

  • Includes all monetisation opportunities, capturing real-world performance impact

  • eCPM and uplift % provide a clear, auditable measure of incremental revenue per opportunity

Measurement Summary

• Accuracy: Denominator captures all monetisation opportunities, not just served impressions

• Relevance: Numerator captures AdX revenue only, isolating PPID/PPS effect

• Transparency: Key-value filtering ensures the signal vs hold-out comparison is fully auditable

• Simplicity: Single eCPM metric captures both fill rate and bid-level improvements

• Alignment: Fully aligned with GAM reporting structure; avoids metrics that are not key-value–filterable

• Outcomes: Produces a robust, repeatable, auditable measurement of PPID/PPS signal-lift performance, clearly demonstrating revenue and yield uplift