Measurement reference

Know what each number counts.

Compare reports using the same event, denominator, date range and currency. The definitions below explain common calculations; your approved platform and agreement determine the actual reporting fields.

Advertiser metrics

Common calculations and their limits
MetricCalculation or meaningKeep in mind
ImpressionsCounted instances of an ad being served or rendered, according to the reporting system.An impression count is not automatically a count of people or viewable exposures.
ClicksRecorded interactions treated as clicks by the reporting source.Agree filtering and duplicate handling; a click is not a completed business action.
CTRClicks ÷ impressions × 100.Compare like formats, placements and audiences.
Average CPMAdvertising spend ÷ impressions × 1,000.Keep currency and the billable impression definition consistent.
Average CPCAdvertising spend ÷ clicks.An average calculated from a report differs from a configured maximum bid.
Click-based CVRAttributed conversions ÷ eligible clicks × 100.Some systems use other interactions as the denominator. Label which base you used.
CPAAdvertising spend ÷ the defined attributed actions.Specify whether an action is submitted, qualified, paid or retained.
ROASAttributed conversion value ÷ advertising spend.Express as a multiple or percentage. Revenue-based ROAS does not account for all costs or equal profit.
ReachUnique audience count under the system’s identification and measurement method.Cross-device identification and coverage affect the count.
FrequencyImpressions ÷ measured reach, when both use a compatible scope.An average can hide people with very high or very low exposure.

Publisher metrics

Delivery, yield and measurement bases
MetricMeaningReview question
Ad requestsRecorded requests for advertising in a defined placement and environment.What makes a request eligible and does one request contain several opportunities?
Filled requestsEligible requests for which an ad response was returned under the chosen reporting definition.Does the report count responses or creatives actually rendered?
Request fill rateFilled eligible requests ÷ eligible requests × 100 in the example below.Does your platform instead calculate fill using impressions or a different opportunity base?
Impression eCPMPublisher revenue ÷ counted impressions × 1,000.Is revenue estimated or finalized, and gross or net of the agreed deductions?
Viewability rateViewable impressions ÷ measurable impressions × 100 under the measurement rules used.Which impressions were measurable, and which format-specific standard applied?
Unfilled opportunitiesEligible opportunities without a filled response under the selected definition.Was the cause demand, targeting, a floor, format compatibility or an integration error?

Worked examples with invented numbers

These examples teach arithmetic. They are not observed GaiaAds performance, a forecast, a customer case study or a promised return.

An advertiser report

Assume USD 1,000 spend, 100,000 impressions, 2,500 clicks, 125 attributed purchases and USD 2,500 attributed purchase value, all under one agreed scope.

CTR
2.5%
2,500 ÷ 100,000 × 100
CPM
USD 10.00
1,000 ÷ 100,000 × 1,000
CPC
USD 0.40
1,000 ÷ 2,500
Click-based CVR
5%
125 ÷ 2,500 × 100
CPA
USD 8.00
1,000 ÷ 125
ROAS
2.5× / 250%
2,500 ÷ 1,000

The 2.5× revenue-based ROAS does not show profit: product, fulfillment, returns and other costs are not included in that calculation. If purchases are not settled or attribution is inconsistent, the comparison also changes.

A publisher delivery report

Assume 150,000 eligible requests, 100,000 filled requests, 90,000 rendered impressions and USD 600 publisher revenue.

  • Request fill rate: 100,000 ÷ 150,000 × 100 = approximately 66.7%.
  • Impression eCPM: USD 600 ÷ 90,000 × 1,000 = approximately USD 6.67.
  • Response-to-render gap: 10,000 filled requests did not appear as rendered impressions in this simplified example. Review counting rules and delivery behavior before explaining the gap.

For a zero denominator, show “not available” rather than manufacturing a rate. For an uncertain event definition, resolve the definition before presenting the number as comparable.

Reconcile two reports in a repeatable order

  1. Scope

    Match account, campaign, placement, markets, devices and date range.

  2. Time basis

    Align time zones and determine whether conversions are grouped by interaction time or event time.

  3. Event rules

    Check attribution window, view-through versus click-through inclusion, deduplication and invalid-event filtering.

  4. Money basis

    Align currency, fees, gross versus net treatment and estimated versus finalized revenue.

  5. Data maturity

    Allow for agreed reporting delays and changes such as cancellations or validation adjustments.

A platform report may attribute an action without proving that advertising caused an additional action. If incrementality is the question, discuss a suitable comparison or study design separately from routine attributed reporting. Google’s conversion lift overview explains this distinction.