An impression can be delivered, measured as viewable and still produce no useful response. Another can reach a real person without meeting a viewability threshold. These are different observations, and treating them as one verdict makes campaign decisions harder.
A useful quality review separates four questions: was the activity legitimate, could the ad be seen, did someone engage, and did the campaign achieve its purpose? Each requires different evidence. Start with those questions before comparing suppliers, changing placements or increasing a budget.
Give each metric a job
“High-quality traffic” is too broad to serve as a measurement plan. Translate it into the specific evidence your team needs. This prevents an inexpensive click, a high viewability rate or a successful sale from becoming an excuse to ignore the rest of the journey.
| Question | Useful evidence | What it cannot establish alone |
|---|---|---|
| Is the activity valid? | Invalid traffic detection and filtration | Whether a person noticed or valued the ad |
| Was there an opportunity to see it? | Viewability and measurement coverage | Attention, understanding or interest |
| Did someone engage? | Clearly defined interactions or on-site behaviour | Whether that interaction created business value |
| Did it meet the objective? | Qualified actions, verified purchases or another agreed outcome | Whether advertising caused every recorded outcome |
Choose a primary outcome, then keep the other measures as diagnostics. A publisher may investigate a poorly visible placement; a buyer may investigate weak landing-page engagement. They can use the same report without needing the same decision.
Invalid traffic is not just bots
Google’s AdSense definition of invalid traffic includes activity that artificially inflates advertiser costs or publisher earnings. It covers automated activity, but also examples such as accidental clicks and artificially encouraged interactions. Human activity is therefore not automatically valid, and an absence of sales is not evidence that traffic is invalid.
The MRC’s invalid traffic standards distinguish general invalid traffic, identified through routine filtration, from sophisticated invalid traffic, which requires more advanced detection. Those categories describe detection needs, not a simple ranking of campaign performance.
When reviewing a supplier’s report, ask which kinds of invalid traffic it addresses, which environments it covers and when filtration happens. Keep the reporting period and version: later adjustments may change the comparison. If a suspicious pattern appears, preserve the evidence and ask for investigation rather than labelling an entire source fraudulent on the strength of a low conversion rate.
Read viewability as an opportunity
The MRC’s desktop viewability guidelines describe an opportunity to see an ad, using visible area and time in an in-focus browser tab. They do not certify that a person looked at the creative. The guidelines also require invalid activity to be excluded appropriately; a raw pixel-and-time test is not a substitute for filtered measurement.
Thresholds depend on the format and measurement framework. For example, Google Ads describes these Active View thresholds for supported inventory:
- Standard display: at least half of the ad’s area is visible for at least one continuous second.
- Large display: ads of at least 242,500 pixels use a 30% area threshold for at least one continuous second.
- Video: at least half of the ad’s area is visible for at least two continuous seconds while playing.
Do not apply that list indiscriminately to every device, format or platform. Confirm the definition attached to the report you are reading. Meeting a minimum threshold is a starting condition, not proof that the creative communicated its message.
Check what could actually be measured
Before celebrating a viewability percentage, inspect its denominator. In Google Ad Manager’s Active View reporting, measurable impressions are those eligible impressions for which measurement succeeded. Viewability is calculated from measurable impressions. An impression that could not be measured is not automatically counted as non-viewable.
This distinction changes comparisons. A source with limited coverage might show an attractive rate for the subset that could be measured. That does not tell you how the missing portion behaved. Ask whether differences in environment, implementation or supported inventory explain the gap.
Keep measurement coverage beside viewability in your reporting template. When either changes sharply, investigate the reporting setup and placement mix before treating the movement as a change in audience quality.
Work through a fictional report
Imagine a fictional campaign with consistently filtered counts: 10,000 eligible impressions, of which 8,000 are measurable and 5,200 are viewable. Using the Active View definitions above, the figures mean:
- 80% measurability: 8,000 divided by 10,000.
- 65% viewability: 5,200 divided by 8,000.
- 52% of eligible impressions measured as viewable: 5,200 divided by 10,000.
The remaining 2,000 eligible impressions are unmeasured. Another 2,800 were measured but did not meet the viewability threshold. Combining those two groups into “people who never saw the ad” would go beyond the evidence.
Now suppose the report includes clicks and purchases. Those counts answer additional questions; they do not explain the unmeasured impressions. Analyse the campaign by comparable placement and device groups, then investigate the specific stage where performance changes.
Keep attention, engagement and outcomes distinct
The IAB’s attention measurement guidance treats attention as a complementary signal, considered alongside delivery and outcomes. Ask how a proposed attention metric is produced: direct observation, modelling, interaction signals or a combination. Understand its coverage before comparing it with a differently named score.
Analytics labels need the same care. Google Analytics defines an engaged session through configured duration and event conditions. That is a product-specific classification, not a universal definition of a person’s interest. A session can qualify without a purchase, and a purchase can have influences outside the ad being reviewed.
For practical decisions, connect the evidence to your objective. If a campaign generates clicks but few meaningful visits, examine the experience after the click, tracking and offer before changing every placement. If visits are useful but purchases lag, inspect the purchasing journey separately.
Turn a quality review into a decision
- Confirm comparable reporting. Align periods, currencies where relevant, formats and counting rules.
- Review filtration. Record which invalid traffic checks the report includes and whether figures are provisional.
- Pair coverage with viewability. Keep unmeasured impressions visible in the discussion.
- Inspect useful behaviour. Define meaningful visits or actions before looking for a winner.
- Test one explanation. Change a placement, creative or landing-page element with a clear hypothesis and a limited scope.
Document the next action and the evidence that would change your mind. A repeatable testing process is more useful than chasing a single quality score that tries to answer every question at once.
Sources & further reading
Explore advertising with Promoweb.
Meet our self-serve platform