| Situation | Viewable condition | Planning use |
|---|---|---|
| Standard display ad | At least 50% of the ad size on screen for 1 second or more | Use as the default test when judging whether a display placement was viewable. |
| Very large display ad (over 242,500 pixels) | At least 30% of the ad area visible for 1 second | Judge big formats against their own rule, not the half-size test. |
| Video ad | 50% of pixels on screen for 2 seconds | Compare video results with other video results, not with display. |
| Clicked ad | Counts as viewed even when the time or pixel thresholds are missed | Check click-heavy placements against viewable rates separately. |
The display and click conditions follow Optad360’s viewability guide, and the video condition follows AdLibrary’s glossary entry.
What is an ad impression? It is one counted instance of an ad being served, while a viewable impression is one that also had a chance to be seen. The gap between the two drives most reporting decisions.
Impressions sit inside the wider system covered in this introduction to what is ad tech, where ad servers deliver and count ads.
What the thresholds change in practice
A served impression and a viewable impression are different counts, so the figure in a campaign report depends on which one is being reported. The choice affects how a placement is judged, what is compared and what is trusted.
Under the Optad360 guide, a standard ad counts as viewed when at least half its size appears on screen for one second or more. Ads over 242,500 pixels need only 30% of their area visible for the same time. ExchangeWire describes viewability as the share of an ad unit’s pixels in the active viewport divided by the unit’s total pixels. It stresses that the measurement reports whether an ad had the opportunity to be seen.
- Buying on viewable impressions. The viewability rate is viewable impressions divided by measured impressions, multiplied by 100. A low rate means many served ads never met the threshold, so cost per served impression flatters the placement.
- Large formats. Because ads over 242,500 pixels use the 30% rule, a tall or wide unit should not be marked down by the half-size test.
- Confirm whether a video figure is a plain impression or a viewable one before comparing it.
- Click-driven campaigns. A clicked ad is treated as viewed even when the thresholds are missed, so click-heavy placements can look better on viewed counts than their visibility alone would justify.
Impression versus view
An impression records that an ad was served, whereas a view records that the ad met a visibility condition. The two counts answer different questions, and treating them as interchangeable is the most common reporting mistake.
ExchangeWire notes that viewability measurement aims to report whether an ad had the opportunity to be seen. The Optad360 guide adds that viewability focuses on that opportunity, while attention metrics go further. A viewable impression is therefore not evidence that a person read or noticed the ad. It suits comparing the visibility of placements, not judging the quality of the creative.
Click-through rate sits on top of these counts. It relates clicks to impressions, so choosing served or viewable impressions as the base changes the rate for the same campaign. When two reports show different click-through rates, check the impression base first.
Whether the traffic behind an impression is genuine is a separate check, covered in what is ad fraud.
Why ad server and publisher counts differ
Ad server and publisher counts differ mainly because each system applies its own counting rule at its own point in the delivery chain. A mismatch is not automatically an error.
The MRC reconciliation guidelines are designed to minimise counting discrepancies between vendors. ExchangeWire says following them should bring the discrepancy range on standard formats to around 5%. A gap near that level is normal, and a much wider gap points to a definitional difference worth investigating.
Definitions can differ sharply. Digiday reported that Pinterest counted a display impression when at least one pixel of an ad was on screen for at least one second. That is far looser than the half-size standard. In June 2021 the MRC denied accreditation for Pinterest’s video impression metrics, saying they did not meet industry standard minimum requirements. Digiday also notes that the MRC oversees audits by outside firms such as Deloitte and EY, and that video impression metrics were approved for at least a dozen ad servers and verifiers.
Wording on the report matters too. The Optad360 guide describes a publisher dashboard that shows the number of ads Google counts as served and viewed. Read the column label before comparing it with a buyer’s figure.
- Match the base first. Confirm whether both reports count served impressions or viewable impressions.
- Match the threshold. A report using a looser visibility rule will always show more counted impressions than one using the standard rule.
- Prefer audited measurement for disputes. Where one party uses MRC-accredited counting, use it as the reference point.
Following one impression from request to count
A single impression passes through several stages, and different systems record different stages. The served count comes from the ad delivery side, while the viewable count comes from a measurement tag running in the page.
- Request. The page or app asks an ad server for an ad to fill a slot.
- Serve. The ad server returns an ad and the impression is counted as served. This is the figure behind the served side of a publisher dashboard.
- Render and track. According to the AdLibrary glossary, a measurement tag runs client-side JavaScript to track what percentage of the ad unit is inside the active viewport.
- Timer. The glossary describes the timer as starting once the visibility threshold is reached. If the viewer scrolls away before the required time passes, the impression stays served but never becomes viewable.
- Report. Served and viewable totals feed the viewability rate described above. A click at any point counts the ad as viewed.
The practical result is that an ad placed far down a page can be requested and served yet never meet the viewable condition. Its served count and viewable count then disagree for a legitimate reason, which is the same pattern behind many cross-report gaps.
Frequently asked questions
Does a viewable impression prove someone looked at the ad?
No. Viewability reports whether the ad had the opportunity to be seen, and attention metrics go further, according to Optad360 and ExchangeWire.
Does the MRC require the viewable condition for a video impression?
According to Digiday, the viewable condition is not required by the MRC for a video ad impression, so check whether a video figure is a plain impression or a viewable one.
What viewability rate should a publisher aim for?
Optad360 says publishers should aim for viewability rates above 70%, as set by leading supply-side platforms. It cites Statista for a mobile app display average of 83.8% in the second half of 2023. Treat these as benchmarks, not guarantees.
What is the Open Measurement SDK?
Optad360 describes Open Measurement (OM) SDK as an Interactive Advertising Bureau standard that streamlines third-party access to measurement data.
