Completion rate is not attention.
That sentence does more work than it looks like. Most CTV campaigns today are evaluated on completion rate — what fraction of the creative played all the way through. CTV passes that test almost universally. Which is precisely why it has stopped telling you anything.
This post sets out what actually differentiates attention across CTV, non-CTV video and display, using data from 5.1 million delivered impressions across two Indian campaigns, and explains how to apply that thinking to a cross-device campaign.
TL;DR
What is attention measurement in advertising?
Attention measurement is the practice of estimating how much active human focus an ad impression actually received, expressed in time or in the probability of being seen or heard, rather than in delivery. It sits between delivery metrics like impressions and viewability — which confirm an ad was served — and outcome metrics like recall and conversions, which confirm it worked. Attention is the missing middle step that explains why two campaigns with identical delivery produce different results.
The industry moved here for a practical reason. As budgets shifted to connected screens, the old proxies stopped discriminating between good and bad inventory. A CTV impression is almost always fully viewable and almost always completes. When every placement scores near-perfect on your quality metric, that metric has stopped telling you anything.
Why impressions, viewability and completion stopped being enough
What viewability actually confirms
Viewability confirms a technical opportunity to see. For connected TV, the standard requires half the ad's pixels to be in view for at least two seconds — and because CTV creative renders full-screen by default, most CTV inventory is treated as fully viewable.
That is a floor, not a signal. It tells you the ad rendered. It cannot tell you whether a person was in the room, facing the screen, or awake.
Why CTV breaks traditional assumptions
Three properties of the living room make CTV structurally different from every channel before it.
**There is no click.** CTV has no engagement proxy. On mobile you can at least infer interest from a tap. On a television the feedback loop between delivery and outcome is missing entirely — which is the gap attention metrics were built to fill.
**One impression is not one viewer.** A CTV ad served into a room with four people counts as one impression. TVision, which measures second-by-second engagement with TV and CTV, tracks this as Viewers Per Viewing Household, and has found that people watching with at least one other person pay more attention than people watching alone. Co-viewing simultaneously under-counts your reach and understates your attention.
**The ad plays whether or not anyone is watching.** Wurl found that roughly 4% of CTV viewers leave each minute of an ad break, meaning a three-minute pod loses around 12% of its audience by the end — while every one of those impressions still reports as a completed view.
Attention vs engagement vs viewability: what each metric measures
These terms get used interchangeably in media plans. They measure genuinely different things.
| Metric | What it measures | Unit | What it cannot tell you |
|---|---|---|---|
| Viewability | Whether the ad had a technical opportunity to be seen | % of impressions | Whether a human was present or looking |
| Completion | Whether the creative played to the end | % (VTR / LTR) | Whether anyone watched it play |
| Audible-and-visible delivery | Share of play time both on screen and with sound on | % of play time | Whether a human was in the room |
| Attention | Estimated active human focus on the ad | Seconds, or a 0–100 score | Whether the message was understood |
| Engagement | Deliberate user action on or after the ad | Clicks, QR scans | Passive impact on brand memory |
| Recall / lift | Whether the ad changed memory or intent | Survey-based % | Which placement caused it |
The practical consequence: viewability and completion are hygiene checks you should pass and then stop optimising against.
What the data says: completion rate hides the real gap
Here is the case for why this matters, using Visora's own campaign data rather than an industry average.
Across two campaigns — one financial services advertiser, one health and wellness advertiser, roughly 5.1 million delivered impressions between them, running CTV, non-CTV video and display across metro and Tier-2 India in 2025-26 — the completion picture looked like this:
| Format | Completion rate (both campaigns) |
|---|---|
| CTV / OTT | 90–93% |
| Non-CTV video | 64–66% |
| Display | n/a |
A media buyer reading that table concludes CTV is about 1.4x better than non-CTV video and moves on. But the financial services campaign also carried IAB Active View data, which reports whether the ad was actually on screen with the sound on. That tells a different story:
| Exposure-quality metric | CTV | Non-CTV video |
|---|---|---|
| Completion rate | 90.28% | 65.6% |
| Average viewable time | 27.9 sec | 23.1 sec |
| Audible and visible at start | 93.5% | 48.9% |
| Play time visible | 99.9% | 75.1% |
| Play time audible and visible | 99.3% | 63.1% |
| Audible-and-visible seconds delivered | 27.7 sec | 14.6 sec |
Basis: single advertiser, ~1.01M video-format impressions, upto-30-second non-skippable unmuted creative, 2026 flight. Audible-and-visible seconds derived as average viewable time multiplied by % play time audible and visible.
Three things fall out of that table.
Fewer than half of non-CTV video impressions began in a state where anyone could both see and hear them. 48.9% were audible and visible at start — yet 65.6% of them went on to record a completed view. A paid impression can complete while being neither properly seen nor properly heard, and completion rate will never tell you.
The ranking survives but the magnitude collapses. On completion, CTV is 1.4x better. On audible-and-visible seconds actually delivered, CTV is 1.9x better. Completion understated the true quality gap by roughly 38%. If you were reallocating budget on completion rate, you were under-weighting CTV by a third.
CTV's advantage is consistency, not just level. In the financial services campaign, CTV completion across pincodes in a single metro ranged from 92.5% to 96.5% — a spread of under four percentage points. Non-CTV video in the same campaign ranged from 43% to effectively 100%, a spread above 50 points. On CTV your risk is whether anyone was in the room. On non-CTV video your risk is where it ran. Those are different failure modes and they need different metrics.
To see how this applies to a live media plan, explore Visora's CTV advertising in India.
What metrics actually measure attention?
There is no single industry-standard attention metric yet — the IAB and MRC have published measurement guidelines, but each vendor still reports on its own scale. These are the metrics you will encounter.
If you adopt nothing else, adopt audible-and-visible play time. It requires no new vendor, it appears in reporting you already receive, and as the table above shows, it separates placements that completion rate rates identically.
How attention measurement differs by environment
Blending attention into one campaign average destroys the signal. Each environment produces attention differently and fails differently.
| Environment | Primary attention signal | Most common false positive |
|---|---|---|
| CTV / OTT video | Completion + audible-visible play time | Full completion into an empty room |
| Non-CTV video | Audible and visible at start | Muted autoplay recorded as a view |
| Digital audio / podcast | Audible completion, listen-through | Autoplay in the background |
| Smart speaker audio | Completion, post-ad voice query | Device on, no listener present |
| Mobile / laptop display | Dwell time in view | Long in-view time on an idle tab |
| pDOOH | Modelled dwell + footfall | Traffic counts treated as viewers |
How to measure attention across a cross-device campaign
Single-channel attention is a solved problem you can buy. Cross-device attention is where campaigns quietly go wrong, and it comes down to four decisions.
**1. Define attention once, before launch.** Lock one definition and one vendor methodology across every channel. Vendors measure differently and report on different scales, so a panel-based CTV number and a modelled display number cannot be added or averaged. Make this a line item in the media plan, not a post-campaign question.
**2. Deduplicate at the household level, not the device level.** A home that sees your CTV spot in the living room and your mobile pre-roll on the sofa is one reached household, not two reached users. Device-level counting inflates reach and hides frequency.
**3. Plan on attentive reach, not impressions.** Reallocate against audible-and-visible seconds rather than raw completion. In the campaign above, that single change moves CTV's relative value up by about a third.
**4. Optimise mid-flight, weekly.** Pull exposure quality by app, daypart, pod position and geography every week and cut the bottom decile before it spends through to full delivery.
Attention benchmarks: what the numbers actually mean
There is no single published attention benchmark for Indian CTV inventory. Any number on a benchmarking slide without a stated methodology and basis cannot survive contact with a real media plan.
What you can benchmark is relative performance between channels. On Adelaide's Attention Unit scale, connected TV has been rated in the high 50s, ahead of linear television in the low 50s, online video in the high 30s and standard display in the low 20s — roughly three times the attention capacity of display. Integral Ad Science, using Lumen's eye-tracking data in its Quality Attention product, has reported that high-attention impressions can deliver conversion lifts of up to 130% against low-attention impressions.
Against that industry backdrop, here is what Visora observed across the two campaigns described above:
| Format | Completion rate | Play time audible and visible | Audible-and-visible seconds |
|---|---|---|---|
| CTV / OTT | 90–93% | 99.3% | 27.7 sec |
| Non-CTV video | 64–66% | 63.1% | 14.6 sec |
| Display | n/a | n/a | n/a |
Basis: two advertisers (financial services; health and wellness), ~5.1M delivered impressions, metro and Tier-2 India, 2025–26. Completion range spans both campaigns. Audible-visible figures are from the financial services campaign only, measured via IAB Active View on upto-30-second non-skippable unmuted creative.
Use these as a directional reference for Indian CTV and video inventory, not as a target. The methodology note matters as much as the numbers — a benchmark without a stated basis is a marketing claim, not a measurement.
Where this data stops, and what we are adding
None of the above is eye-tracked attention. Active View audible-and-visible delivery confirms an ad was on screen with sound on; it cannot confirm a human was looking at it. It is a materially harder test than completion and it is available today in reporting most advertisers already receive — but it is a proxy.
Two gaps remain, and they are the honest limits of this dataset:
Visora's Echosphere connects CTV, mobile, laptop display and smart speaker audio under a single household identifier, which is the precondition for measuring exposure across screens rather than adding up incompatible per-channel reports. In the financial services campaign above, delivery spanned mobile, desktop, connected TV, tablet and smart home devices in one flight — and the household-level view is what makes those five device classes one audience rather than five.
That matters more here than in most markets. Kantar's Media Compass report put Indian CTV reach at 166 million monthly viewers aged 15 and over in the first quarter of 2026, up 23% year on year, with 59 million of those viewers — around 36% — watching connected TV exclusively and unreachable on linear television at all.
Five mistakes that make attention data useless
Frequently asked questions
Is viewability the same as attention?
No. Viewability confirms an ad had a technical opportunity to be seen — for CTV, half the pixels in view for two seconds or more. Attention estimates whether a person actually looked. Because CTV renders full-screen, most CTV inventory is treated as fully viewable, which is precisely why viewability cannot separate good CTV placements from bad ones.
Is completion rate a good measure of attention?
No, and the gap is measurable. In Visora campaign data, non-CTV video recorded a 65.6% completion rate while only 48.9% of those impressions were audible and visible at start. Completion tells you the file played. It does not tell you the ad was on screen, unmuted, or in front of anyone.
How is attention measured on CTV?
Through panel measurement and modelling. Panel providers use opt-in devices in real homes to capture presence and eyes-on-screen data, which then trains models that score placements at scale using signals like app, content genre, daypart and pod position. Most advertisers buy a modelled score validated against panel data rather than measuring directly.
What is a good attention rate for a CTV campaign?
There is no universal figure — scores are vendor- and environment-specific. Benchmark against your own campaigns using one consistent methodology. As a directional reference, Visora has observed CTV completion of 90–93% with over 99% of play time audible and visible, against 64–66% completion and 63% audible-visible play time on non-CTV video.
Can you measure attention on audio ads?
Yes, with different metrics. Audio has no eyes-on-screen equivalent, so attention is proxied through audible completion rate, listen-through rate, and for smart speakers, post-ad voice interactions. Benchmark audio against audio.
Does co-viewing count as attention?
Co-viewing affects how many people your attention applies to, not whether attention occurred. Two people watching attentively is one impression delivering two attentive exposures. TVision research indicates co-viewers pay more attention than solo viewers, so co-viewed inventory can be undervalued on impression-based pricing.
Do you need eye-tracking to measure attention?
Not to start. Audible-and-visible play time is already in standard Active View reporting and separates placements that completion rate scores identically. Eye-tracking panels underpin the more advanced attention models, but the first improvement over completion rate costs nothing new.
What is the difference between attention and engagement?
Attention estimates how much active human focus an ad received, measured in seconds or as a 0–90 score. Engagement is a deliberate action on or after the ad, such as a click or a QR scan. CTV has no click, so engagement says little about a CTV ad. Attention is the more useful signal on the big screen.
What is attentive reach?
Standard reach counts everyone an ad was served to. Attentive reach counts only the people estimated to have actually paid attention to it. Two campaigns with the same reach can deliver very different attentive reach, depending on where and how the ads ran.
How do you benchmark CTV ad inventory for attention instead of viewability?
Viewability cannot separate CTV placements, because almost all CTV ads play full-screen. Compare placements on audible-and-visible play time, which is already in standard Active View reporting. If you use an attention vendor, compare attentive seconds per thousand impressions (APM) and attentive CPM. Always benchmark CTV against CTV, using one consistent method.
Key takeaways
Want to see audible-and-visible delivery across CTV, video and display in one deduplicated household view?
