laptop, movie, popcorn, video streaming, streaming media, streaming service, series, video, media, screen, advertisement, glasses, 3d glasses, movie theater, entertainment, netflix, video streaming, streaming service, streaming service, streaming service, netflix, netflix, netflix, netflix, netflix
Photo by Frank_Rietsch on Pixabay

Household Reach

Part of CTV frequency and household exposure in Australia

Comparing frequency across streaming publishers: 9 steps

Publisher frequency reports are comparable only after their denominators match.

This is an editorial pen name used by the publication, not a claim of an independently verified personal identity. Original editorial credit: Measurement Desk.

Why frequency figures from different streaming publishers will not line up on their own

People move between television, connected TV, digital publishers, social platforms and streaming audio within a single week.

Those channels have historically been measured in separate silos — the pattern Comscore described for streaming audio before adding it to cross-platform reporting.

The publishers in a comparison do not report alike. Hulu, Peacock and Paramount+ sit among the top-tier streaming publishers advertisers buy, while connected TV inventory is fragmented across Netflix, Prime Video, Disney and a long tail of apps that no single buying platform reaches in full.

A frequency figure from one streaming publisher is not comparable with another's on its own.

"Advertisers can't afford to measure channels in silos," Comscore chief commercial officer Steve Bagdasarian said when the company expanded that reporting to include streaming audio and Meta's platforms.

The fix is procedural rather than analytical. Before two publishers' numbers go side by side, agree one unit of count, one reporting window, a record of which metrics each publisher could supply, and one rule for co-viewing and duplicate audiences.

Nine decisions make two publishers' frequency numbers comparable: what is counted, over what window, whether the metric exists, how exposure is distributed, how audiences are deduplicated, how co-viewing is treated, which operational gaps are logged, whether the creative held constant, and how the result is read.

Decide what you are counting: people, devices or impressions

Frequency only means something once you name the population it is divided by. Decide whether a publisher's figure counts people, devices or impressions, and hold that unit for every publisher in the comparison.

In Display & Video 360, unique reach and frequency metrics measure the total number of people shown an ad and how often they were shown it, going beyond basic cookie measurements. Average Impression Frequency per User divides impressions across those people.

Buying platforms do not share a currency either. One TV platform lets buyers switch between incrementality, a view-through built for CTV and a digital view-through comparable to Facebook, and reports reach and frequency alongside CPV, CAC, CPI and ROAS. Mixing those modes compares measurement modes rather than publishers.

A device or cookie count divides the same impressions across screens or identifiers and usually reads higher, because it cannot tell that two devices belong to one person.

Connected TV adds a second distortion: co-viewing occurs when multiple people watch ads together on one connected TV device, so a single screen can deliver more than one viewer.

Fix one unit for the whole comparison and state it in the report.

Fix the reporting window before you compare anything

Most reach and frequency metrics in Display & Video 360 can only be reported for a date range of 92 days or less. Select a longer range and reach data will not appear in the table.

There is also a shorter frequency view. Average Impression Frequency per User (7 days) is listed among the preferred metrics for judging a frequency campaign's performance.

Because the calculations involve modelling, reporting is delayed and early pulls are partial. Re-pull after the data settles rather than comparing a mature figure from one publisher with a fresh figure from another.

Choose a window every publisher in the comparison can support, and note it. A 92-day window and a 7-day window describe different behaviour and should not be mixed in one table.

Check the thresholds that decide whether a number exists at all

Reach and frequency metrics may not appear at all for a given line item.

Display & Video 360 attributes this to factors including the availability of data by country and whether the ad has reached a minimum threshold of impressions and unique users. Its unique reach models do not support every country.

Coverage is market-specific. Nielsen's connected TV measurement of YouTube ads runs in 11 markets — Australia, Germany, India, Indonesia, Italy, Japan, Mexico, the Philippines, South Korea, Thailand and the UK — so one publisher can carry a metric in Australia that another cannot supply at all.

For each publisher, record whether the metric is reported, modelled or unavailable. Treat a gap as missing data rather than zero, or a publisher that cannot report will look like one that reached nobody.

Frequency Measurement Across Key Australian Streaming Platforms

Nielsen CTV Measurement (YouTube)
Available in Australia (via Nielsen One Ads)
Comscore Cross-Platform Reporting
Includes streaming audio and Meta platforms; supports Australia
Google Display & Video 360
Supports unique reach and frequency metrics with deduplication
Hulu / Peacock / Paramount+
Top-tier publishers; may not report same metrics across markets

Compare the distribution, not just the average

An average frequency compresses the whole audience into one number. It hides whether a campaign reached many people once or a few people ten times.

Frequency distribution buckets — 1+, 2+, 3+ and upward to 10+ — show the share of the audience reached at each level of exposure. They are available in Display & Video 360 through pivoted total reach tables.

Report the distribution alongside the mean. Otherwise the heavy-exposure tail disappears into the average, and two campaigns with the same mean look identical when they are not.

Key Metrics for Frequency Comparison in Australia

Reporting Window
92 days (max in DV360); 7 days preferred for performance analysis
Co-Viewing Adjustment
Modelled in Nielsen One Ads (Australia supported)
Deduplication Method
Statistical models using cross-device behaviour and census data
Creative Consistency
Critical — dynamic creatives can skew engagement, not frequency

Strip out the double counting

Deduplication makes cross-publisher totals possible.

Display & Video 360 uses statistical models that account for user behaviour across many browsers and devices. These are built by observing aggregated behaviour across Google products to determine cross-device usage patterns.

Those behaviour observations are combined with other signals and local inputs, such as census and probability surveys, to deduplicate an audience across sessions, formats, networks and devices. The result is the number of unique users, not cookies.

Separate publisher figures become one figure only after this step. Comscore's cross-platform reporting aligns campaign-level metrics from Meta's Facebook, Instagram and Audience Network with television, CTV and digital channels so audiences who see ads in more than one place are not counted twice.

A control-group design is an alternative route to reading cumulative effect. Comscore's 2009 studies compared the behaviour over time of people exposed to display campaigns with a control group who were not, and found lifts in site visitation, trademark search queries, and online and offline sales.

Adjust for co-viewing and where the screen sits in the home

On connected TV, a device-based impression count and a person-based frequency figure diverge whenever more than one person is watching. Display & Video 360's unique reach and frequency metrics account for co-viewing on connected TV devices for that reason.

Co-viewing metrics are now available in some measurement products across multiple markets. Nielsen integrated YouTube CTV measurement, complete with co-viewing metrics, into Nielsen One Ads across 11 markets, Australia among them.

Record for each publisher whether the figure is co-viewed or device-based, and whether the adjustment is modelled or drawn from a direct count.

Normalise the operational differences between publishers

Operational capability varies and belongs in context columns beside the numbers.

Dynamic ad insertion is available on major streaming device platforms but not on all vMVPDs. Where it is available, it may not be supported by fully integrated ad serving, third-party measurement, or the availability of data for targeting.

Publisher structure matters too. Most publishers decentralise their OTT business between digital operations, corporate distribution and licensing, and often between major divisions or brands, which makes coordinated strategy and investment hard.

Log these differences rather than adjusting for them silently. A publisher without third-party measurement cannot be placed on the same footing as one with it, but the gap should be visible in the record.

Hold the creative constant or record how it changed

Creative and format differences show up in the response metrics that sit beside frequency. In Innovid's 2024 Advertising Insights Report for Pharma, interactive CTV advertising delivered an engagement rate almost 2x that of interactive online video, and Haleon's dynamic creative optimisation delivered a 2.5x better click-through rate on YouTube than standard video.

Those are response measures, not frequency measures, which is exactly why they belong in a context column rather than in the frequency figure itself.

Keep creative consistent across publishers where you can.

Where formats must differ, record the variation — format type, interactivity, whether creative was dynamically optimised — so frequency differences are not attributed to the publisher when the creative changed.

Read the result against benchmarks and against effect, not clicks

Vertical benchmarks give context, not targets. Innovid's reports publish average frequency, reach, video completion rates, click-through rates and engagement rates by sector, drawing on impressions delivered by its ad server.

The screen mix has shifted fast: almost half (45%) of Innovid's pharmaceutical video ad impressions in 2023 were CTV, an increase of 36% since 2019.

Clicks are a poor test of frequency. Comscore cited DoubleClick figures showing click rates for display ad campaigns averaging only 0.1%, down from the 3% to 5% of the early digital era.

What frequency is meant to build is the latent effect: the cumulative impact over time of repeated exposure.

Report each comparison with the same short template: unit of count, reporting window, thresholds met or missed, frequency distribution, deduplication method and status, co-viewing treatment, operational caveats, and creative notes.

Fill the template once per publisher — YouTube and Hulu, say — and the two records can be read side by side.

Pre-Comparison Checklist for Frequency Data

  • Unit of count fixed?Yes – e.g., unique people
  • Same reporting window used?Yes – e.g., 7-day or 92-day
  • Thresholds met for all publishers?Confirmed via metric availability logs
  • Frequency distribution reported?Yes – including 1+, 2+, 3+ buckets
  • Co-viewing adjusted?Yes – modelled or directly counted in supported markets
  • Creative held constant?Yes – or variation recorded

More in Household Reach