Upwave is trying to bring a Google Analytics feel to brand metrics measurement – less like post-campaign research consultancy and more like hands-on, inflight daily insights into brand awareness and sentiment using a software solution that can also support top of funnel optimisation. “If we do our job, we make brand campaigns feel like performance campaigns, which you can constantly look at and optimise every hour in some channels. Brand was missing that,” says Kelly.

The next step in this ‘brand-treated-like-performance’ approach is the use of programmatic to optimise brand campaigns. Upwave has generated a persuadability score that can be applied to a brand and its favoured KPIs at a household level. Currently in laboratory testing with partners, this gives a household a score between zero and one based on how likely it is that consumers can be persuaded by a brand or a message. The score spins out of data analysis and modelling that is underpinned by a combination of deterministic exposure data (including ACR or set-top box data) and consumer opinion survey data.

The programmatic buying will be optimised against the persuadability score (with Upwave feeding the data to the DSP – the company does not buy or sell media itself). “Based on how persuadable a household is, a decision can be made on what to bid [to win an ad spot].” Upwave’s ambition is that programmatic buys can be optimised for brand outcomes as well as for conversion.

Upwave has industrialised consumer surveys to generate the data that helps it understand brand impact and persuadability, and this includes full automation. Kelly says the company surveys millions of consumers every year. “We don’t maintain a panel; there is no single group of consumers. For every campaign we find a large representative sample of consumers to get their opinions,” he explains. For any campaign, tens of thousands of people are surveyed, including both exposed and non-exposed homes.

Upwave does not try to link brand-level outcomes to actual business outcomes, leaving others to fuse these data sets if they want to. But the brand-level outcomes can be used as a predictor, by showing propensity to become a customer, Kelly claims. When Upwave conducted some tests to gauge the accuracy of its uplift-to-customer assumptions, using grocery sales data, the connection was very accurate, he notes. “We believe we can be used to forecast sales, though we are not tracking sales ourselves. Our early tracking with sales data has been very accurate.”

Kelly says the TV industry is spending a lot of time worrying about audience measurement and currency but needs to put as much effort into demonstrating outcomes – and that includes top of funnel outcomes, where his company specialises. “If a CMO wants to protect their TV budget, what do they want to show: currency statistics proving that they got what they paid for [they reached their target audience] and they were not ripped off, or metrics showing their top of funnel outcomes and that it was effective advertising?” he asks.

Returning to his American Football analogy, Kelly explains: “When a player scores in the end zone, that is a sales conversion. But what about all the plays that get you down the field into the end zone? We measure the movement down the field.” He points to the potentially huge impact of brand advertising, which at its best generates demand and increases lifetime value of customers you do secure, which pays back over years (like when people churn from other brands but stick with yours).

The long-term impact of brand advertising may be understood but when the CFO knocks on the door, CMOs may need some answers today. “Within a short timeframe, we can accurately forecast that you are generating the basis for customer lifetime value,” Kelly claims.

The paper, entitled: ‘Addressable TV: Digitalisation of an offline media’, is available to download now. It is the result of interviews with key players in the French addressable ecosystem including executives from Mindshare, Altice Media Ads & Connect, Gamned, Novembre, Canal+ Brand Solutions, M6 Publicité, SNPTV, Values, Jellyfish, Bouygues and France Televisions Publicité. This is the first time the annual white paper has been published in English and it is a good read for anyone that wants to understand the French addressable market since the law was changed in 2020 to permit household targeting.

The paper declares that, “Addressable TV has brought television into the digital age, sparking a renewed debate about its position in advertisers’ media mixes. This development allows the all-digital world to incorporate TV into its strategies, bridging the gap between a traditionally offline medium and the digital world.”

According to the SNPTV, the French TV advertising advocacy group, there were more than 1,300 addressable TV campaigns in France during 2022 – a three-fold increase over the previous year (the 2021 figure was 376 campaigns). Developments impacting reach include permission for prime-time ad substitution and the introduction of multiple targeted ads per ad break (in a market that opened with a rule allowing only one per break), while work is underway on a unified addressable TV solution that will get smaller TV sales houses into the game. Measurement is one of the big priorities ahead, and there is significant interest in what CRM onboarding (so use of advertiser data matching) can achieve and how that is implemented.

The paper, which is authored by Realytics and extensively quotes the interviewees, believes France now serves as a model for the rest of Europe when it comes to collaboration as the basis for addressable TV market growth. It points out that France is the only European country where all television channels and ISPs (telcos/service providers with Pay TV) have collaborated and agreed upon common specifications for addressable TV implementation. This includes clear rules and unified reporting for advertisers and agencies, and technology alignments to ensure a seamless viewer experience. SNPTV and AF2M have overseen the development of consistent standards. France has a standardised CPM buying method and uniform targeting, the paper observes.

Five of the TV sales houses (TF1 Publicité, France TV, M6 Publicité, Canal+ Brand Solutions, Altice Média Ads & Connect) have signed with Orange, SFR, and Bouygues Telecom. M6 Publicité is already in the process of signing with Free. Addressable TV can be bought direct or programmatically. Third-party players are heavily involved in this ecosystem to support data (e.g., LiveRamp), performance measurement (e.g., Realytics) and programmatic buying (e.g., Adkymia and TheTradeDesk).

The paper outlines how addressable TV is being used in France today, with targets based on geolocation, household characteristics (e.g., income, family with children) and consumption (heavy to light TV users). By 2021, half of all campaigns used geolocation (and the figure was 52% in H2, 2022, and is expected to be 50% in 2023) with one-quarter based on household characteristics or TV usage.

Rapid growth has been recorded since 2021. Olivier Roberdeau, Head of Multiscreen at Mindshare, is quoted from the previous year’s white paper as he recorded how substitutions grew from 0.5% in January 2021 to 10% in January 2022. By 2022 addressable was gaining traction among traditional TV advertisers as well as new-to-TV advertisers, as witnessed by Isabelle Vignon, General Delegate for Marketing at SNPTV. One executive says the priority is to bring more of the 2,000 existing TV advertisers into this new market. Half of addressable TV advertisers in France are still new to TV, as of 2022.

By 2023, half of addressable TV advertisers came back for more, according to Nathalie Dinis, Deputy Executive, Director of Trade at France Télévisions Publicité. Five sectors stood out as addressable TV users in 2022, the paper says: tourism, retail, services, culture & leisure, and automobile. Beyond geo-targeting, use-cases include reach extension beyond linear TV (including into light TV households).

The white paper notes the importance of measurement. Since March 2021, Médiamétrie has been re-evaluating the corrected Gross Rating Points of broadcasted advertisements that were impacted by addressable TV spots. This recalculation is based on the number of impressions served during these spots.

Mindshare’s Roberdeau is quoted saying: “Advertisers treat addressable TV like digital media and anticipate receiving data similar to digital campaigns to maximise their investments across campaigns. In this regard, the BEE measurement study developed by Realytics presents an excellent opportunity”.

As the paper explains, with BEE, Realytics can deterministically measure the performance of national linear TV, addressable TV, IPTV and Replay TV campaigns in households with Bouygues Telecom TVs. The purpose is to show how complementary different TV-activated devices are, in terms of both audience and online engagement. This includes measurement of audience conversion into website or mobile app activity. In 2022, BEE measured the effectiveness of addressable TV campaigns, including when combined with either a national linear TV campaign or an IPTV Replay campaign.

The white paper suggests that the effectiveness of addressable TV will come under the microscope in 2023, with most sales houses focused on demonstrating the impact via test campaigns conducted near the end of 2022. TV sales houses and ISPs (Pay TV operators) are collaborating in these tests.

Another focus for 2023, and looking ahead, is the use of advertiser CRM data to enhance targeting and personalisation. This first-party data matching comes with both technical and legal hurdles (including the need for new consents), and one executive points to the need for large customer bases to even begin this journey. Additional costs, and therefore higher CPMs, may also need to be factored in. Nevertheless, the paper reports that 2022 saw a significant increase in discussions around this subject.

As of 2022, France Televisions Publicité was offering more addressable TV inventory than BVOD, and between H1 2021 and H1 2022 the French addressable customer base grew by 54% to 6.3 million households, the paper reports. Free’s entry into addressable TV could potentially add 6.9 million more subscribers to the audience pot.

The white paper is 71 pages long, so this summary barely touches the surface. If you want to get the deep-dive, you can download the English language version here. It includes full interviews with all the contributors, in addition to analysis of all the research. The paper argues that ‘addressable TV is strengthening the convergence between TV and digital advertising. The brand-safe environment and high completion rates with digital metrics are reviving the TV medium. TV is positioning itself as a valuable complement to video strategies.”

ViewersLogic claims that its panel-based, single-source data represents a revolution in measuring media effectiveness. On its website, it declares: “Reach, frequency, attention and emotional response are inaccurate proxies that are used to try to answer the real question – how did people change their behaviour after being exposed to your campaign? Stop measuring uncorrelated proxies and start measuring the actual uplift of your campaign.”

It gives examples of real outcomes, like product purchase, sign-up or website visits. Sales lift, and incremental app store visits and app downloads, can be linked back to TV ad exposure. As well as helping to demonstrate campaign ROI, the ViewersLogic technology shows how each cross-platform channel contributed to success.

The company offers a long list of how exposure measurement and/or advanced attribution can be used to boost media planning and campaign strategy. These span customer targeting, competitor customer targeting, inflight media mix optimisation, assessing competitor campaign performance, benchmarking against past campaigns, benchmarking against competitor campaigns, and more. All of this is underpinned by some smart technology for demonstrating when someone is in front of a television when an ad is played, and inventive use of GDPR to generate a waterfall of precious, free purchasing data.

ViewersLogic recruits its panellists online and they download the ViewersLogic app to their smartphones. They are incentivised to give access to their data: data collection is rewarded with points (the more data, the more points) and these are turned into Amazon vouchers. In return, the phone app is permitted to sync with a set-top box and track channel changes (by ‘speaking’ with the STB via the home Wi-Fi router, rather than directly). This syncing is also possible with some Smart TV models from LG and Samsung if viewers are using the Freeview service (UK free-to-air television) on them.

At the same time, the app senses the signal from all local Wi-Fi routers (in the home and from neighbours) to create a reception fingerprint that is unique for each room in a building. This Wi-Fi triangulation means the phone, carrying the app, can be located to a specific room. ViewersLogic looks at where the channel change takes place, and as 70-80% of the channel changes happen when the phone owner is in a specific room, this is used to determine that the television set is in this room. The company can therefore demonstrate ‘person present in front of television’. You can read more detail about the methodology at the bottom of this story – ultimately the system uses a probability threshold that is acceptable to advertisers.

Using BARB data, ViewersLogic then interrogates channel playout schedules to see which ads were shown at which times on the tuned-in channel, to make the connection between person present and the precise ads they were exposed to.

According to Ronny Golan, CEO and Co-Founder at ViewersLogic: “It takes two minutes for a panellist to sign up. The whole process is totally GDPR compliant because the app does nothing except collect your data and pay you money. There is a very clear value exchange there. The app tracks television behaviour, phone behaviour and location to show offline behaviour. We track activity right through to purchasing.

“Think of our app as a remote-control app that is not controlling anything. To see if our panellist is in front of the television, we look at Wi-Fi signal strength, using AI. You need to be carrying the phone, and most people do when watching TV, but if the phone is not present we will not consider it a view, as we cannot show someone is in front of the TV.”

ViewersLogic does not rely on ACR (automatic content recognition) for its UK operations, but Golan says this can be part of the solution and the company is talking to ACR vendors in other markets. “There are several technologies we can use – the architecture means we can change how we collect the data.”

ViewersLogic has found a way to access some of the most valuable purchase data available anywhere to show the downstream impact of the ad exposures. It gets free access to supermarket loyalty card data, like from Tesco Clubcard, by working GDPR laws to its advantage.

“GDPR is our greatest friend,” Golan declares. “When the law came out, everyone was afraid, but we were counting down the days until it arrived. Loyalty cards know everything you buy, but it is not their data, it is your data, and they must give it back to you in machine readable format on request. You [the consumer] can make a request through their website, so with the full consent of our users [panellists], we make a GDPR request on their behalf and ask for the data to be sent to ViewersLogic. The law says the data can go to a third-party – that is data portability’.”

Golan says loyalty card data is available for most of the ViewersLogic panel, and a subset of the panel can be assessed using both loyalty card data and NielsenIQ (Retail Measurement) data, which tracks sales in physical stores as well as ecommerce. He says his company is processing 2 billion data points every day. The solution keeps data live for a year so brands can interrogate the impact of older campaigns. Viewers can be split into exposed and non-exposed groups to show campaign effect.

The company claims it offers a better understanding of exposure-to-impact than alternative solutions and has various case studies to show what has been achieved. Proven brand uplift is one example, and another shows the value of a sponsorship when sponsorship was combined with regular TV ads (vs regular TV ads without sponsorship). ViewersLogic says new-to-TV brands can learn from what other brands have achieved in their sector, too.

The company, which has a UK HQ and R&D in Tel Aviv, claims it is disrupting ‘traditional’ attribution models that fuse data from different silos to find what it describes as “weak correlations between ad viewing and sales”. The innovation includes the option to extend the attribution window beyond the five-minutes considered typical on ‘traditional’ attribution models to one-week and beyond.

There are no patents on the ViewersLogic solution, with Golan declaring that “in software, patents don’t protect you, only good software engineering”. He points to barriers to replicating the ViewersLogic approach. “With data and AI, you have to develop technology to collect the data and then collect the data for a year before you can create a model to train your AI on, and you also have to build the panel,” he points out.

The ViewersLogic methodology – some more detail

To establish ‘person present in front of TV’, the mobile phone (containing the ViewersLogic panel app) constantly checks for Wi-Fi networks and every couple of minutes relays the results of this ‘survey’. ViewersLogic compares the fingerprint of the survey with the fingerprint of the rooms in the house (to determine which room the panellist is in).

What if the phone is on charge in the kitchen, where there is a television set playing, but the panellist is in their front room reading a book?

Golan admits this would cause a wrong reading, “but when you look at the entire data, these cases are so rare they do not affect the results of our analysis. We can know if the phone is being charged and the last time you touched it, and, for example, decide not to use the TV data if the phone is being charged and was not used for more than an hour. From our research, these cases are very rare.”

To cope with someone moving from watching television in one room to watching a different television in another room, each viewing instance would be treated as a separate process, with the viewing combined at the data analysis phase. Golan explains: “We connect to all the supported TVs and STBs in your house and constantly collect their data. We constantly get Wi-Fi survey results from the phone. We do nothing in real-time, but when we analyse the data, we only consider TV viewing if you are in the room with the TV.”

What about if someone is hanging around in the doorway, watching the television on their way to leaving the room – how does ViewersLogic understand whether you are ‘inside the border’ where you can still be watching TV, or outside it? Golan says: “The AI algorithm that looks at the in-house location gives us a probability that you are in the room – so if you are in the middle of the room, the probability that you are in the room will be very high, and when you are in the doorway it will be smaller.

“Clearly, we do not have an exact map of the house, so we decided that we consider you to be watching TV only if the probability is very high. We have performed a lot of testing to define this.”

But this promise has not been fully delivered quite yet. Attempting to match ad exposure to consumer action reveals glaring holes in the data, while any attempt at cohesion across a CTV campaign — let alone the wider media mix — collides with the messy reality of a fragmented market. This should be CTV’s moment in the limelight, but shortcomings in measurement capabilities are holding it back from realising its full potential.

For CTV advertising to make the most of the consumer attention and ad spend coming its way, media owners need to recognise its current measurement shortcomings and fill the gaps with reliable, detailed, high-quality data on actual campaign effectiveness and the role CTV had in it.

IP and device graph matching: a match made in hell?

IP and device graph matching are currently used as the basis for many CTV measurement solutions. The assumption is, if the IP address of a household where an ad for an item was aired matches the IP address of a household where the item was purchased, it can be reasonable to infer that there was a relationship between the two.

However, IP and device graph matching suffer from several problems that make it almost impossible to really understand the effectiveness of a VOD campaign. On the TV side, the IP addresses identify the household and not the individual, so it is impossible to know who in the household was exposed to the TV ad. Furthermore, when the user leaves their house, they will be assigned a generic IP address from their mobile network which can be shared by multiple users, making it impossible to identify the user.

Brands are therefore ultimately not able to understand from IP matching whether the person who bought the product is the same person who saw the ad.

IP matching and device graph matching are also unable to detect attribution in cases where a brand sells through third-party websites such as Amazon. Finally, IP addresses are considered Personal Identifying Information and using them may create a GDPR liability for brands.

CTV has a fragmentation problem

But even if all the matching problems can be solved, CTV also has a fragmentation problem, as it’s another silo in the media mix. For example, a single consumer might have been exposed to an advertising campaign while watching TV, scrolling through social media, listening to a podcast, travelling on public transport, or while playing a game—isolating the effectiveness of the CTV exposure is close to impossible.

But the problem is even larger. A CTV advertising campaign will run across multiple VOD services, each with a different, siloed measurement system. For example, a user could see the same ad on Netflix, Samsung TV, ITVX, and YouTube. If IP matching is used, each of the services would claim 100% of the attribution. The brand will be unable to understand what worked and what didn’t.

Knowing the effectiveness of each component of the media mix is essential to campaign planning and in-flight optimisation but isolating the impact of CTV advertising in the current advertising environment is almost impossible.

It’s time to go back to the source

Much of the hype around CTV is focused on its likeness to the data-driven insights and scale of digital advertising, but to truly unlock its potential we need to use single-source data and measure it within the context of the entire campaign.

Thanks to several recent technological advancements, it’s now possible to build single-source panels that can track actual behaviours and actions in far more detail, right down to the ads they were exposed to, what sites they visit, apps they use, the online and offline purchases they make, and what physical stores they visit. With this approach, panellist consent is provided from the start and their participation rewarded so that data privacy is assured.

This model overcomes the inaccuracies of IP matching. By tracking individuals every step of the way, single-source data reveals the impact of campaigns on offline store visits, CPG sales, or online purchases through third-party websites, enabling accurate and deterministic measurement of CTV advertising for the first time.

The holistic view of consumer behaviour created by single-source data also solves the fragmentation issue, revealing how CTV advertising fits into the larger picture of the customer journey to purchase. By understanding the performance of the channel in context, marketers can use CTV advertising more effectively in their campaigns.

For too long, the CTV advertising ecosystem has been attempting to complete a puzzle with half the pieces missing. By breaking down silos and using single-source data, the industry can fill the gaps in their measurement capabilities beyond just reach and frequency to give advertisers a complete picture of campaign performance. Only then will CTV truly achieve its potential as the best of both the digital and linear worlds.

NBCUniversal calls the expanded Olympics measurement a test-and-learn initiative. “Our partnership with iSpot is completely changing the TV and advertising business model,” the media owner claims. Select NBCU advertisers are being given access to impact measurement including granular attention and interruption rates, business-outcome reporting and creative performance that includes pre-testing and brand-lift analysis.

The statistics being released daily by NBCUniversal during the Games are designed to show what can be measured thanks to the iSpot.tv initiative, but they also reveal the stunning success of this event for the broadcaster-streamer and its advertisers. And this is not just smart adtech – the data demonstrates how lower ad loads are helping to drive engagement and how that combination is economically viable.

During primetime on Tuesday, NBC had a 24% lower ad load and delivered 134% more impressions per unit than the other three U.S. broadcast networks combined, according to the figures released by NBCU. Throughout the Games up to that point, the Olympics coverage on NBC had a 27% lighter ad load and delivered 169% more ad impressions per unit than the other three broadcast networks combined.

On February 7, NBCUniversal demonstrated how it can link ad exposure to business outcomes. Data revealed that viewers were 34% more likely to search for brands that advertised in NBC Primetime Olympics coverage vs what it calls competitive broadcast/cable norms. On February 5 there was a 39% greater likelihood that viewers would search for brands that advertised in the Primetime Opening Ceremony versus competitive broadcast/cable norms. The February 5 data from NBCU showed that brand recall was up 53% and message memorability was up 70% versus “competitive broadcast/cable norms”.

Tuesday (February 8) delivered 982 million household TV ad impressions within Olympic Games coverage (nearly four-times the nearest show that day). The broadcaster-streamer had delivered 6.7 billion household ad impressions since the Games opened on February 2. Completion rates on Tuesday were 98%. NBCU is claiming 10% higher engagement/attention rates for advertising during the Games versus the industry average for the Games so far. During Week One of the Olympics, advertisers have seen a 9% uplift in purchase consideration compared to the week before the Games, while the same figure for non-Olympic advertisers if flat.

Across all TV on Tuesday, the Olympics were the No.1 programme by ad impression delivery with a 12% share of voice. On Feb 5 they delivered 23% SOV. On Tuesday, the Olympics accounted for around one-ninth of the TV ad impressions that day. The Olympics had a 39% share of voice against the other broadcast networks, and 449% more impressions than the No.2 show, Celebrity Big Brother on CBS.

The test-and-learn initiative builds on NBCU and iSpot.tv’s business outcome arrangement established in 2018, designed to provide full-funnel attribution that correlates TV and video ad exposures to the actions taken as a result. “From digital activity on websites and apps to in-store visits and call activity, brands will understand the lift in activity and conversions from their investment in NBCU vs. their total TV and video spend during the same time period,” the media owner explains.

iSpot.tv has been broadly adopted by the video industry and marketing world. Thirty per cent of brand advertisers and half of the top 100 television and video advertisers utilise iSpot.tv. Krishan Bhatia, President & Chief Business Officer at NBCUniversal Advertising and Partnerships, says: “iSpot.tv has set a high benchmark for measuring ads at scale and sharing information broadly with the industry. That transparency and dedication to accuracy has established a foundation of trust on all sides of the industry and made them an ideal partner to conduct such a massive test.”

Using iSpot.tv, brands who advertise with NBCU will gain access to real-time airing data for linear, streaming and time-shifted viewing, and receive comprehensive reports on a next-day basis that include ad impressions, reach and frequency, linear and streaming overlap and incrementality, among other metrics. Measurement will be provided at the household level as well as at the person level for age/gender demos and/or for customised audience segments.