The missing layer in European TV Advertising
Smart TV ad data without the server costs, the surveillance or the US-only economics.
Smart TV manufacturers in Europe have a revenue problem they did not create and cannot easily solve on their own. The devices they build carry more intelligence than ever, the advertising market around those devices is growing and yet most of the household-level advertising data that American platforms have been selling to media buyers for a decade simply does not exist in European markets at any useful scale. Not because no one wants it and not because the technology is fundamentally out of reach but because the infrastructure built to generate it was designed for one specific (and huge) market (the US) with specific economics and those economics have never translated to Europe.
SyncMint, founded a year ago by Mehmet Eroglu and Pieter Oonk, is trying to close that gap with a technical approach that works at European scale and a privacy by design architecture. I sat down with them to get to the bottom of this.
The problem with the American blueprint
The dominant model for TV advertising data in the US was pioneered around 2010 and built around a very specific set of market conditions. A module on the device captures a sample of what is on screen every few seconds, sends that sample to a central server and matches it against a database of known content covering hundreds of channels, thousands of hours of programming, and every ad creative in active rotation. The match tells you what a household was watching and which ads they were exposed to and that data flows to broadcasters, advertisers and device manufacturers as commercially actionable intelligence.
The problem is that this infrastructure is expensive to run, requiring server capacity and content databases that cost millions to maintain and those costs are only justifiable when you have tens of millions of devices contributing data simultaneously. When your addressable footprint is 400K TVs in the Netherlands or 200K in Belgium, the math falls apart entirely. The vendors who built the US system have shown little interest in adapting their operating model for smaller markets, because there is no version of that model that works at European scale without a fundamental rethink of the architecture.
Why TV manufacturers need this more than ever
To understand why this matters commercially, it helps to understand just how thin the economics of Smart TV manufacturing actually are. The bill of (hardware) materials on a mid-range set is substantial and after accounting for manufacturing, logistics, retail and OS licensing, the margin available to the device maker is narrow enough that the business has always depended on supplementary income streams to stay viable.
For a period, those supplementary streams came from the operating system arrangements that defined the last decade of CTV. When Google was actively building Android TV’s device footprint, it paid bounty fees to manufacturers like Sony and TCL, a fixed payment for every set sold with the OS installed, because it needed their distribution reach. Smaller OS providers without the same financial firepower structured similar arrangements around revenue share, promising manufacturers a cut of the monetisation generated on their hardware.
Now layer in a supply chain disruption that has nothing to do with streaming or software with AI companies purchasing available memory at scale and the downstream effect on consumer electronics manufacturers is genuinely severe.
"A customer told us that their cost was around 1$ for a certain memory type and it's now at $44. That's the immense difference they're facing right now. If you don't have the means to subsidise your brand with other income streams, you’re in for a struggle."
Mehmet Eroglu, co-founder
For a manufacturer without the balance sheet depth of a Samsung or LG, that kind of shock does not get absorbed into the P&L. It threatens the entire unit economics of the business and pushes the question of supplementary revenue from consideration to necessity. Data revenue shared directly with the device manufacturer, generated from advertising intelligence that already exists on the hardware they produce, offers a way to address that pressure without requiring them to build a platform, acquire content rights or compete with the OS providers already running on their devices.
How the detection actually works
SyncMint’s technical approach starts with a deliberate constraint that turns out to be a significant advantage. Rather than trying to identify everything on screen, the module embedded in the device’s chipset is trained to detect one specific type of content: advertising. Not sport, not news, not drama, just the ads.
This is not an arbitrary limitation. Advertising content is structurally distinct from everything else on television precisely because it is designed to be: every piece of ad creative is built to stand out, to be immediately recognisable and to not sound or look like the programming it interrupts. That distinctiveness makes advertising far easier to detect than content. Football matches are notoriously difficult to distinguish from each other because the first five seconds of almost every game feature a whistle and crowd noise. Identifying ad creative A versus ad creative B is a far simpler problem and a far lighter one to run on a small portion of a chip’s processing power.
The local (as in ‘on device’) fingerprinting architecture compounds this efficiency. Rather than sending samples to an external server for matching against a vast catalog, the algorithm runs on the device itself, checking incoming audio against a library it already holds locally. On top of the audio detection layer sits a contextual classification engine powered by machine learning, which adds a second dimension to the data. This engine does not identify specific content, it classifies it. It recognises that what is on screen is football rather than tennis, that a Rolex ad is airing during a sports broadcast with Sinner visible and a sponsor logo on court, that the viewing context is premium sport rather than late-night infomercial. It builds meaningful context around each advertising exposure without attempting to build any kind of profile of the person watching. For most planning purposes, that classification is exactly what advertisers need: confirmation that their message reached a household in a specific viewing context and a record of how many times that week it already had.
Privacy as architecture, not compliance
Systems operating in the US can combine a household’s IP address, device identifiers, screen size and detailed viewing history into a behavioral profile that tells you the address, the hardware specification and what that household was watching at eleven on a Friday night. That is one-to-one profiling of personal behaviour, it has already attracted legal scrutiny in the US including an ongoing case in Texas and it reflects a data philosophy that sits awkwardly with how Europeans think about the relationship between their devices and their private lives.
SyncMint’s architecture makes those profiles structurally impossible to build rather than merely difficult to access. The module detects advertising exposure and nothing else. It does not log what content a household watches, how long a viewing session lasts or any behavioural signal outside of which ads have appeared on screen and in what context. There is no viewing diary being assembled anywhere in the system, no content fingerprint being transmitted to a server and no persistent household identifier linking sessions together into a longitudinal profile. The data that flows out of the device is limited, by design, to what is commercially useful for advertising measurement and excludes everything that would make it useful for surveillance.
“The data harvesting that legacy competitors do always felt uncomfortable, especially coming from a European perspective. So we took the approach of: what if we look at advertising exposure only? No one particularly minds that they drive past a billboard and see it. The advertising is there to support.”
Pieter Oonk, co-founder SyncMint
Oonk’s billboard comparison captures the underlying logic cleanly. Nobody objects to driving past a billboard and being counted as part of the audience measurement for that location. They did not choose to see the ad but they understand the implicit exchange between free content and commercial exposure. The data SyncMint collects is the television equivalent: a record of advertising exposure, measured in aggregate, without any of the behavioural inference that has made US Smart TV data practices a target for regulators.
What advertisers do with the data
The most immediate commercial use case is frequency management and the planning gap it addresses is more significant in Europe than most media buyers will acknowledge.
Every major advertiser has a view on optimal frequency for a given campaign. According to Oonk, for a major retailer, historical data pointed to an optimal frequency of seven impressions per week, roughly one per day, and hitting that target consistently across both linear and streaming was something their current media buy could not verify.Staying within that range matters both for efficiency and for how the brand is perceived by the audience. What advertisers have consistently lacked is a reliable way to measure whether they are hitting that target, because linear and CTV are planned and bought through separate systems that do not share household-level data with each other. The linear buy gets optimised against panel reach data. The CTV buy gets optimised against digital impression metrics. Nobody reconciles the two at the household level, because until now there was no data that made that reconciliation possible.
The practical result plays out in every major European market: a household ends up seeing the same supermarket ad ten times in three days because the linear team and the streaming team each optimised their buy in isolation, while a competitor’s message reaches that same household zero times because no one had the visibility to identify and fill the gap.
SyncMint provides a single view of advertising exposure across both environments, which closes the loop and makes it possible to plan against reality rather than against assumptions.
The CPM uplift this kind of data enables is well evidenced in the US market. Oonk points to Vizio as the clearest example: household-level targeting and incremental reach measurement delivered three to four times higher CPMs compared to standard run-of-network rates, a difference that reflects the premium buyers will pay when they can verify they are reaching genuinely unduplicated audiences. On the buy side, a planner managing a major consumer brand campaign can use the data to set different bid prices for households already at frequency saturation versus those that are hard to reach because they are predominantly streaming-only viewers and that kind of bid optimisation creates real efficiency at a scale that European CTV has been promising for years without having the underlying data to support it.
Linear inventory, notably, tends to be the underserved channel in these analyses. The industry narrative around CTV targeting has pulled planning attention and budget toward streaming, but most European households remain cord-shavers rather than cord-cutters, spending a meaningful share of their viewing time on linear while also using streaming services. Those two environments live in separate planning silos and the data that would connect them across a single household view has simply been absent. That absence is what SyncMint is building against.
We’ll discuss that in more details during Streaming Made Easy Live as part of Stream TV Europe.
That’s it for today. Enjoy your weekend and see you on Thursday for another edition of Streaming Made Easy.