FreeWheel today published a blog post examining the structural limitations of IP address-based targeting in streaming television advertising, urging the industry to move toward deterministic identity solutions rather than continuing to rely on signals that a major study found were accurate barely one time in eight.

The post, authored by Barrie Brandt, Director of Product Marketing at FreeWheel, arrives as the connected television advertising market grapples with a fundamental tension: demand for household-level precision on the one side and an identity infrastructure widely acknowledged to be unreliable on the other. The piece draws directly on research commissioned by the Coalition for Innovative Media Measurement (CIMM) and Go Addressable and conducted by data validation firm Truthset, which found that IP-to-postal address linkages were accurate just 13% of the time, and IP-to-email linkages only 16%.

"Imagine sending a birthday gift to someone based on a home address that you're only 13% sure is accurate or an email address that you're only 16% sure is correct," Brandt wrote. "Chances are the gift ends up at the wrong house or email address."

That analogy captures the core problem with considerable economy. IP addresses are assigned dynamically by internet service providers, rotate for security and software update reasons, and can be masked entirely by virtual private networks. When a household resets its router - a reflex response to buffering during a live sports broadcast, for example - it may receive a new IP address immediately. The address previously linked to that household in an advertiser's targeting data could then belong to a neighbor. Geographic precision compounds the problem: IP-based geolocation, according to FreeWheel, typically resolves only to city level, which creates acute inaccuracy in dense urban areas and in locations near administrative borders.

None of this is, strictly speaking, new. The Truthset analysis, released in November 2025 at the Go Addressable Summit, examined nearly one billion IP address records sourced from six major data providers and cross-referenced them against deterministic data from two internet service providers and one multichannel video programming distributor. The study found that providers agreed on the same IP-to-household linkage only 6.4% of the time - a figure that reveals how fragmented and inconsistent the underlying data infrastructure actually is. State-level analysis in that study showed IP-to-postal accuracy ranging from 5% to 17% depending on geography, with states whose population centers sit near borders performing worst.

FreeWheel's contribution today is not to restate those numbers, but to explain the technical mechanism behind them - and to describe what a more reliable architecture looks like.

The mechanics of IP address instability

At the core of the problem is the nature of dynamic IP allocation. Most residential internet customers are assigned addresses that change whenever their router restarts, whenever their ISP performs a software update, or simply according to the ISP's own rotation schedule. Dynamic Host Configuration Protocol, the standard used to distribute IP addresses within networks, was never designed with advertising identity in mind. It was built for efficient network management, and the addresses it assigns are temporary by default.

VPNs introduce further distortion. When a consumer routes traffic through a virtual private network for remote work or privacy, the IP address visible to an ad server may correspond to a server farm in a different city or country entirely. The household the advertiser believes it is reaching bears no relation to the actual viewer. For frequency capping - a core function of any campaign management system - this is particularly damaging. A campaign designed to limit exposure to three impressions per household may deliver many more to the same viewers if those viewers appear, from the data, to be separate households.

According to FreeWheel, the widespread practice of inferring IP-to-household linkages from probabilistic matches against email or postal records is where the system most visibly breaks down. An identity provider might know a consumer's home address from a loyalty card registration, then use that address to infer an IP address, then use that IP to target the consumer on CTV. Each inference introduces error. Stacked together, those errors compound to the point where targeting accuracy falls to the figures Truthset measured.

Deterministic sources as the alternative

The distinction FreeWheel draws is between probabilistic and deterministic data. Probabilistic linkages, which characterize much of current IP-based targeting practice, rest on statistical correlations rather than verified assignments. Deterministic linkages, by contrast, come directly from the entity that issues and manages the address - an internet service provider.

"When the source of IP address data comes from deterministic sources, IP-based targeting is as good as any identifier used for targeting," Brandt wrote. "Internet service providers, like Comcast, are deterministic sources because they generate, assign, and manage IP addresses for users."

FreeWheel occupies a structurally unusual position in this debate. As a subsidiary of Comcast, it has direct access to Comcast's IP-resolution data - not as a probabilistic inference but as an authoritative record. The company's Identity Network, described in the post, incorporates that deterministic IP data alongside verified household addresses from external data providers and first-party IDs from streaming subscriptions. The result, according to FreeWheel, is a verified IP-to-email linkage path that bypasses the probabilistic intermediary steps most of the industry relies on.

The scale of the difference shows up in FreeWheel's operational data. According to internal figures cited in the post, IP addresses account for only 17% of audience targeting within FreeWheel's platform. Most targeting activity within what FreeWheel calls Buyer Cloud - the cloud-based, modular infrastructure formerly known as Beeswax's Bidder-as-a-Service - relies on stronger identity signals. FreeWheel renamed and restructured its product suite in May 2025, consolidating its programmatic offerings under the Publisher Suite and Advertiser Suite architectures, with Buyer Cloud sitting on the buy side.

Why this matters for the broader market

The advertising industry is not moving away from IP addresses entirely. Google, for example, announced in December 2024 that it would modify its platform program policies to permit IP address use for CTV targeting, with changes taking effect on February 16, 2025. Google's approach leans on privacy-enhancing technologies - confidential computing, trusted execution environments, and secure multi-party computation - to allow IP-based targeting without re-identifying individual users. Subsequent analysis on PPC Land noted that Google's move was in part competitive: the company's CTV capabilities had lagged behind rivals who already deployed IP-based targeting.

FreeWheel's position, implicit in Brandt's post, is that deploying IP addresses more widely does not solve the accuracy problem if those addresses are sourced from probabilistic graphs. The relevant question is not whether IP addresses appear in a targeting workflow, but where the IP data originates. A verified IP assignment from an ISP and an inferred IP linkage built from email matches may look identical in a bid request but deliver dramatically different targeting outcomes.

FreeWheel has built its identity architecture around that distinction. The Identity Network's native integration of Comcast deterministic data positions FreeWheel differently from identity providers that assemble probabilistic graphs without ISP-level source data. For publishers working within FreeWheel's ecosystem, that integration surfaces as an addressability input that scales across varied inventory without requiring each publisher to maintain its own identity infrastructure.

The broader movement in the industry runs in the same direction. Smartclip integrated a deterministic TV measurement platform into its supply-side platform in April 2025. Utiq and Visoon launched consent-based identifiers for HbbTV and CTV environments in September 2025. Comscore and FreeWheel had already partnered in August 2024 to develop privacy-resilient, ID-free audience targeting for CTV. The pattern points toward a market that is progressively layering deterministic data at the base of identity graphs and using probabilistic data to fill gaps rather than as the primary targeting signal.

What FreeWheel recommends

The practical recommendations in the post are directed at both sides of the advertising transaction. Advertisers, the company argues, should ask technology partners specific questions about which signals are prioritized in audience targeting and how those signals connect back to an attributed conversion or sale. The quality of that signal path - not simply its existence - determines whether frequency caps hold, whether measurement reflects actual exposure, and whether a campaign reaches the household it intended to reach.

Publishers, meanwhile, are encouraged to partner with audience providers whose data can scale accurately across varied inventory and to share deterministic identity signals wherever possible. The argument is that publishers who contribute higher-quality identity data to the ecosystem will, over time, attract more demand and enable better outcomes for their advertiser clients.

There is no single solution to streaming addressability, according to FreeWheel. Interoperability and multiple durable data inputs are required, and no single company holds complete data across all households. That acknowledgment runs somewhat against self-interest for a company describing its own identity solution in the same post - but it reflects the technical reality of a market where even the largest identity providers maintain meaningful gaps in household coverage.

The November 2025 Truthset study found that the best-performing data providers achieved substantially better accuracy than the worst, with performance differences of roughly 4.5 times between top and bottom providers. That spread suggests that the market is already differentiating - that deterministic sourcing does produce measurably better targeting - even if the average accuracy figures remain low in absolute terms.

For the marketing community, the implications are direct. Campaigns that assume IP-based targeting is delivering household-level accuracy are, in most cases, operating on assumptions the underlying data does not support. The 87% miss rate documented in the Truthset study translates to wasted impressions at scale. As CTV advertising spending approaches $33.35 billion in 2025, the cost of that inaccuracy accumulates rapidly.

Timeline

Summary

Who: FreeWheel, a Comcast subsidiary and major technology platform for television advertising, through Barrie Brandt, Director of Product Marketing. The broader context involves research commissioned by CIMM and Go Addressable, conducted by Truthset, examining data from six major identity providers.

What: FreeWheel published a blog post today detailing why IP addresses are unreliable as primary targeting identifiers for CTV advertising, citing Truthset data showing 13% postal accuracy and 16% email accuracy. The company argues that deterministic sourcing - specifically ISP-level IP assignment data - is the appropriate foundation for identity graphs, and that its own Identity Network is built on Comcast's verified IP-resolution data, first-party streaming subscription IDs, and trusted household address data from external providers. FreeWheel disclosed that IP addresses represent only 17% of audience targeting activity within its platform.

When: The blog post was published on February 28, 2026. The underlying Truthset research was released at the Go Addressable Summit on November 5, 2025, covering a 90-day data collection period from December 2024 through February 2025.

Where: The post appears on FreeWheel's website. The issues it addresses affect the CTV advertising ecosystem broadly across the United States and, increasingly, international markets. FreeWheel operates globally with offices in New York, Chicago, London, Paris, and Beijing.

Why: IP-based targeting is pervasive in CTV advertising despite fundamental structural problems rooted in dynamic IP allocation, ISP rotation, VPN usage, and the probabilistic inference methods most identity providers use to link IP addresses to households. With CTV advertising spending approaching $33.35 billion in 2025 and the industry moving toward deterministic measurement as a standard, the accuracy of underlying identity data has direct financial consequences for advertisers spending at scale. FreeWheel's post positions its Identity Network - built on Comcast's deterministic IP data - as a more reliable alternative at a moment when the limitations of probabilistic IP graphs are receiving growing industry attention.

Share this article
The link has been copied!