If you’re using IP addresses as an identity signal for connected TV advertising, you’re doing it wrong. Or more accurately, you’re probably getting wrong identity matches.

IP addresses are effectively the CTV version of the web’s cookie. They are often used to match to other, more deterministic forms of identity, like email addresses and even mailing addresses. That way a brand can track if someone saw a brand’s CTV ad and then logged into its site to purchase a product. 

Problem is, most of the time IP address matches are wrong, according to data intelligence company Truthset. The company sources data from 22 data providers – including Experian and Publicis Groupe’s Epsilon – on a quarterly basis and compares that against its own research panel of 1.4 million people.

“In order to assign an audience attribute to a connected television device, you have to go through three or four match processes in the data supply chain, each of which introduces probably a minimum of 50% error,” said Truthset CEO Scott McKinley.

The error rate goes way beyond 50% for some match types, though. 

• 84%: Error rate for IP-to-email matches.
• 87%: Error rate for IP-to-mailing address matches.

Truthset identified other identity inaccuracies as part of its “State of Data Accuracy” report. Matching email addresses to mailing addresses gets it wrong 49% of the time. And even audience demographic segments – from age and gender to behavioral segments like pet and car owners – can be prone to error. 

Truthset analyzed identity data against 85 million U.S. adults over the past five years, McKinley said. Across 25 demographic categories, the company found an average error rate of 55%.

Not great. 

When taking into account these various identity error rates, Truthset estimated that 40% of ad dollars spent buying CTV ads in the open programmatic market in the U.S. is going to waste, which would equate to $7.4 billion based on eMarketer’s CTV ad spend projections.

“Most of the error, particularly in CTV, is coming from these identity hops. They’re just not that accurate. Nobody has good data on how likely a given IP address is to be associated with a given household. So that turned out to be an even bigger problem than the error in the demographic attributes,” McKinley said.

And the IP address problem is only getting worse. Internet service providers in the U.S. are beginning to adopt National Architecture Translation, which has the effect of bundling and obfuscating household-level IP addresses. Last year one major U.S. ISP had bundled at least 50% of its household footprint into collectives of 16 households per public-facing IP address, according to a person with knowledge of the matter.

Meanwhile, Truthset has seen how unintelligible IP addresses can be in other ways.

“We picked out one household. It should have had one to two-and-a-half IP [addresses] over a nine-week period. Instead we had 11 IPs. That means nine or more of those IPs were incorrect and were in reality associated with other households who got those ads that were in question,” said McKinley. “This is not tenable.”

What we’ve heard

“Most of the advertisers tried to take some big swings. Some of them didn’t land as probably as they were intended.”

— Motto’s Sunny Bonnell on this week’s Digiday Podcast episode recapping the Super Bowl’s best and worst ads

Numbers to know

48: Average latency, in seconds, for Peacock’s stream of Super Bowl LX compared to cable TV’s 38-second delay.

$65: Monthly subscription price for YouTube TV’s new sports-only tier.

1%: How much Fox’s ad revenue grew year over year in the final quarter of 2025.

$11.38 billion: YouTube’s ad revenue in the fourth quarter of 2025.

474 million: Number of daily active users that Snapchat had at the end of 2025.

What we’ve covered

As YouTube chases TV budgets, its still holding onto its roots:

• Google’s sales reps are now being incentivized to prioritize pitching YouTube, according to three ad execs Digiday spoke with in exchange of candor for anonymity.
• As YouTube moves further upstream for TV ad dollars, it’s careful not to weaken the lower, mid-funnel mechanics that make those budgets defensible.

Read more about YouTube’s TV ad strategy here.

Even in the age of AI, the Super Bowl rewarded the human touch:

• Dig found that 92.4% of production-related posts favored traditional produced ads featuring real actors and celebrity-led storytelling, reflecting a clear audience preference for human-made creative.
• AI is turning culture into a production shortcut — faster, cheaper and, at times, familiar. But the Super Bowl is one of the last places where brands are reminded that cultural likeness is easy but shared experience is earned.

Read more about the Super Bowl’s ad trends here.

The Athletic leans on live blog, video to guard sports media from AI:

• While live blogs have been around for years, The Athletic sees live formats as a way to keep audiences on its platform, while insulating its reporting from getting scraped and repurposed by AI tools.
• The site had 16.9 million unique visitors in January 2026, up 59% year over year, according to Similarweb data.

Read more about The Athletic’s live content strategy here.

Brands invest in creators for reach as celebs fill the Big Game spots:

• As the creator economy grows, marketers have put new energy into Super Bowl LX outside of the traditional prime-time ad spot — from hosting creators at Super Bowl watch parties to pop-up events on the ground.
• The U.S. annual creator economy ad spend is set to reach $43.9 billion this year, according to the IAB.

Read more about creators’ role in Super Bowl marketing here.

Bloomberg’s new video hub aims to keep audiences on its turf:

• Bloomberg wants its video content to be easier to find and harder to leave.
• Bloomberg has 707,000 paying subscribers, according to a recent internal memo on the publisher’s 2025 finances, written by Bloomberg Media CEO Karen Saltser and shared with Digiday.

Read more about Bloomberg’s video strategy here.

YouTube users have already noticed changes on the platform:

• The changes show just how much YouTube is willing to flex to keep itself ahead of the game — whether that’s competing with big streamers like Netflix and platforms like TikTok, cracking down on AI slop, or expanding upon the way creators can get paid.
• Days after Mohan’s blog promised to clearly label AI-generated products and manage AI slop, 16 of the top 100 most subscribed slop channels were removed from the platform, according to a January 2026 study published by online video editing platform Kapwing.

Read more about YouTube’s 2026 changes here.

What we’re reading

NFL’s next round of rights talks:

The NFL is looking to get an early start on its next round of rights negotiations with an eye toward non-traditional media companies, according to CNBC.

YouTube’s TV ad spend challenges:

YouTube may be the predominant streaming service on TV screens, but the platform is still only receiving a fraction of the money brands are spending on traditional TV ads, according to The Wall Street Journal.

Amazon’s AI-powered TV and film production:

Prime Video’s and MGM’s parent company will start testing a program next month to use AI to cut production costs while claiming to still involve humans, according to Reuters.

Hollywood’s next round of labor talks:

The entertainment industry’s worker unions are set to sign new agreements with film-and-TV studios this year after the last round led to a historic work stoppage, according to The New York Times.

Meta’s standalone Vibes app:

Meta is testing a standalone app for its AI-generated video feed Vibes a la OpenAI’s Sora, according to TechCrunch.