Refacto

Podcast episode

Rock Bottom = Good News

by Ken Rona · ·AdTech AdTalk

dsp performance-marketing programmatic publisher-economics ssp

TL;DR

Adam Heimlich (buy-side, Chalice) and Gareth Glaser (sell-side, Gamera) survey a gloomy open-web landscape — declining publisher traffic, agency RFIs that unknowingly parrot Chalice's own pitch, and contracting consumer ad categories — before landing on Taboola's deal to manage NBCUniversal's programmatic sales as the most structurally interesting development of the cycle. The thesis throughout: closed-loop, performance-oriented models beat open-auction programmatic for SMB demand, and AI will accelerate brand in-housing at scale.

What was covered

  • Open-web publisher distress: Gareth Glaser described universal traffic declines across open-web publishers, characterizing the mood as "publisher existential dread" — stuck on a sinking 150-year-old ocean liner with no escape. Gaming sites (both content-about-games and in-browser game platforms) and local news were called out as rare bright spots, with gaming benefiting from direct type-in traffic independent of Google.
  • LTV tracking as publisher survival tool: Gareth highlighted that Gamera's first product was lifetime value (LTV) tracking — measuring revenue per reader and return visit rates — and that publishers are only now beginning to adopt it four years after he first pitched it.
  • Google AI Overviews and affiliate review economics: Adam Heimlich and Gareth discussed how AI Overviews are throttling search referral traffic. Product-review sites, they argued, already rely more on affiliate revenue than programmatic display, and the real shift is toward direct licensing of review content to LLMs rather than ad-supported models.
  • Agency RFIs echoing Chalice's own pitch: Adam noted that major holding-company agencies are sending Chalice RFIs that verbatim describe Chalice's modular, platform-independent, incrementality-focused AI approach — after seven years of the company operating with that positioning. His diagnosis: aggregated agency data stacks output average results, which fails large enterprise brands with unique optimization needs.
  • Contracting consumer ad categories: Both hosts flagged that beer, snacks, DTC upstarts, and consumer finance are shrinking as ad categories, with growth migrating to pharma/GLP-1s, dermatology, sports betting, cannabis, and gaming.
  • Taboola / NBCUniversal programmatic deal: Adam disclosed he is personally taking a long position in Taboola (explicitly labeled "not financial advice"), arguing Taboola is "grotesquely undervalued" at roughly a $1 billion market cap with ~$485 million in quarterly core revenue, versus The Trade Desk's ~$710 million quarterly revenue. The Taboola/NBC arrangement is a guaranteed-revenue deal where Taboola's Realize DSP (its own bidder) manages programmatic sales, keeping the revenue split at roughly 65% to publisher / 35% to Taboola with no additional supply-path middlemen.
  • Agentic ad tech / ADCP debate: Adam pushed back on ADCP (Ad Commerce Protocol, a proposed standard for buyer–seller agent-to-agent transactions) and companies like Swivel, arguing that without committed buy-side adoption the standards are "fighting for real estate on the moon." Both hosts were bullish on narrower agentic use cases — AI agents that take a media plan and auto-traffic campaigns — but skeptical that middleware aggregators add value in that stack vs. foundational AI tools like Claude.
  • AI-driven in-housing acceleration: Adam predicted that enterprises spending $500M+ annually on media will accelerate in-housing as agentic AI lowers the execution barrier, putting further pressure on agency fee models.

Notable claims & predictions

  • Adam Heimlich: "I personally am taking positions in Taboola right now… their court revenue last quarter was 485 million dollars. The Trade Desk revenue last quarter was 710. That's close." — Framing Taboola as severely undervalued relative to DSPs at a ~$1B market cap.
  • Adam Heimlich: "If you aggregate data then you'll output averages — and big clients are dissatisfied with Google and Meta because averages don't work for them." — Arguing holding-company agency data pooling is structurally incapable of enterprise-grade optimization.
  • Gareth Glaser: "Traffic is universally down… most of the damage has occurred. The drops have been pretty significant, but they're flatlining." — Suggesting the worst of the open-web traffic collapse is behind publishers.
  • Adam Heimlich on agency in-housing: "It is crazy to me that a company that spends 500 million dollars a year on ads is going to pay a percentage back to someone else to do their ads" — predicting AI makes in-housing by large brands inevitable and imminent.
  • Gareth Glaser on Taboola's publisher guarantee: "Taboola is the master of guarantees… They will give NBC a number. They're basically buying access with their balance sheet." — Positioning the revenue-guarantee model as the structural reason NBCUniversal defected from open programmatic.
  • Adam Heimlich on open-web contraction: "The agency world is looking at something like 50 percent contraction over the next couple years… I see a lot of denial in the buyer space."

Fact check

  • Adam Heimlich's revenue comparison — Taboola ~$485M quarterly core revenue vs. The Trade Desk ~$710M quarterly revenue: Unverified from this transcript. These figures are stated as current without a sourced quarter or filing cited in the episode. Listeners should verify against each company's most recent earnings release before acting on the comparison. Adam also explicitly discloses a personal long position in Taboola immediately before making this valuation case — a clear misaligned incentive. His characterization of Taboola as "grotesquely undervalued" and the stock comparison to The Trade Desk should be discounted accordingly; he is talking his own book.
  • Adam Heimlich on Beeswax being acquired by Comcast: Accurate in substance — Beeswax (a DSP, or software advertisers use to buy digital ads, known for its configurable bidding infrastructure) was acquired by FreeWheel, which is owned by Comcast. The characterization "where things go to die" is editorial, not factual.
  • Gareth Glaser's claim that weather companies "inflate" rain probability because "people don't like the national data because it's too accurate": Unverified and stated flippantly. There is legitimate academic research on probability-of-precipitation presentation biases, but the causal claim that weather publishers deliberately skew forecasts upward for user preference is not supported by sourced evidence in this episode.
  • Gareth Glaser's Taboola revenue split of "65/35 publisher/Taboola": Unverified in the transcript. This is stated as a known fact without citing a source. The actual contractual splits in Taboola's publisher agreements are not publicly standardized, so listeners should treat this as an estimate, not a confirmed figure.

Names mentioned (from the watchlist

  • The Trade Desk — compared to Taboola on quarterly revenue and market positioning; Adam suggests Taboola's NBC deal "should terrify" The Trade Desk.
  • NBCUniversal (Comcast) — announced Taboola will manage all programmatic display sales; discussed as a strategic defection from open-web SSP stack.
  • Beeswax (FreeWheel/Comcast) — discussed as an early modular DSP pioneer, now growing via market timing despite Comcast ownership limiting investment; Adam and Gareth referenced using it as early as 2015.
  • LiveRamp — cited as an early identity workflow integration on Beeswax (cookie-free bidding on LiveRamp ATS IDs).
  • Yahoo — cited as a large, healthy publisher with strong finance and weather traffic; described as "unfathomably large."
  • Jeff Green (The Trade Desk) — his public case for open-web brand advertising cited and debated in context of Taboola's NBC deal.
  • AppLovin — not mentioned directly, but the SMB performance advertising model Adam describes for Taboola maps closely to AppLovin's thesis; not explicitly referenced.
  • Scope3 — mentioned as a platform where publishers build packages for advertisers in the agentic ad-tech context.

Why this matters for ad-tech operators

  • Taboola's NBC deal signals a structural shift in how premium publishers may monetize display. If Taboola's closed-loop model (guaranteed floor revenue, endemic SMB demand, Realize DSP as the primary bidder) outperforms open-auction SSP stacks for publishers, other large publishers may evaluate

Full analysis

Taboola signing to run all of NBCUniversal's programmatic display sales is the one development on Adam Heimlich and Gareth Glaser's podcast worth an operator's attention. Strip away the stock talk and the question underneath is real: when a premium publisher hands its programmatic monetization to a performance player that guarantees revenue off its own balance sheet and bids with its own DSP, is that a one-off, or the shape of how open-web display gets sold from here?

Reversibility: Type 1 for NBCUniversal (unwinding a guaranteed-revenue deal and rebuilding an SSP stack is slow and expensive), Type 2 for everyone watching (you can wait and see before copying it). What's actually being decided: whether large publishers keep running open auctions with a chain of middlemen, or hand the whole job to a single counterparty that promises a number. Forcing function: the deal is live, and other publisher CROs will get asked about it in their next board meeting.

The Market Analyst Adam is talking his own book and says so, so discount the "grotesquely undervalued" line to zero. The valuation contest between Taboola and The Trade Desk is noise. What isn't noise: a premium broadcaster chose a guaranteed floor over auction upside. That's a publisher voting that predictable beats maximal. In plain terms, NBCUniversal took a salary instead of commission. For the open-web SSP model, one anchor client defecting to a closed loop matters more than any quarterly print, because it tells you where large publishers think the risk now sits.

The Skeptic Steelman the bear case on this being a trend. Taboola can guarantee NBCUniversal a number because Taboola owns endemic SMB demand that clears against recommendation widgets. That demand doesn't obviously transfer to premium video or display at scale. Guarantees are only as good as the balance sheet behind them, and "buying access with your balance sheet," as Gareth put it, is a cost that compounds every renewal. If the guaranteed number ever falls below what an open auction would have cleared, NBCUniversal's finance team notices within two quarters. This works until the arbitrage stops working.

The Operator For a publisher monetization lead, the appeal is obvious and dangerous. A guarantee kills your revenue-forecasting headache overnight. But you've also handed your yield, your buyer relationships, and your data to a single party whose DSP is the primary bidder into your own inventory. That's the fox running the henhouse and cutting you a check for the eggs. First thing that breaks at 90 days: your direct-sold team and your programmatic guarantee compete with the same demand, and nobody internally can see the real clearing price anymore. You've traded transparency for certainty.

The Customer / End User Two customers here. The advertiser gets a cleaner path: one bidder, fewer supply-path hops, roughly 35% to Taboola with no additional middlemen skimming. That's genuinely better than the ten-tolls-on-one-highway open auction. The other customer is the SMB advertiser Taboola already owns, and that's the tell. This model works for publishers who can absorb performance-priced SMB demand. It works far worse for a publisher whose value is brand-safe reach that brand buyers pay a premium for. The Athletic and a snack brand's DR team want different things from the same pixel.

The CFO The guarantee is the whole story and also the whole risk. Adam's own framing, that a company spending $500M a year on ads paying a percentage to someone else is "crazy," cuts both ways. It's exactly why brands are in-housing, and it's exactly what NBCUniversal just did in reverse by paying Taboola a cut to run its sales. Somebody is always taking the percentage. The real question for a publisher CFO: is the guaranteed floor above or below your trailing twelve months of open-auction revenue, net of the ad-tech tax you're already paying? If Taboola can guarantee a number and still profit, that spread was your money.

Where the council splits:

The Market Analyst and the Skeptic disagree on whether NBCUniversal is a template or an outlier. One sees the first domino of publishers abandoning the open auction; the other sees a deal that only pencils because Taboola has a specific, non-transferable pool of SMB demand.

The Operator and the Customer disagree on the transparency trade. The advertiser genuinely benefits from fewer middlemen. The publisher gives up the ability to know what its own inventory is worth. Both are true, and they point in opposite directions.

Underneath both: does closed-loop performance demand actually clear premium inventory at premium prices, or does it reprice premium down to DR economics? That's the hinge.

What it hinges on: whether Taboola's guarantee to NBCUniversal is above the open-auction number NBCUniversal was clearing, and whether the SMB demand that funds that guarantee holds up against premium video and display. If both are true, the model spreads. If either fails, this is a distressed publisher taking a lifeline, not a strategy other healthy publishers copy. The council leans toward: this is real for distressed and mid-tier open-web publishers, and mostly irrelevant for publishers who still command genuine brand-buyer premiums. Before copying it, a publisher should model the guaranteed floor against trailing net programmatic revenue and stress-test what happens when Taboola's own DSP is the price-setter into its inventory.

Prediction: No other publisher in NBCUniversal's tier (a major US broadcast or cable network group: Fox, Warner Bros Discovery, Paramount-Skydance, or Disney) will hand its full programmatic display sales to Taboola or a comparable performance-recommendation player through the 2027 upfront season (May 2027).

Confidence: Medium. The model fits distressed open-web publishers, not premium-video sellers with direct sales leverage.

Why: NBCUniversal handed programmatic display to Taboola because a guaranteed floor beat the uncertainty of open auctions, and Taboola could fund that guarantee off endemic SMB demand it already owns. But the other broadcast groups still run large direct-sold video businesses where brand buyers pay a premium for reach and safety, and handing that to a DR-priced bidder risks repricing their best inventory down toward performance economics. The peers most likely to move, mid-tier open-web publishers with collapsing traffic and no direct-sales leverage, are not in NBCUniversal's tier. The opposite outcome, a same-tier network copying the deal within a year, would require a premium broadcaster to conclude its own sales team can't beat a guaranteed floor, and that's a concession none of the four will make heading into an upfront where they're still selling scarcity.

Revisit by 2027-05-31: We're right if no Fox, Warner Bros Discovery, Paramount-Skydance, or Disney property has publicly moved its full programmatic display sales to Taboola or a direct equivalent by the 2027 upfronts. We're wrong if any one of them announces such a deal before then.

Listen to the episode →

Comments