The Hardest KPI In Advertising

Full analysis

Your draft

Political advertising has run in a post-cookie world for a decade, and Mark Jablonowski, CEO of DSPolitical, spent this AdExchanger Talks episode with Allison Schiff explaining how. The practical question for the rest of us: are the disciplines political shops built out of necessity, audience-first targeting, aggressive exclusion, and RCT-validated first-party models, worth stealing for brand programmatic? And is the disinformation threat Jablonowski flags a real operator problem or a policy debate we can ignore?

This is a Type 2 situation. Nothing here forces a decision, and nothing here is hard to reverse. The forcing function is the 2026 midterms, with AdImpact projecting $11.6 billion in political spend, past the 2022 midterm record of $8.9 billion and even the 2024 presidential cycle's $11.2 billion. That money moves through the same CTV inventory and identity plumbing brands use. So the spillover is worth understanding even if the direct impact is low.

The council

The Market Analyst. Follow the $2.7 billion of CTV political spend this cycle. Jablonowski says campaigns buy CTV like it's broadcast, entire congressional districts, no audience layering. That's a district-wide buy on inventory sold at a premium for precision. For SSPs and CTV publishers, that's fine in the short run: dumb money at high CPMs is still money. But it tells you the audience-based CTV buying that FreeWheel, Magnite, and the CTV DSPs keep pitching still isn't landing, even with buyers who have the best voter data on earth and a binary win-or-lose incentive. In plain terms: if the most motivated buyers in advertising can't be bothered to use CTV's targeting, the workflow is too hard, not the buyer too lazy.

The Skeptic. The claim that political discipline transfers to brands is where this episode oversells itself. It mostly doesn't, and the reason is the incentive, not the tooling. Campaigns run RCTs and scrub converted voters daily because an election is one day, one outcome, and the data (who voted) arrives clean and fast. Brands have no election day, no voter file, and no clean signal that a customer "converted" and should be suppressed. Jablonowski's own origin claim, first to match the national voter file to online cookies, is unverifiable and self-serving; TargetSmart and Data Trust were in the same lane. Take the man's book with salt. The methods are real. The idea they port cleanly to a retailer's always-on campaign is the oversell.

The Operator. Two things here actually work on a Tuesday morning. First, real-time conversion suppression. Jablonowski's daily feed that stops targeting voters once they've cast a ballot is the grown-up version of the post-conversion exclusion most brand programmatic still botches, retargeting people who bought the thing three days ago. If you run high-velocity campaigns in retail, insurance, or finance, a real-time suppression feed is a concrete build worth scoping. Second, exclusion logic as a first-class lever, not an afterthought. The catch: the plumbing is real work. DSPolitical runs three separate bidders per campaign for match rate. That's expensive over-engineering justified only by binary election stakes. Don't copy that part.

The Customer / End User. For the brand marketer being sold pre-packaged vendor audiences, the useful message is that the campaigns spending billions build their own models on their own data and validate with control tests. That's the standard brands claim to want and rarely deploy. The honest read: they don't deploy it because their CRM is messier than a voter file and their org isn't structured to run experiments. The gap isn't knowledge. It's data quality and organizational patience, and no vendor deck fixes either.

The Pre-Mortem. Say the disinformation thread gets ignored and bites in 2027. Jablonowski's argument is that organic disinformation on platforms with gutted trust-and-safety teams dwarfs the AI deepfake panic in paid media, because organic content faces no disclosure rules, no ad archive, and algorithmic amplification. The warning sign is already flashing: bot-detection startup Spur just raised $200 million from Insight Partners, founded by two ex-Defense engineers in 2017. Capital is flowing to bot and disinformation infrastructure. If brand-safety vendors keep fighting the last war on deepfake creative in paid inventory while the organic bot problem swamps the environment their ads sit next to, they'll be selling the wrong product into 2027.

Where they part ways

The Market Analyst sees a real commercial gap in audience-based CTV; the Skeptic says the gap persists because the incentive to close it is weak even for the most motivated buyers, so don't hold your breath for tooling to fix it.

The Operator wants to steal real-time suppression; the Customer says most brands can't feed it clean enough data to matter. That tension is the whole "does political transfer to brand" question in miniature.

What it hinges on

The transfer question turns on one thing: whether a brand has a fast, clean conversion signal and the org to act on it. Political campaigns have that by nature. Most brands don't. Where a brand does have it, retail with loyalty data, subscriptions with a hard signup event, the suppression and exclusion playbook is worth building. Everywhere else it's aspiration.

On market structure, M&A, and pricing in mainstream programmatic, this episode moves nothing. It's a niche case study. Say that plainly. The one thread with legs beyond politics is the money moving toward bot and disinformation detection, and that's a real signal because the capital is already committed.

No high-conviction prediction this week.

The episode is a niche political case study. The one forward-looking thread with real evidence, capital flowing to bot-detection infrastructure like Spur's $200 million raise, isn't what the episode is actually about, and the transfer-to-brand and CTV-adoption threads are too slow-moving to pin a dated, falsifiable call on without manufacturing false precision.