Last issue I promised that this week we'd stop doing my arithmetic and start doing theirs. So I went looking for a company that does the arithmetic on its own website, in public, where anyone can read it, and then prays you won't. I found one. It's called tvScientific. I'm not naming them because they're the worst. I'm naming them because they're the clearest to examine as all their records are now in public. They say the quiet part in a press release, then say the even quieter part in the fine print, and if you staple those two sentences together you get a confession with a logo on it.

Here's the pitch, and I'm being generous, because the pitch is genuinely gorgeous. tvScientific is a "performance advertising platform for connected TV." An Idealab company, born out of Bill Gross's startup factory, run by Jason Fairchild, a programmatic-search lifer.

The promise: buy the biggest screen in the house like it's Google. Declare the outcome you want, name your price for it, and their AI, four years in the making, will hit that number or beat it. They call it Cost-Per-Outcome. As of last summer they'll guarantee it. "You only pay for outcomes." On average, they say, campaigns beat the target by 30%. They reach 95% of ad-supported streaming. They call the whole apparatus "radical transparency" and "unprecedented ROI."

Beautiful. Reach for a pen. Now watch the hand leave the table.

Because here's the validation the whole industry pointed at: in December 2025, Pinterest agreed to buy them. Terms undisclosed, expected to close in the first half of 2026. And everyone nodded, the way we nod now, big check, must be real.

But sit with who bought them for one second, because it's the whole joke. Pinterest's entire reason to exist is person-level intent. Six hundred million people, fifteen billion boards, telling Pinterest exactly what they are planning to buy next. That is the purest person-level intent signal in advertising. And Pinterest just married that signal to tvScientific's television engine, the one that connects your ad to your sale on what they call a "1:1 deterministic basis."

So let's talk about that word. Deterministic. It's the word doing all the work, the word that earns the premium. It means same verified human, at the ad and at the sale, no guessing. It's the difference between "we know" and "we inferred." You are paying, at a markup, for we know.

Now go to their own measurement page. Not a leak. Their page, for their buyers. Scroll down a little. The matching, it explains, is done "based on IP address."

There it is. The "1:1 deterministic" identity you're paying up for is an IP address.

And in November, CIMM, Go Addressable, and Truthset benchmarked nearly a billion IP records from six providers against the real truth sets held by the phone and cable companies, and measured IP-to-household accuracy at 13%.

Not ninety. Thirteen. Their own summary sentence: roughly three cents of every dollar spent this way reaches the intended household.

So run the picture all the way through, because it's beautiful in the worst way. The most precise intent data anyone owns, Pinterest's, about to be poured through the leakiest pipe in the business, tvScientific's, which finds the right household 13% of the time. That's not my metaphor anymore. That is a merger. The industry called it a masterstroke. Pinterest called it immaterial to its earnings. Both of those are on the record. Read both sentences.

And now the thesis of this entire issue, the one worth the price of the rest:

It is not that the fancy layer is overpriced. It's that, a stunning amount of the time, the fancy layer makes your advertising worse. It lowers your return. It lowers the lifetime value of the customers you buy. And a dumb, cheap, broad campaign, the exact thing "TV Advertising 2.0" was invented to shame you out of, frequently beats it on the only number that pays your mortgage: profit per actual customer. The premium isn't buying precision. It's buying a more expensive way to reach people who were going to buy anyway, plus a fresh haul of coupon-clippers who will never buy again, plus a scoreboard operated by the same company selling you the game.

Big claim. Mean claim. In Part II, I prove it with three experiments run by eBay, Facebook, and a stack of economists who wanted advertising to work and couldn't make the numbers cooperate. I show you exactly how the outcome-optimized, affiliate-network machine goes hunting for your worst customers and files it as a win. And I hand you the single tell that separates a real vendor from a guy palming a card.

We put it to Truthset's Scott McKinley straight: when Pinterest and tvScientific say "deterministic," is that a real word or a costume? He went with costume.

Real determinism, he says, is narrow.

It needs an IP in the moment tied to actual PII (a streaming-app email, a Comcast billing address, a Roku or Xbox device ID) sitting on hardware that stays put, like a TV or a console.

Everything past that is, and we quote, "pretty much probabilistic and terrible."

The numbers he brought to the fight come from Truthset's own CIMM study. IP-to-hashed-email: 13% correct. IP-to-household: 16%. That is the load-bearing beam under a word marketers are charging a premium for. The honorable exception is Roku-grade matching, where a Pinterest ID snaps 1:1 onto a real ID. If your activation partner actually has that, wonderful. Most are selling you the 16%.

His larger point is structural, and it is the part nobody in the deck wants underlined. Pinterest's whole premium is person-level intent, what people are planning, not what they already did. Push that through an IP and it resolves to a household at best. The bride-to-be pinning centerpieces and her father-in-law watching Tubi on the same router become one indistinguishable blob the instant the signal hits the TV. You did not degrade the precision. You changed what the word "audience" means.

On the glossy 27% and 65% lift stats: when one company sells the audience, runs the media, and grades the homework, the first question is who picked the losing team it beat. And incrementality testing, the industry's favorite exorcism, only half-works. Randomize before the graph touches anything and you get a real answer. Build your control group out of the same broken IP graph and you just cloned the error into both columns.

The kicker, straight from Scott: a CMO buying this tomorrow is measuring one genuinely real thing, that Pinterest has honest intent data, delivered through a pipe nobody validated, scored by the person who owns the pipe. "Deterministic" is how "we made some probabilistic joins" gets rewritten as "we know." The money moves in the gap between those two sentences.

Everything above was the setup. The free part. The part tvScientific's PR person can forward around with a shrug. What's below the line is the subtraction, and it's the part they'd very much prefer you didn't do.

Here's what you're not reading right now:

The three experiments that end the argument. eBay turned its ads completely off in 68 markets and watched what happened to sales. Facebook graded its own homework across 15 experiments and 1.6 billion impressions and caught itself off by 300%. And a Quarterly Journal of Economics study ran the numbers until it proved something genuinely disturbing: with a normal-sized campaign, a healthy 25% return would fail to even show up as real 74% of the time. These aren't my opinions. They're peer-reviewed, and they detonate the entire "deterministic performance" pitch. They're below the line.

The one product feature the literature specifically tells you to run from, that tvScientific put on its marquee. I name it. I show you the paper that warns against it. You will not be able to un-see it in a pitch deck again. Below the line.

The machine that hunts for your worst customers, mapped from their own partner list. I walk you through exactly how an outcome-optimized system, pointed down coupon and cashback and cheap-install rails, goes looking for the lowest-lifetime-value humans in the market and files it as a win, while your dashboard glows green. This is the part that explains why your CPO looks great and your business doesn't. Below the line.

The five questions that clear a room. The exact wording, laminated-card ready, that makes an honest vendor relax and a con artist suddenly need to "circle back offline." Plus the one question that ends the meeting on its own. You walk into your next pitch and run the room instead of getting run. Below the line.

Your Monday-morning playbook. Not theory. Four concrete moves, in order, the holdout to demand in writing, the three-cell test that never makes the deck, the LTV cohort audit that catches the damage the dashboard hides, and how to reallocate on the only number that pays you. Below the line.

Next week's number, teased. The fee scam stacked under the measurement scam. Where half your budget goes before the ad even serves, and the 15% that a Big Four firm with a two-year mandate literally could not find. Paid subscribers get to that first.

And the disclosure I'm proudest of. The one where I tell you what happened to this newsletter's ad revenue the moment we started asking these questions. True story. It's the whole reason this thing stays honest, and it's below the line too.

Get A FREE MONTH OF ADOTAT+ by using this link