Liquid Death plans to leave LiveRamp. PRWeek reports the trigger was Publicis's $2.2 billion takeover, and rivals are now asking whether the platform can stay neutral.

That one story sums up the week.

A data layer that used to sit between everyone now answers to one of the parties. Neutral was the product. Once it's in doubt, so is the data.

The referee keeps joining a team

Reddit launched first-party tools to measure incremental impact, according to MediaPost, justified by Reddit's own internal analyses. DoorDash and Instacart are joining forces, which means one bigger seller reporting sales results on its own ads. OpenAI signed deals with Magnite and PubMatic so ChatGPT ads can be bought through the same pipes as everything else.

Different companies, same structure. The party selling the media is the one grading the test.

To be fair, Reddit going after incrementality is the right goal. I've argued against last-click for years, and a platform measuring lift beats one counting clicks.

But that doesn't make their numbers right or wrong. It makes them unverifiable. Nobody outside Reddit can check the work, including the person who has to defend your budget to finance.

Both numbers can be true

Digiday reports that Google search, the line item buyers never questioned, is finally getting questioned. It escaped scrutiny for so long because the dashboard always looked good.

I just got off a call with a client who couldn't square two reports. Our dashboard showed zero incremental impact for a campaign. The publisher's dashboard showed it working.

Both were accurate. They answer different questions.

The publisher shows last-click credit: who touched an ad before converting. Incrementality shows what wouldn't have happened without the spend. In that campaign, the conversions would have happened anyway.

Brand search is the cleanest example. Someone already looking for you types your name, clicks the ad, and the platform takes the credit. Retail media runs on the same logic. Platform-attributed sales count every buyer who saw or touched an ad, and plenty of them were going to buy regardless.

The dashboard isn't lying. It's answering a different question than the one your CFO is asking.

Spend isn't impact

Adweek reports political campaigns spent a near-record $331 per voter, and the GOP's spending advantage didn't translate into votes.

That's the problem in its purest form. A lot of that money landed on people who'd already decided. Last-click would have called every dollar a win.

Brands carry the same blind spot. Publisher-reported credit can't separate the dollars that changed behavior from the dollars that just got counted. And the more of your plan that sits inside walled gardens and merged platforms, the more of your reporting comes from someone with a stake in the answer.

ChatGPT adds another source of impressions to that mix. There will be gaps. Claude doesn't run ads, so nobody will publish how many people ask it for the best barbecue sauce. That's fine.

Measurement was always imperfect. All models are wrong, some are useful. Be wary of anyone selling you a perfect picture.

Get a second read before budgets lock

If you're setting next year's budget now, here's the discipline.

Before you move spend on any platform's lift report, get a second read from someone with no stake in the result. Tie it to your own analytics data, not theirs.

Run holdouts where you can. Ask the one question a platform dashboard can't answer: what changes when the spend goes away?

Then put the two numbers side by side. If they match, you've gained confidence. If they don't, you've learned how much you were paying for sales you'd already won, before you doubled down.

That's why we built a direct API integration with publishers. The advertiser authenticates Google Analytics, the data flows in, and an independent vendor does the measuring. The publisher still gets evaluated. It just doesn't hold the pen.

Pick data partners by the same test. If one has a stake in the answer, discount the answer.

The second number only matters if people use it

An independent read does nothing if your team keeps opening the old dashboard.

This week a client fully switched to Provalytics as their primary source of truth, after several months of running it alongside GA4. Nobody outside our team saw it, but it's the win that mattered most to us.

What made it work was the rollout. They gave everyone a safe space to say what they liked and didn't like about the platform. That opened real conversations, and people came to understand what incrementality is for: seeing the difference and acting on it.

If I could change one thing about most rollouts, I'd ask for the complaints in week one. A team that hasn't voiced objections hasn't adopted anything yet.

Search might still earn its budget. So might Reddit, ChatGPT, and DoorDash and Instacart.

They just have to prove it to someone who isn't selling it.

Sources

More from Jeff

Discuss it on LinkedIn →