There is always a new platform on the shortlist. ChatGPT, Reddit, CTV, whatever your team brings into the room next quarter.
The question underneath it never changes. Is this worth a real test, or are we about to spend two months and $30,000 confirming that our customers don't live there?
We've been early on a few platforms. Some of those bets paid. The ones that didn't tend to fail the same four ways.
Does the audience actually exist there? Presence isn't audience. Your customers may have Reddit accounts and never once be in buying mode on it. The bar is whether someone can show you five real examples of your customer expressing intent on that platform. A demographic overlap chart doesn't count.
Is the window still open? Inventory is underpriced when a platform still needs proof it works. Once that's settled you're paying market rate to learn an unfamiliar system, which is the worst trade available. If your three closest competitors are already spending there, you missed it.
Can you make creative built for it? This is where most tests quietly die. You resize Meta assets, performance comes back weak, and the platform gets written off. You tested your creative. If the plan is "we'll adapt what we have," don't run it.
Is the budget a test or a gesture? Five thousand dollars spread across six weeks produces noise. Someone reads the noise and declares the channel dead. You need enough spend to clear the learning phase and generate real conversion volume inside the window.
These are gates, not weights. Three out of four is still a no.
Failing any one gate doesn't produce a bad result. It produces an uninterpretable one.
And the $30,000 is the smaller loss. The expensive part is the conclusion you now carry into every future planning meeting. "We tried Reddit, it doesn't work for us" outlives the person who ran the test, and it will keep you out of the channel three years later when the window is actually right.
A test you can't read is worse than no test, because no test leaves the question open.
If you want an outside read on whether something on your shortlist clears all four, that's what our audits are for.
Have you launched your creator affiliate program this holiday season? Don't miss out on building demand and landing the best partnerships. Levanta's 90-Day Holiday Sprint breaks it all down. Download the Guide.
What to ask this week…
Two questions when someone brings a platform into the room.
"Who brought this to us?"
Go: someone on your team noticed customers showing up there. No-go: a platform rep with a deck. The first-mover claim is the product they're selling.
"What would make us kill this, and by when?"
Go: a number and a date, agreed before a dollar moves. No-go: "we'll see how it performs." Tests without a kill criterion run until someone loses interest, and the conclusion belongs to whoever talks loudest in that meeting.
Are ChatGPT ads the right fit for me?
The framework applied to the platform most of you are being pitched right now.
ChatGPT ads went live across 31 European markets on Monday, six months after the US pilot. OpenAI ad revenue is reportedly up more than 25% since the start of August.
Gate | Read |
|---|---|
Audience exists | Partial. 25.94% of 50,006 commercial prompts returned an ad in July, so reach is real. But 14.35% of impressions matched their prompt no better than a random one would have. |
Window open | Yes. Auction is thin. Europe launched on opt-in consent rather than legitimate interest, so the pool is smaller and higher intent. |
Creative ready | Depends on you. It's a conversational placement, not a feed. Resized Meta assets will underperform and you'll blame the platform. |
Budget is real | This is the gate most brands fail. Match quality that noisy means you need more spend to reach signal, not less. |
Three clear, one conditional. That reads as a test, not a channel.
Assign one person a small budget to learn the mechanics before the auction fills, and keep it out of your revenue forecast until the match quality improves.
Elsewhere…
Google's mass arbitration exposure. A firm is building a case that Google owes advertisers money going back a decade. Worth watching. Nothing to decide.
YouTube changing how it counts views. Long-form and live are moving to the Shorts method, counting a view at playback start. Your view counts will jump and nothing will have improved. Keep it out of any deck where it could read as growth.
Not ignorable: AI disclosure is now law in four jurisdictions, including California's SB 942 and Article 50 of the EU AI Act, both live since August 2. The IAB's updated framework draws the line at AI that changes what a viewer could reasonably believe is real: synthetic voices, avatars, and digital twins trigger disclosure, AI-drafted copy doesn't. If you're shipping AI-generated creative, someone needs an asset-level record of what the AI did and where it runs.
What's the first thing you check when a new platform comes up? Hit reply. We read them.
