Playwire

Every content moderation vendor walks in with an accuracy number. Almost none of them will tell you what it's made of.

Here's the trick. A tool can hit "99% accurate" by flagging everything above a low risk threshold. It didn't solve your problem. It just moved it from your comment section to your review queue.

And for publishers, this isn't a trust-and-safety line item. Unmoderated UGC gets flagged by brand safety scoring, demand partners lower floors or pull out, CPMs drop, and yield ops spends weeks getting them back. A moderation tool is a revenue protection decision that happens to look like a community management one.

So before you sign, get answers to these seven:
1. Accuracy: false positive and false negative rates, separately, on representative test data
2. API performance: p95 and p99 latency under load, plus fail-open vs. fail-closed behavior
3. Compliance: who tracks regulatory changes, and is there a changelog?
4. Customization: can you train classifiers on your own labeled data?
5. Human review: what triggers escalation, and who's actually reviewing?
6. Transparency: self-serve audit logs, or a support ticket?
7. Integration: a realistic timeline, not the demo timeline

If a vendor can't answer all seven clearly, that is the answer.

The full procurement checklist, with what a strong response looks like for each: https://hubs.li/Q04xFl9c0

"Interstitials are a mobile format."

Say that in a layout review and we'll show you the chart.

August 2026, each device indexed to its own ecosystem average:
Mobile interstitial: 2.68x
Desktop interstitial: 2.76x

Desktop holds the premium as firmly as mobile does. Arguably a hair better. If your desktop traffic is meaningful and you excluded it from the interstitial plan by default, that's revenue sitting on the floor.

A few other things the guide clears up that trip people up:
➡️ 300x250 inside a full-screen overlay isn't a typo. It centers on a dimmed background, which is exactly how Google recommends rendering it.
➡️ Interstitial, prestitial, and pop-up are three different formats with three different policy outcomes. Two of them are banned outright on mobile web.
➡️ GAM web interstitials have been supported in single-page apps since June 2025. No full page reload required.

Judge the unit on revenue per session, not on its own fill rate. A format serving 1 impression per 246 page views will always look small in isolation and still move your total.

Read the full breakdown: https://bit.ly/3TmqAYP

You're running 12 ad units. Three of them are paying the bills. The other nine are decoration.

We pulled a month of RAMP ecosystem data across 40 publishers and 568M sessions. At the median site, 90% of ad revenue came from just three placements. That number held from the largest publisher in the sample to the smallest.

Which means layout work isn't a 12-unit optimization problem. It's a 3-unit problem, and most publishers are spending their testing budget on the 10%.

What the data says your three should be:
- Your anchor: the bottom rail. 23.5% of sampled revenue, median rank #1.
- Your in-content workhorse: the repeating unit that scales with content length.
- Your high-viewability specialist: corner video (87% viewable) or a well-positioned rail.

Get those three right. Then, and only then, worry about unit number four.

Full placement, viewability, device, and density data in the guide: https://hubs.li/Q04xD_7j0

If you missed the last Playwire Pickleball Open, now's your chance.

Playwire and Admiral are running it back on Sept 24 at City Pickle in Times Square. Pickleball, drinks, industry peers, and one very shiny trophy.

All skill levels welcome. Invite only, and spots go quickly.

Apply for your invite: https://hubs.li/Q04t90Zn0

We’re excited to welcome Maida Khan, Senior Machine Learning Engineer to Playwire!

With deep experience in machine learning, generative AI, and building scalable data-driven products, she brings an incredible mix of technical expertise and product thinking to the team.

Join us in giving Maida a warm welcome!

The AI moderation tool flagging your inventory isn't broken.

It's working exactly as designed. For someone else's site.

A combat tutorial on a gaming site trips the violence classifier. A pediatric dosing guide trips the adult-content classifier. Injury reporting on a sports site gets blocked at peak demand.

Every one of those is legitimate, high-quality editorial content, and every one of them gets the same treatment as content that should be blocked.

Generic moderation is trained on the broadest possible definition of unsafe. That's fine for a general-audience news feed. It's a liability if you're in gaming, health, education, sports, legal, or finance, where your core content reliably triggers classifiers built without you in mind.

And the bill shows up in your revenue, twice. Once when flagged pages get excluded from curated deals and PMPs, and again when lower fill drags down the quality signals buyers use to price everything else.

The fix isn't lowering the bar. It's calibration: contextual classifier training, vertical-specific thresholds, category overrides, and a feedback loop that keeps it tuned.

The publishers hurt most by generic moderation are the ones investing most in content quality. → Here's how to close that gap: https://hubs.li/Q04x8csK0

Interstitial ads get treated one of two ways. Either they're a revenue cheat code you bolt on to save a bad quarter, or they're a UX crime you refuse to touch.

Both takes are lazy. Here's what the data actually says.

Across our ecosystem, the web interstitial returned 2.84x the average CPM in August 2026. It held between 2.6x and 3.0x every month sampled over the past year.

It did that while appearing on roughly 1 in 246 page views. 0.14% of impressions, 0.40% of revenue.

That's not a coincidence. The premium exists BECAUSE it's scarce. One advertiser owns the whole viewport, it fires between page states instead of inside content, and it never floods its own auction.

Which means the second you raise frequency to "get more out of it," you're trading the premium for volume you didn't need.

It's a price instrument, not an inventory play. Plan it next to your floors, not next to your slot count.

Full guide on what it is, where it's allowed to fire, how to cap it, and how to measure it: https://bit.ly/4ieUU1I

You've spent all year optimizing for competition. Time to do it in person.

On September 24th, Playwire and Admiral are once again taking over City Pickle in Times Square for The Playwire Pickleball Open.

The details:
- When: Thursday, Sept 24th, 3:00–5:00 PM EST (arrive by 2:45 to warm up)
- Where: City Pickle, Times Square, NYC
- Skill level: All of them. We'll match you accordingly.
- Cost: Complimentary. Spots are curated and limited.

Ad tech is a serious business. Pickleball doesn't have to be.

Last event was a big hit, and this one's invite-only. Apply for a spot below. We'll review and confirm before the event.

👉 https://hubs.li/Q04t8_NF0

Most publishers find Google Search Console the way they find a slow gas leak. Something smells off, someone says go check, and the problem's been sitting there for months.

Backwards. GSC isn't an SEO tool — it's a revenue forecast one layer upstream of your ad server.
📌 Rising impressions: monetizable pageviews incoming
📌 Falling impressions: advance warning, days before your revenue reports notice
📌 "Crawled — currently not indexed": content you paid for that can't serve a single ad impression
📌 High impressions, low CTR: you ranked. You just didn't collect the click.

It's free. First-party data straight from Google. Fifteen minutes a week catches most of what quietly drains publisher revenue.

Full publisher's guide: https://bit.ly/4he181g