∞

Meta is bringing generative AI to ads with:

  • Backgrounds for product images
  • Magic crops to stretch one asset across Meta’s many aspect ratios
  • Text variations

The first one is welcome news for any company with white background product shots, those don’t work well on social.

The last one is the AI-indication of an existing feature.

The second one is what I’m most excited about. Anything to streamline asset creation and optimization across placements is very welcome.

Magic crops plus text variation with more aesthetic automatic overlays would be a great next step to make it even easier to advertise across placements.

∞

A good listen on how to think about the current landscape of AIs and LLMs.

Add some Douglas Rushkoff quotes for additional context:

AIs are probability engines.

If AIs are based on us (our writing, knowledge, etc.), the only way to raise better AIs is to be better ourselves.

∞

Emotional labor has become a competitive advantage. Our commitment to showing up as a human, even (especially) when we don’t feel like it is precisely how we create value. And it’s this human work that helps us feel seen and valued as well.

-Seth

∞

Why am I extra bullish on podcasts right now?

TV and film producers prepare for potential post-strike production logjam

Once the actors’ strike ends the shows and movies still need to be made and the backlog will be big.

Content owners (Disney, Nickelodeon, Dr. Seuss, PBS, etc.) have been turning to audio to fill the entertainment gap and ad dollars are pulling back from TV because of the lack of new content.

The longer it lasts the more likely habits are to change in a meaningful way towards channels producing new entertainment.

Shocks cause shifts.

∞

More social shifting to messaging as Meta starts testing broadcast channels on Facebook, following the features success on other platforms.

Distribution of social content by users is increasingly moving to private channels.

∞

Everything old is new again, which is why Media Mix Modeling is the new attribution.

measurement has been so hampered by recently enacted privacy restrictions that Meta, Google and Amazon are finding that any measurement tool is better than not demonstrating attribution at all.

∞

Marketing nerds unite! A chewy piece comparing ecomm promo language performance.

Thrown in the ring:

  • $ off vs. % off
  • Free vs. Fast delivery
  • Promo code error codes
  • “Black Friday” vs. “Cyber Week”

And maybe some more ideas you can steal.

∞

Google is further signaling the decay of attribution modeling with the announcement:

First click, linear, time decay and position-based attribution models are going away

None of these were great. Data-driven replaced position-based and GA4’s shift to user (and event)-based measurement over session-based made first click redundant.

But I think this is an admission from Big G that (aside from last click) accurate attribution is hard to do in a post-cookie world, so it’s all about data modeling now.

∞

Streaming’s big pain point: there’s too much stuff to watch
/
Despite more available programming than ever, 1 in 5 viewers abandon streaming sessions out of frustration, according to Nielsen.

The Great Unbundling into the infinite shelf space of the internet still has a major problem: discovery.

Modern search has largely the same issue.

Maybe AI can fix it, but what does the knowledge graph underlying the model look like?

Pandora always underwhelmed me when, and this feels like the likely analog for TV and movies. But podcasts could be prime for AI discovery thanks to LLMs and transcription.

∞

TikTok doesn’t want marketers obsessing about last-click attribution. In part because TikTok looks really bad when viewed as a direct performance channel. But also, attribution is less helpful these days.

It’s not about what a specific channel does. It’s about what it adds to your marketing mix.