The Week Meta Overspent

🌟 Editor’s Note

Last Monday, Meta had a lot of media buyers losing sleep. Ad accounts across the board saw massive overspending (we were in the same boat). It wasn’t just manual bidding campaigns, automated campaigns with set budgets also overspent like crazy.

This explains why CPAs spiked so hard when we closed last week's performance:

The overspend didn't give us better performance, it just burned more budget on worse conversion rates. Let's hope this week runs more smoothly.

· · · · · · · · · · · · · · ·

Back in February, I mentioned Meta's new integration with Manus AI, the autonomous AI agent that started rolling out in Ads Manager.

Last week, on March 18th, Meta officially announced new beta tests across their family of apps to bring Manus to businesses, creators, and entrepreneurs.

Meta says the core challenge for advertisers isn't a lack of data, it's the manual work required to extract insights from it.

The Ads Manager connector securely accesses your performance metrics and turns them into presentation-ready reports. It identifies actionable insights and creates output formats like websites, dashboards, slides, and infographics.

Gives businesses access to over 2 million creators with authenticated, first-party insights. Manus evaluates creators and audience alignment to help identify the right partners for your marketing objectives.

Lets businesses ideate, generate, publish, and analyze content across Posts, Carousels, Stories, and Reels on Instagram. A unified workflow that eliminates friction between creation and distribution.

Manus is now available in the WhatsApp Business app as your "project partner for deep work, automation, and professional-grade results." It works in the background, connecting to your tools to get things done.

Example: A client messages you on WhatsApp. You send it to Manus. It checks your calendar, pulls up your pricing, and drafts a professional response you can send in seconds.

These integrations are designed to "securely execute multi-step, time-intensive tasks such as reporting, research, and content development, so business owners and marketers can stay focused on strategy and growth."

This is a beta. Early reports from agencies show the tool is prone to hallucinations. Don't trust outputs blindly, verify before you send anything to clients or use it to make strategic decisions.

· · · · · · · · · · · · · · ·

At the 2026 Upfront Summit in LA, Shopify President Harley Finkelstein said the company is preparing for AI shopping agents to fundamentally change e-commerce.

His thesis: Only 18% of U.S. retail happens online. AI agents can change that by acting as personal shoppers by discovering, comparing, and buying products with context that traditional search can't provide.

Example:

You search for running shoes on Google. You get Foot Locker ads, big retailers, paid placements. An AI agent learns you prefer On running shoes. Next time, it surfaces On products directly, not whoever paid the most for ad placement.

Finkelstein calls this "merit-based shopping." The agent optimizes for fit, not ad budget.

The data: Shopify's Q4 2025 earnings results reported AI-driven orders on Shopify grew 15x year-over-year in 2025. Gartner predicts 20% of all transactions will be executed by AI platforms by 2030.

If AI agents prioritize best fit over ad spend, smaller brands get a fairer shot at discovery. But only if your product data is machine-readable and protocol-ready.

Finkelstein said the initial rollout will be slow. But according to Shopify enterprise research, most companies don't have a formal strategy for AI-ready product data yet. The ones who do will be first to market when agents become the default shopping interface.

· · · · · · · · · · · · · · ·

ChocBox is a TikTok Shop brand doing millions in organic sales. Their problem: zero Meta presence, no Facebook ads, no data, nothing.

We took on a public challenge to scale their Meta account and hit $66K in revenue in 8 weeks.

The numbers:

Our approach: keep the native, organic TikTok feel, but add a structure optimized for Meta: stronger hooks, clearer setups, and storytelling that builds intent before the product shows up.

Each ad gave just enough context to stop the scroll, while still feeling native to the feed (not like an obvious ad).

I sat down with our Creative Strategist to break down hook analysis, retention curves, what statics are actually good for, and the angles that drove performance. Full breakdown is available on YouTube.

Want to be our next case study? Apply here.

· · · · · · · · · · · · · · ·

In the ChocBox case study, hook score was one of our leading indicators. If you don't stop the scroll in the first 2 seconds, nothing else matters. No hook, no hold. No hold, no conversion.

But most brands test 3-5 hooks and call it a day.

I've compiled 400 hooks that have driven performance across different categories. These are proven patterns you can adapt to your product and audience.

Reply to this email with "HOOK" and I'll send them over.

· · · · · · · · · · · · · · ·

That’s it for this edition. Meta’s getting more autonomous, ecommerce is moving to AI agents, and creative still matters more than budget.

Til next time,

Uri & The Growth Collective Team