Your Ad Platform Is Grading Its Own Homework

Meta just quietly fixed a conversion measurement problem you didn't know you had. Starting November 12, the Meta Pixel gets a new on-device conversion recovery feature, and you don't have to change a single thing in your setup to benefit. Here's what's happening: some purchases never make it back to Meta's servers because of browser privacy settings, dropped connections, or tracking blockers. Those sales disappear from your reported numbers, which means Meta's algorithm thinks your campaigns are underperforming and optimizes accordingly. The new feature captures those missed conversions on the customer's device instead, using information the customer already provided (like an email address) plus temporary non-identifying device data. Neither piece of data leaves the device. Meta only learns a purchase happened and which campaign should get credit. If your reported conversions have felt low relative to actual sales, this is designed to close that gap, which could mean better campaign signals and lower cost per acquisition without you lifting a finger. You can opt out in Events Manager if you prefer. Full details here.

Now for the story that explains why "let the platform handle it" is a dangerous instinct. Business Insider sat down with three small business owners who've been running AI marketing tools in the real world, not in a case study. Their consensus is blunt: Meta's built-in AI has a structural problem because it optimizes for ad spend, not your margin. It's a restaurant critic who also owns the restaurant. ChatGPT and Claude produce generic copy unless you bring serious brand context to the prompt. The three owners agreed on one rule you should write on a sticky note: use AI for fast diagnostics and copy variations, but never let the platform evaluate its own budget or positioning decisions. That "don't let the platform grade its own homework" principle applies everywhere, not just Meta. Every ad platform's AI recommendations deserve that same skepticism. Read the owner breakdowns here.

Build your AI incident response plan before you need it. Shelly Palmer has a piece worth reading this week on what he calls an AI Apocalypse Playbook. The core argument: within 18 months, capable AI models will be accessible enough that even unsophisticated bad actors can cause real damage, and self-organizing agents doing unexpected things may cause more harm than intentional attacks. For small owners, the practical translation is this: if you deploy an AI agent and it sends wrong pricing to your whole customer list, or processes billing at scale with a bad rule, do you have a plan? Not just a technical fix, but a communication plan, a customer narrative, a chain of accountability? Your existing business continuity plan is the starting point, but it probably wasn't written with AI mistakes in mind. Palmer's full playbook framework is here.

-Scott

SmartOwner is published (almost) daily by the team at DigitalTreehouse. Want AI consulting or automations for your business? Reply to this email.