After investing heavily in Meta advertising to launch a new product, the author found that traditional marketing work could be rethought through AI and more disciplined testing.
At a global fintech in London, the author served as director of growth, overseeing technical marketing duties such as measurement, signals engineering, analytics, and performance tracking. The company had a budget of tens of millions of dollars per month and a large marketing team, including dozens of developers. Now, the author says, many of those tasks can be handled with AI. Where the author once wrote SQL and relied on dashboards, much of the work now happens through chatbots.
The author applied this approach to RunInk, an app that turns running routes into posters, spending about $1,700 on Meta ads in the first month. Sales reached approximately $200. The author acknowledges that testing is involved and that initial results may be weak, but the approach became more profitable as Meta’s systems gained more learning data.
Previously, dozens of designers produced and tested many creatives. Now, AI can generate many variations and Meta can test them. AI-generated visuals are not always visually appealing, but with the right prompting, AI can produce concepts even for those without design experience. The author could focus on promising variations and broader testing.
Signals engineering was another area that changed. Ad platforms depend on machine learning, which requires sufficient optimization events. Meta needs about 50 optimization events a week to move out of its learning phase. New products may start with too few signals, especially when selling few high-ticket items. The author suggests breaking the funnel into smaller steps, assigning monetary value to each step, and gradually shifting optimization toward purchases.
Question Your AI Like You Would an Intern
AI is a powerful tool but can make mistakes. It should be questioned, validated, and given appropriate context. The author would ask team members whether reported numbers seemed high or off. The same probing approach should be applied to AI. The author also notes that AI often lacks context, so important details should be supplied.
Use Advertising to Get Your First Customers
It is difficult to build a product without users. After the first 100 or so users, the author surveyed customers who had made posters and found demand and satisfaction, though many considered the free digital file sufficient. The author then offered a digital product, which became a best seller. Finding the right price point and product requires talking to customers.
Budget for More Than $1,000
Launching a new product with a fresh brand may require more than $1,000 in advertising. The author describes the investment as a useful way to direct people to the product.
Key Takeaways
- Marketing tasks that once required dozens of developers and designers can now be handled solo, with AI generating ad variations and Meta testing them.
- Meta needs about 50 optimization events a week to learn, so break the funnel into smaller steps, assign each a monetary value, and gradually shift optimization toward purchases.
- Question AI the way you would an intern, give it context, and talk to customers to find the right price point and product.