Building useful things with AI, code & automation.
Software engineer and founder of Luxmatico. I build practical tools with AI, code, and automation—and share what works, what doesn’t, and how you can build your own.
The step that stands out is asking ChatGPT or Claude to compare your cheapest and most expensive ads. One thing I would add is to give it the same fields for every ad: spend, purchases, cost per purchase, audience and placement, and ask it to list what the winners share only where the sample is big enough. Otherwise it tends to explain noise with a confident story. I would also log each idea as a hypothesis, for example "UGC-style hook lowers cost per purchase", then check that against the linked new ad after a fixed spend, say a week or a set number of purchases. That turns the loop into a record of what you actually learned, not just a stream of new creatives.
Yes, good ideas, thanks! I will extend with these for sure.
Btw I already have already small application which collect all the relevant logs of the business, like purchase from the invoicing system, phone calls from Bitrix, GA4, Meta and Google Ads data and so on, and create a summary every day from those. May I need to connect them. I will write a small article about that too :)
Charlie from xSquads
The step that stands out is asking ChatGPT or Claude to compare your cheapest and most expensive ads. One thing I would add is to give it the same fields for every ad: spend, purchases, cost per purchase, audience and placement, and ask it to list what the winners share only where the sample is big enough. Otherwise it tends to explain noise with a confident story. I would also log each idea as a hypothesis, for example "UGC-style hook lowers cost per purchase", then check that against the linked new ad after a fixed spend, say a week or a set number of purchases. That turns the loop into a record of what you actually learned, not just a stream of new creatives.