As I talk to more SMB advertisers, I keep seeing a pattern which I call the 𝗗𝗮𝘁𝗮 𝗣𝗮𝗿𝗮𝗱𝗼𝘅 — where chasing more data leads to worse outcomes.
As I said in my last post: modern ad-tech has moved from a white-box system that was hard to operate but easy to reason about, to a black-box system that is easy to operate but hard to reason about. With Meta removing most of the manual levers, the one control left is creative — so agencies test creatives, aggressively.
One small advertiser showed me their dashboard: ~$100/day budget, the agency was running 40 creatives in parallel! Dozens of creatives with under $10 of spend each. They asked, “Why didn’t they get a purchase for creative X?”
So I asked their historical CPA. $45.
That’s the whole problem. Splitting $100/day across 40 creatives is ~$2.50 per creative per day — but it takes ~$45 of spend on average to produce one sale and you may not even get that with unproven creatives. Most of those creatives would need weeks of the current budget just to have a fair shot at a single purchase. Asking why $10 of spend didn’t convert isn’t a data question — 𝘁𝗵𝗲𝗿𝗲’𝘀 𝗻𝗼 𝗿𝗲𝗹𝗶𝗮𝗯𝗹𝗲 𝗱𝗮𝘁𝗮 𝘁𝗵𝗲𝗿𝗲 𝘆𝗲𝘁.
Testing lots of creative variations and staring at metrics 𝘧𝘦𝘦𝘭𝘴 productive. But data only helps when it’s reliable — and I always say 𝘂𝗻𝗿𝗲𝗹𝗶𝗮𝗯𝗹𝗲 𝗱𝗮𝘁𝗮 𝗶𝘀 𝘄𝗼𝗿𝘀𝗲 𝘁𝗵𝗮𝗻 𝗻𝗼 𝗱𝗮𝘁𝗮, because it invites confident wrong decisions.
Here is the fix: use marketing intuition to cut 40 creatives down to 4, then spend enough on each to gather 𝘀𝘁𝗮𝘁𝗶𝘀𝘁𝗶𝗰𝗮𝗹𝗹𝘆 𝘀𝗶𝗴𝗻𝗶𝗳𝗶𝗰𝗮𝗻𝘁 results — and then decide.
When was the last time someone showed you 𝗱𝗮𝘁𝗮 you would never actually make a decision on?
P.S. Words are mine. The Data Hungry monster is made by Claude.
