October 30: The ad eating your budget, clicks that go nowhere, and the Sunday myth
Three things we learned running ads this week. Short version: the numbers you look at first are usually not the numbers that matter.
-
One ad can quietly eat your whole budget
Meta does not split your budget evenly across your ads. It picks favourites. This week we watched one video ad in an account soak up a huge share of the budget while dozens of other ads barely got shown. Meta clearly loved it. Customers did not. It cost more to show (the CPM was far higher than the rest) and the cost per sale was worse than the ads sitting underneath it starving. A handful of those starved ads were actually selling, and cheaply. They just never got the chance.
The tricky part is that the campaign looked fine at the top level. You only saw the problem when you opened the ad level and sorted by spend.
The favourites got the money. The ads that sold cheaply got scraps. Once we paused the favourite, the budget flowed to the other ads and the cost per sale came down within a day.
What to do: Once a week, open your campaign at the ad level and sort by amount spent. If one ad has most of the money and a worse cost per result than the ads below it, pause it. Meta's favourite is not always your customer's favourite.
-
A great click through rate can be a bad sign
An ad in another account had a click through rate that would make any marketer proud. Dig one layer down and the story changed. One placement, Facebook Reels, took a small slice of the budget and produced more than half of all the link clicks. Almost none of those clicks turned into a sale.
More than half the clicks came from one placement. None of the sales did. Reels traffic is cheap, scroll-happy and quick to tap. That inflates your click numbers without doing anything for your revenue. If you judge the ad by clicks, you will scale the wrong thing.
What to do: In Ads Manager, use Breakdown by Placement before you celebrate a click through rate. Judge every ad on cost per purchase (or cost per lead), not clicks. If one placement is producing lots of clicks and no sales, exclude it or move it to its own ad set with its own budget.
-
Everyone "knew" Sundays were bad. The data disagreed.
A client and I both had the same gut feeling: Sundays were rough, and maybe we should just turn ads off that day. Before touching anything, we pulled daily results since September into a spreadsheet and grouped them by day of the week.
Sunday wasn't the problem. It brought in more sales than any other day, had the best conversion rate of the week, and was one of the cheapest days to reach people. Thursday was the weak one, and not because ads got more expensive. Ad costs were normal. People just bought less on Thursdays.
Thursday's cost per sale spikes because people stop buying, not because ads cost more. Sunday holds its own. One wrinkle worth knowing: Meta credits a purchase to the day someone clicked the ad, not the day they bought. So a strong Sunday might partly be people who clicked Sunday and bought midweek. Which is exactly why pausing Sundays would have hurt.
What to do: Before you pause a day or dayparts, export daily results and pivot by weekday. Look at conversion rate, not just cost per result. If a day really is weak, trim the budget by less than 20% instead of pausing, so you do not reset learning. Gut feelings are a hypothesis, not a decision.
That's it for this week. If something here saves you money, reply and tell me. If you think I'm wrong, reply and tell me that too.