-
How many impressions is enough to be statistically significant enough to eliminate a creative?
Posted by on July 19, 2024 at 12:49 pmI'm trying to find the best images to use in my ads. I have 60 images spread across 6 META ads (each ad can hold 10 images).
I'm trying to eliminate the worst performers based on number of clicks. The number of conversions isn't high enough to factor in as most don't have any conversions.
My problem is that when I look at images which have received 0 clicks, all of them have received less than 100 impressions total. There are some images with over 1,000 impressions.
First question:
For these low performers that are also below 100 impressions total, how many impressions is enough to eliminate them? I have some as low as 9 impressions! Surely Facebook can't know if an image is bad based on only 9 impressions.Second question:
I'm thinking about pausing the images that already have 200+ impressions so that I can get more data on the poor performers. Is there a way to pause just an image within an ad, or would I need to build completely new ads to do that? -
1 Reply
-
I like to use this calculator: https://neilpatel.com/ab-testing-calculator/
-
Your problem is you have 60 creatives for 1 campaign lol
-
Rule of thumb: 500-1000 impressions for statistically significant results, but it depends on ad spend.
-
You have way too many variables to test.
you need to reduce the number of creatives to maybe 4 or 5. Run that set for a few weeks. Pick the one that works best. Test that against another 4 or 5. Etc.
It takes time. Sometimes it takes a really long time. You can’t hurry the process up unless you’re willing to spend $$$$.
-
There’s a lot here, you’re doing a Dynamic creative test. You didn’t mention budget but I doubt you’re spending enough to test 60 ads at once.
I like to test a ton of ads too but when you have 60 ads, if you’re spending less than a couple hundred a day you gotta do it in phases. Like test 10 one week, pick a winner, test another 10 next week etc
How do you pick a winner? I’ll pick my winners based off CTR & CPC – if however in your example you have some ads with impressions and others without you can do 2 things. Trust facebook does know that the ad isn’t worth shit after 9 impressions, or break out the low performers and put them all in one ad group and give them a budget to see if anything happens.
Spoiler alert, these platforms are pretty smart and I usually trust them to tell me which creative is worth it or not.
-
5x-10x your cpa / AOV if your rlly want to have accurate data. However an experienced marketer will be able to tell if an ad will work with 3x cpa /aov.
So if your selling 10$ aov items, you should know within 50$-100$ in spend if your ads are “working”
If your selling 200$ items , it will take a solid 1k to 2k in ad spend to rlly know .
However as I mentioned , experienced marketers will be able to know within 3x AOV in spend as most know their ad metrics and what a profitable ad looks like
-
CTR, or Click-Through Rate, measures whether your image stops the scroll. I typically run my ads until they reach 500 impressions due to budget constraints, but higher would be better. Once they hit 500 impressions, I check their CTR. If all are below 1.5%, I would start a new batch.
The reason why some ads get lots of impression other get few is because Facebook doesn’t evenly distribute the budget. What i do is I stop ads at 500 impressions to allocate budget to those with fewer impressions, or I run each ad in its own ad set to evenly spread the budget without using campaign budget optimization.
I recommend getting at least 500 impressions before deciding. If budget is tight, a traffic campaign might be more cost-effective for gathering data (that’s what I did). However, don’t expect immediate sales—it’s possible but unlikely.
-
I like to keep it simple. Flat 200 clicks minimum over at least a week.
Comparing ads on rotation isn’t quite like a proper stat sign CRO test. Sometimes you just eyeball it and go “that’s enough data”. Waiting for perfection is a good path to getting no testing done
Log in to reply.