Home > Paid Traffic Sources > Facebook & Instagram

---

Facebook Ads: How to test images/copy in a statistically relevant way (7)

---

Hi guys, this is my first post and I ask for help because I have a doubt about how to test different images/copy on Facebook Ads.


I have read here that each Adset targets a specific pool of people.


I have also read that to test variants effectively, only one Ad should be used for each Adset.


Now, suppose I have a lookalike of 320,000 people and I want to test the same copy but with 3 different images. I have to create 3 Adset with the same audience, with one Ad each, with the same copy each but with different image.


After a week the Adsets have different performances: € 1 CPA, € 1.5 CPA and € 3 CPA.


Now, how can I understand that the Adset that has obtained the best result, 1 € CPA, has reached it because it is the image that works better and it is not the pool of that Adset to be more in target?

---

You don't know. But within the 320.000 audience it is likely people have common interest/behaviour etc. So yes, you can say the 1 cpa image did best when it comes to hitting the best CTR. Doesn't mean it ends up being the ad bringing in most conversions ofcourse.

I mostly test 2-3 images per ad set, with same text/targeting/setup by the way, but everyone got his/hers own preference. Best not to exceed more than 3 images since FB will focus on certain images only and give others not the right attention in an ad-set with 6 images for example.

---

Hi Stickupkid thank you for your reply.
It is true that often high CTR = low CPC = low CPA but it's not always true.

Why have I to choose the best Ad looking to CTR and not to CPA or ROAS (that is the final goal of selling campaigns)?

For example, with an "open bank account" white offer if i use two images, a good looking girl and a credit card illustration, we can suppose that the girl will have a higher CTR but at the same time a lower CR too because the people i let enter are less interested in financial products.

What's wrong?

---

Honestly I have no idea since I always choose traffic to website, never on conversion or such. A high CTR means you hit the right audience for sure, with the most attracting creative. Again, the best performing image doesn't necessarily mean it gives the most conversions. Make sure your whole funnel is consistent so people are not surprised when clicking through to the prelander and offer page.

Why not combine the creditcard illustration with the girl for example in one image?

---

This is a good idea .
Anyway i'm a newbie in AM so maybe i'm saying something deeply stupid but i think that CPA and ROAS reflects the performance of all the funnel (the consistency of AD with Landing with Offers ecc.) so should be the mother of the KPI to observe and optimize.

---

My bad. But you are right eventually yes. Don't look at CTR only ofcourse, look at which ads contribute to your conversions. Try to find the balance between low CPC and high CR!

---

Great discussion!

I'm not yet an expert in FB, but will add what I know.
@mobbyagency If you really want to be more certain about ad performance, you can duplicate each adset one or more times, then take the average of all of the same ads across those adsets.

Better, you could just pause the adsets and ads that are inferior, and leave the superior ones running.

A potentially more effective approach would be to just forget about striving for that extra accuracy when split-testing. Instead, just make up for it by testing a lot more ads - 3 of the same ads in each adset, and different ad between adsets (i.e. the "normal" or "typical" approach).

Then just keep on 1)adding new ads, 2)duplicating the best ones (to scale) and 3)pausing the worst ones.

It's also important to note that decisions should be made based on the most-recent stats and not lifetime stats, as ads will saturate.

As for whether to cut based on CTR or ROAS: Of course ultimately we want to optimize towards profits, but if we keep an eye on CTR we can eliminate the worse ads earlier to save on spend.

Do keep in mind though (as you've pointed out) that lower CTR can/may result in higher profits. So when cutting based on CTR, we need to be a bit lenient so as to factor this in, by only cutting ads that have CTRs so low that you know they'd be unlikely to turn out superior/profitable.

Cutting too aggressively can result in some profitable ads being cut. Cutting not aggressively enough will result in more spend. Experience will help you to identify a sweet spot somewhere between these 2 ends of the spectrum.

Hope that helps!



Amy

Sent from my SM-G930W8 using STM Forums mobile app

---

Home > Paid Traffic Sources > Facebook & Instagram