(4) Testing for Black Friday

(4) Testing for Black Friday

1. Testing: ABO or CBO?
2. What to test before Black Friday?
3. When to STOP testing

The main change (and this is very confusing), is that now we are not testing to find ONE WINNER and scale it,

We are testing so that we can build diversity, and that is very contradictory.

Why test if then we will build a CBO with lots of different ads?

_________________________________________________

Testing: ABO or CBO?

For the majority of cases, “lean ABO”.

Not an ABO with 30 ad sets,

Not an ABO where you test small details.

But an ABO where you can extract conclusions fast and learn.

Ad set 1_(Concept: Product image with discount)

• Ad 1: , ad 2:

Ad set 2_(Concept: MultiProduct image with discount)

• Ad 1: , ad 2:

The way to analyze this:

1. At the ad set level, I’m testing two different formats and understanding which one has higher CTR and lower cost per add to cart
2. At the ad level, I’m understanding retrieval:
   What ad gets delivery and why, why meta prefers one versus the other.
   Is it because it looks more organic?
   Or because the text is bigger?
   Is it getting more engagement?

Example scenario after 3 days:

• Ad set 1: 90% of the spend goes to the ad “buy 1 get 1 free”, which has better CTR → Because it has a person.
• Ad set 2: 90% of the spend goes to the ad “now on sale” which looks more organic, better CTR and CPM.
• Ad set 2 (multiproudct) has lower cost per add to cart than ad set 1

Conclusions:

1. As a concept, multiproduct ads with a discount are better
2. Meta andromeda retrieval prefers the multiproduct ads that look more organic

2024: here I would literally test every single ad,

Or, here the same, but a little bit better:

What to test before Black Friday?

Clear destination (maximum 7 CBOs):
CBO_(Evergreen)

CBO_(Evergreen + Discount)

CBO_(Black Friday)_(Recommendation)
CBO_(Black Friday)_(5 re-branded winner ads)

CBO_(Black Friday)_(ABO winners 1)

CBO_(Black Friday)_(ABO winners 2)

CBO_(Black Friday)_(Dynamic)

What to test as a fashion brand?

• Multiproduct collage (floating)
• Flat lay multiproduct (floor)
• Split screen model + product
• Collage split in 4 parts
• Video: Model wearing the product (full body or close up)
• Video: Product image with content in the background
• Video: Product + Client Review, Unboxing, Try-On

→ Possible discovery: only multiproduct ads work
→ Possible discovery: only images work
→ Possible discovery: showing a model lowers CPM

I don’t know what conclusions you will extract, but most importantly, think about DIVERSITY when you find 1 winning concept.

Imagine that “Flat lay multiproduct (floor)” works very well,

We CANNOT do 20 flat lays that look similar.

This will make us LOSE in retrieval.

So the conclusion cannot be that “I have to do everything split screen images”.

The conclusion is that single product ads, are better than multiproduct.

What to test as a cosmetic/supplement brand?

• Per angle / Per avatar:

Image/Video: Problem vs Solution

Image: Article Ad

Image/Video: Before & After

Image: Advertorial Images

Video: Listicle, Ranking

Image/Video: MultiProduct

Image/Video: Client Review

Image: US vs Them

Image/Video: Ugly Ad, Meme, Organic

Image/Video: Offer, Discount, Shipping

• Don’t do everything, you have to select

*If article ad is a winner, with that you cannot build an entire CBO diversified.

*If the approach “problem vs solution” works great, this can be expanded into diversification,

Because you can communicate “problem / solution” in multiple ways:

• Black Friday offer with “problem / solution headline”
• Product image with “problem / solution headline”
• Product image + result with “problem / solution headline”
• Before/After with “problem / solution headline”
• Client review with “problem / solution headline”
• UGC video with “problem / solution headline”
• Doctor video with “problem / solution headline”
• Organic ugly ad with “problem / solution headline”

Now there is not a specific way of testing,

Create a “BlackFriday_ABO” and run 15 ad sets maximum.

Extract conclusions that allow you to build DIVERSITY.

When to STOP testing

Challenge:

• If you stop testing on November 20th, you didn’t catch the real Black Friday demand, so you don’t really know what works
• If you stop testing on November 26th, you catched the Black Friday demand, but it’s too late to build the scaling CBOs and fill them up with 30 purchases before BF
  
• Stop testing around November 20th-25th

_________________________________________________

Should we re-test our bid caps, cost caps, … before BF?

Yes, but don’t make this test on November 10th, because that’s basically evergreen, do it when the auction is packed.

November 25th could be the sweet spot.

CBO(andromeda)CC100

CBO(andromeda)MAXVALUE