Facebook Campaign & Achievements

Although I can't share any screenshot related to my current work (NDA), to better demonstrate the work process I decided to run a performance campaign as an affiliate (using my own savings.)

Creatives - A/B testing

Angles are one of the most important elements (aside from images) when we advertise on the Facebook platform.

What the heck is an angle?
The angle is best described as the topic of a blog post or the particular problem you’re trying to solve. It is an insight from your customers – a burning pain or a dream – something important to them.

Having a good angle can make or break a campaign. In another word, the marketing ‘angle’ is how my messaging can connect with the target audience; It is not easy as it sounds. if you spend hours and come up with what you think is a killer angle, it might actually suck. Most of the time, the stupidest angles I thought would never ever work, end up the best campaign that performing 200% better than the one I thought is the best.

There is many ways to test, but this is what I think works best for me.

To test angles, I first dial in the best image to find the best CTR image. Start by doing 10 - 20 ads, with no angle, and see which one gets me the highest CTR. (see screenshot above)

After I find the best image, I duplicate the adset with the winning ad and moving forward to the next important element. The Ad copy for my ad Changing only 1 element. This ad copy should be various angles I’ve come up with for the product or service. (See how to brainstorm angle in 15 mins post). At this stage, If the team has a hard time being creative, I would hire freelance copywriters with different product experience.

My understanding of properly testing angles on Facebook is campaign structure and time. There is no shortcut to finding winner ads in a few hours, a day, etc. Especially important on Facebook, I find from many data-driven tests that it takes a total of 3 full days to properly optimize to a point where the best ad set ‘levels’ out.

Part #1 — Finding your winning image

Step 1) Set ‘Website Conversions’ objective; set it for optimizing for leads if I am doing lead gen, or optimize for purchase if I am doing eCommerce.
Step 2) Create your adset, set the targeting, and create your ad.
Step 3) Duplicate the adset 9x - 19x and change the image on all ads one by one, so they are different per ad.

Part #2 – Got the highest CTR (All) image? Good, now the fun begins.

Step 1) Duplicate the best adset as many angles as the team comes up with.
Step 2) Edit each ad, and place your angle
Step 3) Upload/save the changes.
Step 4) Let it run for 3 days before cutting anything, I typically test an angle at $20 per adset.

Additional Note

1. There are 2 main phases of the FB algo - Phase #1 is Learning, and Phase #2 is Optimizing.
2. The optimization of the campaign is on the adset level.
3. Give the algo time! Let it run for few days, I like to run at least 3 days
4. The algo doesn't like changes. Once I launch, I don't touch the settings of the ca

Optimize campaign

No matter how experienced you are advertising on Facebook, You can't predict exactly what is going to happen when you launch a new campaign. Of course, some parts of the campaign element are a bit predictable (proven creatives, certain targeting settings), 

After launching the A/B test for angles, the big question is when do I shut off, adjust or scale certain ad sets?

Learning phase
Facebook is giving us signals about newly launched ad sets, "the learning phase". During this phase FB is showing ads to certain "pools" of the potential audience to see how people respond to it (like? click? comment? spam report?). At the learning phase, I normally don't wait for the learning curve to be finished before shut off ads. But if I see some creative work (based on previous/similar tests) I keep them running (See screenshot above).

In the previous section, I talked about when launching A/B testing I mainly look at the CTR. The pixels and/or sub_id's (Please refer to Campaign Tracking post) I am able to see from which ads my conversions are coming. When the ad with the best CTR has the best conversion rate, the optimizing process is easy. but often the ad with a lower CTR (and higher cost) is converting better than the ones having a high CTR and low cost. This makes the whole testing killing and optimizing story a bit different.

If my most expensive ad brings in the most conversions, and my targeting settings are the same, here is where I spend time to find out why that certain img + ad copy combination converts better than another. For example, lay down different possible parts that influence the CR and find a way to reduce cost by testing new/adjusted/similar images, and/or go more aggressive with ad copy so CTR will be higher.

The next step is to test the best ads on different audiences (age range, certain regions, special interest) based on the previous projected target audience. 


The principles of scaling can be applied across different Dsp, One thing worth mentioning is as budgets go up the ROI goes down, it pretty much happens on every DSP. so it is important to have volume focused mindset. Spending $100k with 30% ROI is much better than $10k with 200% ROI. The core of scaling is having a good strategy and systemizing (breaking strategy down into small actionable item for the team)

Incremental Scaling: It is easy and safe. the algo won't punish me if I scale assets this way. The obvious disadvantage is slow. Each time I adjust the budget, the Facebook algorithm needs time to adjust the adset to the new budget. I can't scale adsets to $10k or higher right away, it takes few days. 

On Facebook, I select my winning adset, and increase the daily budget by 15%-30% at least every 2 days. Example:

Day 1 - $50, Day 2 - $50, Day 3 - $50
Day 4 - $60 (20% increase)
Day 5 - $60, Day 6 - $600
Day 7 - $72 (20% increase)

Adsets Duplication (My personal favorite)
The advantage of this scaling method is the ability to scale wider and faster, but one thing that needs to pay extra attention to is the audience overlap(Frequency in Facebook). You can't duplicate the same adset 20 or 30 times ( I paid a terrible price for this) because you will compete with yourself and it will cause a low delivery on your adsets.

Example: When I test and find out my winning adset, with $50 daily budget, I simply duplicate the adset 4-6 times with a 100%-200% higher daily budget, it depends on the ROI of the test the campaign, Wait 2-3 days after ads go live to see the results again, shut off the bad adsets, and repeat the process.

Additional Note:
The best time to adjust the budget or let the campaign running is between midnight to 1 hour after midnight the account's time-zone. 

Final words

As a performance marketer, I care about results. I don't care if the CPC is $0.1 or $10, as long as the performance is good, and I have a decent ROAS i'm good with it.

Real-time Lead Distribution Project

Responsible for Integrating essential elements and help to build real-time lead distribution solutions that allow clients to filter, set real-time bidding strategy, and deliver MQL in a matter of a few seconds after customers submit.

Campaign Tracking & Automation Project

I am lucky enough to be part of developing the software that able to track campaign performance from TOF to BOF with extreme accuracy & industry-specific campaign automation functions.

Professional Portfolio © 2020 Daniel Liu