Deividas Mikočionis | Growth Hacker | Sugatan.io 

March. 18, 2019 • 10 min read

n this article we’re going to share a step-by-step process how we scale our clients’ eCommerce businesses using Facebook Ads.

We’re going to cover 9 essential steps we had to take to make this happen.

The Goal: Scale the account from 15k to 50k monthly ad spend maintaining 2x ROAS at TOF and +3x ROAS throughout the funnel.

Results

$520k+ in monthly sales with 3.79x ROAS throughout the funnel.

2.35x ROAS at TOF spending $100k+/monthly.

410k in revenue during black Friday with 7.27x ROAS.

…

Here is a quick summary of how we managed to achieve it

Focusing on the entire funnel and making the customer experience as frictionless as possible

Having creatives testing template to work from.

Utilizing the power of Facebook’s machine learning using CBOs

Building up bigger remarketing audiences for flash sales.

---

2018 should be highlighted as one of the toughest years for Facebook advertisers.

We’ve had GDPR laws being applied in Europe.

Cambridge analytica scandal, causing CPMs to skyrocket.

& Facebook shutting ad accounts left and right which don’t align to their rules.

BUT…

All of these new challenges bring new opportunities to the brands which care about the customers and have amazing products to display.

Reading this case study, we’ll show you how we managed to scale this particular brand from 15k monthly ad spend to 140k ad spend in only 4 months in 9 steps.

1. Customer-Centric Marketing Funnel

Immediately when we started off, we saw the lowest hanging fruit was – the entire remarketing funnel.

The brand had a very basic remarketing channel set-up, without using video views as audiences, splitting the audiences into the days they interacted with the brand, nor particular place they interacted with the website.

For this reason, we decided to split their remarketing funnel into the days they interacted with our mentioned brand and to the exact place they were on the funnel.

MOF up, MOF down, BOF (you can read this article to find out what it means if you don’t know yet) were split into 0-3; 3-7; 7-14; 14-30 days. For the MOF up we extended it up to 90days.

The theory behind is – not all website visitors are the same. Some of them want the product, but have some objections, so they add to cart but don’t buy.

To find out what were the objections of buying our product we did the following:

• Installed HotJar on to our clients website and while eating popcorn we watched how our customers behaved on the website. What did they read, where did they drop out.
• We went to read the messages we were getting on our Facebook channel and comments on our ads.

This is what we’ve figured out:

• Shipping price was one of the main objections.
• They wanted to know when they’ll receive the item.
• They were not sure if the brand was legit business or just another “cheap Chinese crap”.

We addressed these objections using our DPA and video ads at the time when it was appropriate to do it.

Also, we’ve addressed these objections on the website to make it convert better.

2. Testing Strategy. PPE vs Conversion Campaign

 

Over 8-months ago all of our tests were done starting off with post engagement campaigns. We would look at our cost per landing page view, cost per add to cart, cost per initiate checkout and purchases.

 

We figured out our metrics by using a ROAS calculator (which you’ll be able to download by joining our Facebook group) so that we know what we can pay for each one of the actions our customer take our funnel.

 

However, by using this approach, we faced one problem – we were getting inconsistent results.

 

Even if the customers were adding to cart on our PPE campaign, once we moved them to conversion campaign, they were not converting at all.

This makes a lot of sense looking at it from Facebook’s point of view.

Because Facebook has all of its users segmented into:

• Engagers
• Buyers
• Clickers
• Video viewers

Etc. etc. etc.

So even though the targeting was the same, Facebook was showing our ads to a completely different audience, which was post engagers, but not buyers.

No wonder we were getting inconsistent results.

Since then, we do our tests on the conversion campaigns from the get-go and have much more reliable data on our creatives and audiences.  

3. Letting Go of Our Ego and Letting Facebook Decide What to Do For Us

When we started our Facebook marketing career over two years ago, we were testing everything ultra gradually.

We would test placements, audiences, age, gender, countries, optimization, objectives and much more. 

In other words – everything that was possible to test on the platform.

The reality is – Facebook’s algorithm has become super intelligent with its’ machine learning and knows exactly how to find our ideal customers.

There are thousands of advertisers on Facebook who are feeding the right data across the globe and Facebook gathers it into one place to make use of it across all of the accounts.

The best part – it will only get better, so advertisers will have less and less work to do on the optimization side.

However, we were doing tests and these were:

– Single interest audiences. NOTE: We did not layer it with “Online Shopping” “Engaged Shoppers” and other similar interests. Facebook already knows who’s likely to convert if we choose “Purchase” optimization on your ad-set level.

– Optimization based on bid cap, value or target cost. We usually have all of these campaigns running at the same time on different CBO campaigns, however, for this account, we usually had “Highest value or lowest cost”  and “Lowest cost with bid cap” campaigns running. “Highest value or lowest cost” campaign was providing us with stability over the account and we were able to spend 3k daily hitting 2x+ ROAS on average.

We separate IG stories and automatic placements. If the account is mature enough, we don’t need to worry about choosing the right placement for each one of the ads. Instead, we choose an automatic placement and reap the benefits of sharing social proof across all of the ads.

4. Using CBO (campaign budget optimization) As Our Main Engine of Sales

As I mentioned in the previous point, Facebook doesn’t want us to create many hurdles for it to let the machine learning optimize.

At the same time, we saw an immense push from  Facebook representatives to use CBO and they were not wrong. The function works amazingly!

In theory, you create CBO campaign with a lot of ad-sets. Then you hook up the creatives on them and Facebook finds these audiences for you.

Well…

If that was the case, advertisers would not have a job anymore.

Here is how we approached the CBO campaigns:

1. Strategies to Apply When Using CBO

In order for Facebook to find your winning ad sets, it has to get data on the ad-set level.. The more ad sets you have, the more your budget is spread, hence less data it has.

So we have three options at this point:

a) Start with a much higher budget ($500+ on a CBO) to have more data coming to make it easier for Facebook to decide.

b) Have these audiences already tested and put only the working ones on the CBO campaign.

c) Start with a higher number of ad sets and reduce them quickly to the bigger ones and best-performing ones.

We used a combination of all of these approaches.

We identified the audiences that were already working for our clients and put them on a CBO. We started it with a higher budget to get more data from the beginning and we expanded the targeting so we have all of the hypothesis either approved or disapproved.

During the first 5-7 days, our ROAS wasn’t great.

However, we had this strategy in place, so we knew little by little it was going to improve and get our desire ROAS.

It took us around 5-7 days to trim down the non-converting ads and ad-sets to start getting stable 2x+ ROAS consistently at the TOF(excluding 90-day website visitors).

2. Hooking New Ads on the Already Working Ad-Sets.

Whenever we had a bunch of new winning ads coming in (which I’ll cover on the 4th point in depth), we would put them onto already working ad sets.

This gave more options for Facebook to choose which ads to choose from at the same time reducing the risk of ad fatigue.

Although we read the article from Facebook approaching it as a significant change on the ad-set, which would pull the data – it wasn’t the case. The ad sets seemed to underperform for a few hours or a day at max.

We tried a different option as well – start a completely new ad-set, however, it took too long for it to optimize to the level we wanted it to perform.

3. 40-40-20 or Small, Medium & Large Size audiences on CBO.

Facebook reps are constantly pushing us to give more data to the ad-sets and have at least 50 conversions on each one of them.

We tested it and it worked pretty well for us, hence we started to structure our campaigns in a way to achieve these 50 conversions on each of the ad-set every week (although over 30 conversions gave us enough stability as well).

To achieve this we had small size (1-5 million) audiences, medium size (5-20 million) audiences and large size (20+ million) audiences put on there started our CBOs with a then massive amount for this account – $800.

This type of campaign structure helped us to achieve:

1. The stability we needed to get sales consistently (smaller size audiences).

2. Audiences are big enough so we can scale the budget (medium and larger size audiences).

4. Budget Optimization and Refusing to Pay Zuck’s Grocery Bill

Since we started working, we saw a very strong pattern emerging on the account.

Whenever we were reaching 2x at TOF by the noon, we knew it was going to remain the same for the rest of the day, hence we increased the budget by 20-30% for the rest of the day and would look at the account the following day.

If the account didn’t reach 2x by noon, we would know something is off.

This being a case, we reduced the budgets by 10% forcing Zuck and his team to search us a better audience to serve our ads to.

Changing the budget so frequently may cause confusion to the algorithm, however, we saw it work on this particular account.

5. Using CBOs Across the Funnel

Seeing the potential of CBOs we started using it across the whole funnel all the way to the back-end, splitting the audiences into the days they had interactions with us.

Facebook was consistently finding us the customers who were the most likely to buy from us.

5. Waterfall Method to Test the Creatives

There is one irrefutable law in online marketing – always be testing.

The questions is…

Test what?

As I mentioned above, we didn’t test much on the ad-set level, however we took this rule to the heart for testing the creatives.

Every Monday of the week, our team of assistants would go through the creatives we have running and create a report based on that.

The report would be sent to our creative team and we would discuss what types of creatives worked, which ones didn’t, what was the possible cause of it and what type of creatives we could be testing for the following weeks.

By Tuesday next week, we would get new creatives and have a campaign set-up ready to go testing all of the small, yet super important details on the ads.

We tested the following things:

The Most Important Tests:

• The creative. Video ads, buzzfeed style video ads, photos, carousels.
• 1:1 ; 16:9 ratios for the creatives.
• 3-5 copies approaching each one of them from a different angle.

Second Most Important Tests:

• Thumbnails if we have any videos to test during the week.
• Buttons. We create the ads on PPE campaign so we can create the ads without any buttons, with buttons which are in the white background and other in the conversion campaign to test the headlines.
• Headlines. As we were testing a lot of ads on PPE campaign, we did not have to test the headlines at all times, however, we were still coming up with 2 headlines for every creative.

For these tests, we have our structure of the account set-up so we start testing with our 2nd-3rd best-converting audience, which happens to be interest based combo.

We test from 6 to 8 ads on the ad-set to avoid getting statistically insignificant results.

We would have the ads run for another 3 days to see what kind of results do we achieve by then.

The metrics we were looking at:

• Purchases. Obviously, if the ad had 3 purchases, we would call it a winner from the get-go and immediately transfer it to all of our TOF campaigns to see how they perform there.
• Add to Carts. We don’t get any purchases out of these creatives, we would go to our ROAS calculator and see if the metric aligns with what we can pay for this particular conversion.

Once the creative is shown to work, we move it down to our TOF campaigns and we hook them up on the ad sets that are already working so we have them for the weekend as it’s when most of the purchases come for this account.

If the ads were performing well for a few days, we move them down to MOF up, wait for a few days and move them down to MOF down, BOF and eventually to the back-end.

The underlying strategy behind “watering down” the ads is – we don’t want to be hitting our customers with unproven ads, potentially hurting our brand & hurting the sales.

6. Optimization Techniques to Prepare for Scaling

The main running CBO campaigns is – looking at the data from the campaign point of view.

If you start killing the ad-sets on there, you’ll lose the whole point of doing it.

So what did we do instead?

We were mostly creating, testing and killing the ads.

We were mostly creating and testing new ads.

We would go through the ROAS calculator (we used customized GoogleDataStudio for that) daily for each one of the campaigns, seeing which ads performed well and which ones didn’t in the past 3, 7 and 14 days.

The ones that aligned with our ROAS calculator would be turned off.

The ones that did – stayed on and were happily representing our brand 🙂

Saying that, we would still be looking at the ad-sets and if we had two ad-sets sharing the same audience, we would be killing the smaller one, leaving more conversion for the bigger one to optimize further (and get 50 conversions needed).

This strategy is closely connected to the previous point made, due to us killing the ads fairly quickly and relying on the new ones to substitute them.

7. Manual Bidding Strategies to Scale Up Our Campaigns

All of the things mentioned above were the rules we use to manage every single account we have.

Call them basics or ground rules.

However, the biggest scale was reached when we started implementing scaling strategies.

There is one, in particular, we wanted to share with you, which allowed us to scale the budget quickly to $14k+ ad spend days getting anywhere from 2x to 4x ROAS at TOF.

This is what we call “Get Out of My Way” strategy.

Here is how to set it up:

1. Find the audiences that are already converting well and combine them with the creatives already working on this particular audience. For this, we looked at the data on our main CBO campaign(40-40-20).

We chose in between 3 to 5 ads, to begin with.

2. Creating a new CBO campaign for each of the already working ad-set. Once again, we used CBO to scale using this scaling strategy.

3. We created 8-14 ad sets for each one of the audience we chose starting from 2x manual bid of target CPA  (have a look at your ROAS calculator for that) all the way up to 14x of our target CPA.

For this particular account we did 8x, however, on the other accounts we noticed it working with higher bid cap.

4. This one will be difficult to take.

Start running with the budget of 10x of your target CPA.

If you chose to go all the way up to 14x manual bid, the budget should go up accordingly to 12-14x of your target CPA.

5. If it works well – scale aggressively.