Two times more sales on Facebook by fine-tuning ads

eCommerce Case study: 2x more sales on Facebook by fine-tuning ads

eCommerce Case study: 2x more sales on Facebook by fine-tuning ads

Daniel Daniel

What do you do when an account that went well in the past starts to tank? When at the same budget you start getting less conversions but at a higher cost / conversion?

In this case study we present one of our favorite solutions, which led to doubling sales: changing ads.


In an account that worked well in the past, we saw that performance started to decline. Compared to the previous month, we recorded 50.59% fewer conversions …

… And with 48.27% less revenue:

Although we’ve already done all the budget and audience optimizations that we normally do, this being one of the oldest clients.

Other indicators, such as CPM (cost per thousand impressions) and CTR (click through rate), have also increased, so it has become clear that we need to step in and change something for this account because it’s starting to go seriously wrong.


Diagnosing a problem that has no obvious cause is done by elimination.

We evaluated the following possible causes and found that none explain the situation:

  1. the site loads without problems
  2. prices are unchanged
  3. ad sets have not been changed from the initial settings
  4. all ads are approved and there is no problem with audiences
  5. the ads shown are the best performing in this account
  6. campaigns have been running continuously

So where can the problem be? If the visible indicators won’t reveal the cause, we must look at the least visible ones.

Hidden indicators

On Facebook are some indicators, as important, as hidden. These are “Audience ovarlap” and “Delivery insights”.

Audience overlap shows what percentage of an audience is shared with another audience, so how much overlap. It is located in the Audiences section and becomes active if we select two or more audiences

If the overlap is large, there may be problems displaying your ads because we have multiple ad sets that are eligible to show your ad but only one can do it.

Here’s what the indicator looks like in our case:

Auction Overlap shows that audiences are very well delimited because they have an overlap of only 3%. This means that each ad set shows ads to unique people and there are no overlaps.

So we have to move on and look at the other indicator, “Delivery Insights”. This can be found by hovering on the “Delivery” column to any ad set with over 500 views:

Delivery insights show us, among other things, what percentage of the audience in that ad set saw the ad for the first time. The lower the percentage, the less people saw our ad the first time , so there are many who saw our ad the second or third time, maybe even more. This would be classified as a saturated audience.

The higher the percentage, the more people saw our ad for the first time, so few people have seen it more than once. Such an audience is unsaturated.

A small percentage of new people seeing our ads may exist either because there are few people fitting our audience within a certain amount of time, or because there are many new people but our budget is very large, or for both reasons.

This indicator is important because a saturated audience will react negatively to our ads, no matter what they are, because they see them again and again.

In our case the indicator looks like this:

It should be read from the bottom up because the bottom recording is the most distant and the top one is the most recent. So the numbers start from the bottom (a week ago) and end up (yesterday).

Progression is from 21.24% to 10.05%, which means that there where twice as many people seeing our ads for the first time a week ago.

And now we have discovered the real reason for the decline in ad performance: the audience is saturated!


There are many factors that can lead to an audience saturation, but in our case main two reasons were the size of the audience and the previous scaling of ads.

The audience has approx. 5,500,000 people, so it’s not small at all. But once you have a budget to spend because your ads are doing really well, any audience becomes small because you will burn through it much faster.

What to do in this situation? We can cut the budget. We can increase the audience by adding more general interests or Lookalikes, etc. But our solution is to change the creative (ads).

Changing the creative can solve the problems of a campaign.

We chose to change the creative because from our experience, once we know the audiences are good, the creative has the greatest impact on performance. And we know that audiences are good because we have used them successfully in the past.

Ad optimization

The first ads we need to improve are the ones that spend the most budget:

The most powerful element in an ad is the image/video, followed by the headline and then the text.

That’s why we first look at what images we have in ads, for what products and for which product categories and then we get new images from the following sources:

  1. From the site, from product images and (possibly) descriptions
  2. From the customer (who may have unpublished product images, or from the manufacturer, etc.)
  3. From processing current images in Photoshop (for example, putting a colorful background if it is now white, or adding accessories, etc.)
  4. From stock image sources (if they match)

After we obtain the images we have to evaluate whether it is necessary to process them or not. By processing we mean adding labels with prices or discounts, customer logo, etc.

There are some markets where only the kind of bold colorful ad goes. In other markets a simpler type of ad goes, with the product on a white background.

Types of ads

We’ve created all the main types of ads, from posts to carousels, collections and dynamic creative.

Dynamic creative are interesting because they involve creating up to 5 headlines and texts and adding up to 10 images, and the system will then display them in all combinations. The winning combination will be displayed preferentially.

After all, I was left with 20 ads of various kinds that I divided in ad sets.

And then we closely monitored the metrics in the account.


This change may or may not be successful. Not every new ad will work better than the old one, unless is more powerful and persuasive. Thea’s only one way to find out and that’s called split testing.

We expect dynamic creative ads to be most successful, because we have provided the system with more options to choose from, but the reality turned out to be different.

Static image ads (we call them “-img-” in names) perform by far the best:

Which goes to show that we can never never say for sure what works and what doesn’t and everything needs to be tested. Although dynamic creative is the best ad format on other accounts, two static image ads occupy the top positions in terms of conversions on this account.

The effect of ad optimization

In the first four weeks after placing the new ads over the previous four weeks the results look like this:

Conversions doubled …

The cost per conversion decreased by 28.92%…

The value of conversions increased by 117.15%…

And ROAS (or return on ads spent) increased by 48.98% …

The graph shows the time of launching new ads. The effect of optimizations on conversions (blue line) and cost / conversion (green line) looks like this:


To increase sales, all formats and as many variants of images and titles should be tested. This is not easy but the possibility of a big increase in sales shows that it is worth doing.


Leave a Reply

Your email address will not be published.