Evergreen campaigns: the perfect optimization cycle for Facebook ads

Not surprisingly, there are 1001 „truths“ about the optimization cycle of Facebook & Instagram Ads. As is the case with any perfect account structure, a meaningful setup is key here as well. In order to work out the perfect optimization cycle for a given Facebook Ads setup, we teamed up with Robert and Elodie from nu3 to develop the following questions:

• Which conversion type is used?
• Are we dealing with a seasonal campaign?
• How long had the previous campaigns been running on the basis of which decisions were eventually made?

Of course, every strategy must match seasonality alignments as well as the choice of conversion and advertising types. Nu3 faced in particular the following problems:

• There are seasonal products such as weight loss products (especially popular in January), which so far have only brought satisfying ROAS in the corresponding season. How could you let these campaigns „learn“ even before the season starts?
• Under internal rules, running campaigns were generally exchanged after a month or so for new ones. By doing so, however, it was not possible to gather sufficient data for optimization.
• For longer campaigns, rigid optimization time frames were set that were heavily based on dates and too little on data.
• If the performance was deemed too poor, campaigns were often terminated before any optimization could be reached. The reason for this was that a duration of two to three weeks was believed to be sufficient.

But how do you know that a given campaign has gathered enough data? When should it be paused, restarted or stopped entirely? Together with Nu3, we have developed a smart approach to the perfect optimization cycle!

Using as an example the “Crunchy Peanut Butter Campaign”, an all-season successful evergreen product, we will present our approach to the perfect optimization cycle.

First, we set the campaign to have no end date, thereby not having to pause or restart it. Accordingly, the campaign would be able to run indefinitely. As we were dealing with an evergreen product that could be marketed all year round, we opted for a 4-week optimization cycle. The following parameters were to be optimized:

1. Optimization of the advertising measures: If certain stylizations or types showed good results in terms of ROAS, new advertising measures of the same type were created.
   
2. Optimization of the target groups: New suitable interests were continuously added to our interest target groups during the campaign period. Furthermore, it is important to consider pursuing a new direction if the results are unsatisfactory. For lookalike target groups, the quality of the imported data is absolutely crucial. Accordingly, an optimization cycle could look like this:
   1. All buyers
   2. All buyers that have bought a corresponding product
   3. All buyers that have bought a product several times
   4. etc.
      
3. Optimization of conversion settings: During the course of the campaign, more and more data was generated via the users’ conversion funnel. Based on this, we were able to continue optimizing towards the lower funnel (e.g. shopping cart placements instead of product page visits).

After our analysis, operating on a four-week basis, we evaluated those three above-mentioned parameters in the third week and made adjustments from the fourth week on. In a second round, we evaluated the same three parameters in the third week and made adjustments from the fourth week on again. After just two optimization cycles, we were able to increase sales by 20% while lowering advertising costs, thereby increasing the ROAS by up to 15%.

Applying to all campaigns, we recommend an optimization cycle based on the number of clicks and conversion events. For the perfect optimization cycle, we always recommend nClicks => 1,000 as well as conversion events that are set individually matching the respective industry. In the case of nu3, for example, the conversion events were decided to equal 15 purchases. Depending on the budget and cost per click (or cost per conversion), the number of days / weeks adjusts to the aforementioned optimizations.