In the last couple weeks of Q1 2018 to mid Q2 2018, we ran a digital campaign to help a telco in Southeast Asia grow the user base as well increase the usage for existing customers. There were 2 types of campaigns that we launched for 2 different types of products. One is a starter pack to acquire new users and another is a data package to keep existing customers engaged.
When we launched the campaign, we started off with a very general strategy by splitting the ad sets into 2 set of audiences: existing users for the Data Package campaign and non-users for the Starter Pack campaign & Data Package campaign. We only used a single image as the ad type,with 2 variations of landing pages, and split the ad sets targeting male and female audiences. In the first few weeks of the campaigns, we’re getting quite expensive CPA ranging from $9-$10 against our KPI at $6.1.
3 days after the campaign started, we introduced Carousel ads which instantly impacted to a very good CPA from above $8 down to $3.77. However, the impact didn’t stay for too long before the CPA increased to above $7 in the following days.
We introduced a couple new strategies such as adding new creatives specifically for IG Stories as well as 3rd variation of the landing page for testing purposes and adding video in addition to just still images. This means that we stretched the campaign from running on FB and IG Feed only to include IG Stories too in the middle of the campaign.
Targeting wise, we’ve improved from just users vs non-users to further narrowing down the targeting with mostly Custom Audiences (CA) and Lookalike Audiences (LAL). We introduced the following 5 new sets of audiences to bring user from upper funnel to lower funnel and even remarketing to those who have purchased to make repeat purchases – because we’re selling data packages.
- CA: Website Visitors
- CA: Cart Abandoners
- CA: Purchasers
- LAL1% Purchasers
- LAL4% Purchasers
In addition to that, we’ve utilized our proprietary Automation Engine starting in week 6 to help oversee the campaigns especially over the day nights and weekends to complement the daily campaigns optimization.
Since then, we’ve been seeing improvement in performances and efficiencies as shown below:
Having implemented the first version of the Automation Engine, later we found out that the performance was generally better on weekdays compared to weekends. Hence, we improved the way we triggered the automation engine by implementing stricter rules (decrease bid by 2x higher percentage and decrease budget by 2x higher percentage if the CPA is above the KPI) in weekends compared to weekdays. Thus, the Automation Engine v2.
After we’ve implemented all the necessary changes to improve the campaigns’ performances, we’re seeing a very good trend in CPA in weekly basis up to the very end of the campaigns.
Although we are not able to hit the KPI given by the client, we’re progressing well from week to week to get us really close to the client’s KPI and the following are the learnings that we got from running the campaigns.