Locations
Upon analysing the sales data, we decided to separate the targeting of Warsaw + Masovian voivodeship, and the rest of Poland. Campaigns aimed at key locations (stars) had their own budget for every promotional channel.
Target audience
We divided the target groups based on the pursued aim, as well as new and returning clients. We delivered different messages to those who just found out about the company than to those who had been familiar with it already. For the purpose of optimisation, we excluded the internal traffic, competition, or those users who might have visited the website multiple times but each of their visits was very short.
The campaign was optimised based on demographic data, target audience on the market, and specific personas. We also took advantage of the groups similar to the remarketing lists, both in the case of search ads and GDN ads.
Schedules
We were optimising the campaign based on its effectiveness during different times of the day and days of the week. We divided each day into smaller periods, within which we were adjusting the CPC bids.
Devices
We were optimising the campaign on the basis of the device used. In some cases, we created separate campaigns for desktop and mobile users.
Weather
The bids were adjusted every hour, depending on the current weather.
A/B testing
We ran A/B tests for different bid strategies, and tested different content when optimising the ads. In the case of GDN, we didn’t test different banner types – but we focused on the optimisation of ad targeting methods instead.
Exclusions
When optimising campaigns, we paid attention to excluding non-matching ad placements (in the case of GDN or YT campaigns), as well as negative keywords (DSA and PLA campaigns). Each week, we updated the list of such phrases and placements, until we achieved the desired results.
Machine Learning
The key campaigns aimed at increasing sales used machine learning and advanced bid strategies for the purpose of their optimisation.