Why do some display ads earn better CTRs than others? Do some ads just get luckier than others or is there any explanation available that can answer how some ads perform better than others? To answer this question, we devised a scientific and statistical approach that tagged different ad components (like ad copies, ad creatives, logos and call to action) with numerical weights corresponding to their effectiveness in gaining CTRs. High performing ad components were again used in new display campaigns and voila – these ads performed like magic! And thus, Sokrati Creative Analytics was born.
Ad components and tagging
The Sokrati Creative Analytics process involves taking individual ads and tagging its different variables, right from the color of the image used to the size of the logo. With the help of a scoring algorithm, we rank each of these variables on the basis of their corresponding ad’s performance history.
Following are the different ad components that are tagged:
1. Text – Product specific (Tablet, LCD TV, etc): This text is tagged such that it gives us an idea of what is the product – A phone or a specific brand-name?
2. Text – Product Descriptive (Special offer, surprise gift, discount, etc.): This text is tagged such that it describes the theme of the Ad. Whether the Ad speaks about discounts, is giving a special offer, etc.
3. Image- Product Specific (Ipad, Bottle, shoes, etc.): What is it that the image portrays, i.e., the product being advertised.
4. Image – Product Descriptive (Green, Pink, Model, etc): Any factor that describes the image, i.e., whether it was a green colour stole, or was it a mannequin that was being displayed, or a real model, etc.
5. Logo size (small, medium, large): What is the size of the logo that was used in the ad.
6. Banner Size (320*50, 300*50, etc.): There are different banner sizes that are used according to the positions on the placement. It could be a horizontal above-the-fold ad, a side banner, etc.
Each of these ad variables are scored separately on the basis of the ad’s CTR. The individual scores are then aggregated in order to find out the overall variable score.
The final output can be seen in the form of the following format:
How is this process cost-effective?
The creative analytics process involves testing the effectiveness of variable set comprising of different ad components vis-a-vis another variable set of the components. But is testing so many ads for their performance really cost-effective?
Consider an example where there are 3 ad components and 10 variables of each component. So the total number of ads generated would be 10X10X10. If the cost of launching each ad is 5$, then the total cost for testing 1000 ads becomes $5000. Now that is a LOT of money.
So, instead of testing all 1000 ad combinations, we use a statistical technique called Design of Experiments that scientifically reduces the number of combinations required for testing. Through, Design of Experiments, the 10X10X10 combinations reduce to 10+10+10 combinations, therefore saving a lot of cost and time for the advertiser.
Once the Ads are launched on the basis of the DOE combinations, we monitor it through our dashboard. On the basis of the performance of each of the combinations, we figure out which variables work for the ads. Once these variables are understood, we keep iterating our DOE and try different combinations of the successful variables till we hit the right combinations that work for a particular client.