I’ve spent over $40m on digital marketing in the past 10 years, and anyone in my position will tell you the same thing:
- Creative testing is the key to growth
- It takes a lot of work to get useful results
In some cases, through lots of trial and error, it has been known for our clients to grow 100x over 6 months, purely by testing new variations of ad creative (text and images) until we find the winning combination.
*FYI: To solve this issue and get advanced creative insights at scale, we builtSpotlight.
Big rewards justify the big effort.
However, the process to report on creative performance and then incorporate that back into your process for coming up with new creative ideas to test is painful.
This is what a Facebook creative report looks like:
You get a big ugly list of ads with unintelligible names in a buggy, slow interface, where you have to click into each ad individually to see what you even tested.
So we’re stuck manually extracting the data, cleaning, segmenting and analyzing until it’s in a more digestible weekly report or presentation.
A/B Testing Ad Creatives At Scale
This is such a big problem that anyone who does creative testing at scale (10+ variations per week) develops their own system for making sure every test is marked up in the right way.
You’re constantly reusing and retesting your historic best performers: 5 new lines of copy for your best-performing image, or 3 new images for your best performing copy.
You want to tease out exactly which variables are causing an impact, and to do that you have to run a lot of the same images and texts changing one thing at a time.
So unless you look inside the ads at the individual “assets” (text and images) your ads are built from, you don’t know what it was about that ad that drove performance.
Facebook is on the right track with ‘Creative Reporting’ (which should really be its own tab!)
The platforms have started to adopt this ‘asset-based’ model, for example in Google Universal App Campaigns there is no ‘ad’ level at all: you upload your assets for them to mix and match!
As an agency of 45 people working with clients such as Booking.com, Monzo Bank and Time Out, we are testing thousands of variations a month.
So it’s critically important we have a system for tracking and reporting on the thousands of tests we’re running at any given time.
The Ladder UTM system works by giving each image (creative) and ad copy (message) its own unique name and identifier (c1, c2, c3 for creative, m1, m2, m3 for message) like this:
- LAD-a1-c1-m1: Business Travelers / Spiral eBook / Rock & Roll*
*Note: this has to go in both your ad name and your utm_campaign URL tracking parameters, otherwise you won’t be able to see performance in both your ad and analytics platforms.
So when we look in our analytics reports after the fact, we know how c1 the “Spiral eBook” image performed not just for the Business Travelers audience, but wherever else we used it.
With this naming convention, we can even pivot and segment the data to see how this image performed paired with all of the other messages we tested it against.
Marketing Creative Reporting Conventions
Without the right naming conventions, you won’t see anything useful in those reports, and you won’t be able to learn what worked (or didn’t).
Analytics is recorded once and forever: you can’t go back and change the data. If you made a mistake… guess what? Your data is trash.
It’s a case of garbage in, garbage out.
Let’s run through a real example: do dogs perform better in your ads than cats?
Maybe you’re a dog person and you’re absolutely sure they do, but it’s easier to convince your team if you have proof, and it’s always good to remind your boss of all the ROI you drive.
To set this test up, you have to go through and manually label every ad with “cat” or “dog” in the ad names and tracking parameters. There’s no other way.
If you skip this step, you’re going to be spending an awful lot more time manually inspecting ads one-by-one and performing excel wizardry to report on it afterwards.
That’s a lot of work to answer one question, but it’s worth it – you might find like I did on a past client, that dogs perform +40% better than cats.
One key insight like this can be the difference between getting a bonus or getting fired!
That’s just answering one question: but this gets addictive. You want to keep improving performance so soon enough you’re testing everything.
Introducing and testing variables such as different people, objects, colors, locations, backgrounds in the image, which corner you place the logo, whether you use positive or negative ad copy, different keywords or sentence types…
The list is endless and all of these variables can make a big impact on performance.
So long as you diligently plan these things ahead of time, you can make it work and get the insights you need.
…sure it’s a lot of work, but that’s the job.
All this effort is worth it – for everything you’re learning about customer behavior and all the weird and wonderful things that influence them to buy your product.
Remember this is why you got into marketing in the first place: the ability to experiment and see what really drives human behavior.
This rigorous testing is the difference between you as a data-driven marketer and the other 99% of marketers who just guess which techniques will work (and get it wrong).
Multivariate Creative Testing
So you double down and get even more sophisticated.
Knowing if it’s a dog or cat image is cool, but some of those dog pictures were shot in parks, some cat pictures might have humans in them. Shouldn’t we be layering on these insights too?
What we’re talking about here is not just looking inside the ads to get to the underlying assets, but going one step further again. What is IN those images and IN that text?
Maybe dog images work better in the park and cat images at home? Could park pictures work best with the owner present and if paired with a blue border and a ‘BUY NOW’ call to action…?
Now you’re in trouble.
When you start trying to layer multiple tags on top of each other in the assets, the system starts to break down.
You start to get ad names that look like this: “c1234-dog-in_park-with_owner_male_owner-smiling-wearing_hat-blue_background_red-border||m7483-free_trial-2019_award-low_cost_testimonial-BUY_NOW_CTA”
Not to mention the immense amount of time all of this takes to set up.
You start noticing things that are working, but ignore them because you don’t have the time to manually sift through all of your historical ads to see where else you’ve used that image/text.
Then a new person joins your team and they accidentally (or intentionally) break all your tracking by using the wrong case or naming convention!!!
There has to be a better way, and at Ladder… we’ve built it.
Spotlight – Automatically Report On Creative Performance via Machine Learning Tags
Spotlight is a creative insights tool that uses machine learning to do all of this tagging for you, across your existing ad assets.
Spotlight pulls in your Facebook data and automatically tags your ads with the various machine learning algorithms we use to see what’s inside your images and text.
You won’t just see if these tags are present in your ads (which would be pretty cool in itself) but you’ll actually be able to see how the presence of these labels affects performance!
Here’s what it can do…
See performance by what’s actually in the images.
Click a tag to see examples of which images were labeled with that tag.
Are faces present? What emotions were detected in those faces?
How does color affect performance?
Does positive, negative or neutral sentiment work best in your ad copy?
Do questions perform better or worse than exclamations?
How about the impact different types of entities have in the copy?
Finally, add your own custom tags to retroactively report on key initiatives.
Hopefully, you can see how valuable this is to you, as a marketer on the hook for coming up with new creative ideas and reporting on what worked.
This is the perfect report to put in the brief to your copywriter or designer. Imagine sharing this with your team when they are intrigued to see your campaign results.
Remember all of this analysis and tagging is done for you. You don’t need to adopt the Ladder naming convention – the name of the ad doesn’t matter at all anymore!
The Future of AdTech
Up until now, digital marketing innovation has been all about targeting, finding the right audience through creepy cookie tracking. I believe the next frontier will be creative.
It’s the last big lever left to pull.
Google and Facebook have automated targeting with machine learning, so our only refuge as performance marketers is to get really good at creative testing.
But you can’t get really good at creative testing without the right tools to handle that scale.
Spotlight empowers you to work in a whole new way as a modern performance marketer.
No more diligently tagging thousands of ads ahead of time to make sure you can report on the key insights – the system will label them and do the analysis for you, automatically.
This will free you up to do the higher-value activities, like using these creative insights to come up with new interesting things to test, on the path to a truly data-driven creative strategy.
Spotlight Beta Launch
We built this tool for ourselves, but we know that a lot of other marketers will benefit too. So by the end of the year, we’ll be launching Spotlight as a standalone self-service tool.
We’re looking for beta testers, so if you’re interested, sign up below and I’ll personally reach out to talk to you and set you up with access.