By Rob Drummond with Jonathan Wilson
Estimated reading time: 5 minutes
Download a Print-Ready PDF
If you’re writing ads on a regular basis you can very easily get stuck writing the same ad different ways. The concept of volume and 80/20 applies. Rather than testing a few different ads, it’s better to test a wide number of very different ads, and quickly kill off the ads that look like non-performers.
This is the opposite to the way most ad testing is done. Usually only a handful of ads are created, and the different versions tested until a degree of statistical significance can be proven. That approach is still valid if you are testing a handful of proven ads against each other, (and also infinitely better than doing no testing at all), but if an ad isn’t proven you will lose too much time doing this. There are simply too many ads to test.
With a more regular iteration of ads you accelerate the learning curve. The adjustment here is rather than burning through $5,000 to find a good ad, you’re maybe burning through $500 (which your winning ad should quickly recoup).
The key idea here is to initially test more ads with smaller amounts of data, and limit your spend by using remarketing. An ad that works with a remarketing list will usually work for a cold audience, but you’ll spend less on remarketing to find that winning ad.
The amount of ads you can initially test depends on your level of website traffic. If you have thousands or tens of thousands of visits per day, you can quickly test out hundreds of ad concepts, and cut away losing ads in as little as a day or two. At lower traffic volumes you might test say 15 very different initial ads, and extend the initial analysis period to a week. The important thing here is to test a completely different ad concept in each ad, rather than different versions of the same idea.
Our friend Chad Prince gave us an example from his work with a dentist. One ad said:
Ad 1: I.T services for dentists: that’s worth smiling about
Ad 2: Your ex-wife still has a great smile. We are just that reliable.
According to Chad, both of those ads worked really well, but they approach the topic from a completely different angle. They evoke very different images in the reader’s mind. If you were creating Facebook ads for instance, you would use a very different image for these two ads.
Images matter a lot in remarketing. Your ads on the Google Display Network will mostly appear as banner ads. Facebook, Instagram, LinkedIn and Twitter are all highly visual mediums. Put some thought into your images – don’t just select an image as an after-thought.
It’s also important to test ads that draw on positive and negative emotions. You may find in your market that writing about pain, anguish, guilt and shame is more effective. But test some ads that also paint a picture of the situation your customers are moving towards.
There is a lot of interest in marketing in machine-learning and artificial intelligence. But the reality is that Google can never write ads as well as you can. They can manage bids better than you, but they can’t create ads in this way. Even if you work with an agency, ad creation is one of the things you shouldn’t completely outsource (even if you don’t actually setup and test the ads yourself).
In this initial testing phase, do not ‘set and forget’ your ads. Come back in a day or two prepared to hack away the ads that haven’t worked. You could think of this like farming your ads. Each ad is part of your crop, and because of the volume of ads you’ll be testing you need to pay regular attention.
You could also think of this like a brute force approach. One of the ways that people hack into websites is by trying hundreds or thousands of password combinations, which is a brute force attack. In the same way, you’re attempting to unlock a winning ad by throwing hundreds or even thousands of ad variations against the problem.
It is very easy to get attached to your ads – especially on Facebook where people reply to the ad. You have to cultivate a level of detachment about your ad performance.
A great source of ad ideas is to go into your archive of old marketing materials, and cross-pollinate winners from one medium to another. You could for example look in your email marketing system, and export a list of your most successful email subject lines. Writing a Facebook remarketing ad about each of your most successful subject line headlines is a good thing to test, because the concept itself has already been proven in a different medium. For each subject line you can then test a handful of very different images that evoke different emotions.
You should ask: where are you currently showing up regularly to your audience? If you do a regular Facebook Live, which of those have had the best engagement? Take that and create some Google Ads.
If you publish on LinkedIn regularly, find out which of your posts have had the best engagement, and create a series of Facebook ads. You don’t always need to reinvent the wheel – you may already have all the content you need in different formats. Any existing content piece that has an exceptional click to open rate is golden to work with. Successful emails often convert well to successful Facebook ads. You’re just leveraging data and content you already have, and applying a different twist to it.
The image from a winning Facebook ad will usually still work well on the Google Display Network. You can test this quickly by uploading the image into Google’s responsive display ad format, where Google automatically resize the image for the different display ad formats.
Once you have found a winning ad, you want to increase that ad’s exposure with more of your budget. Now the ad has passed the initial performance validation, you can dedicate more spend to it, and start to analyse its true profitability (rather than just making an initial judgement on click or conversion rate).
This is important on Facebook, because over time your successful ads will generate more engagement, which acts as social proof. Facebook rewards ads that generate high levels of engagement, so the more likes, comments and shares your ad has, the better. But you should only ramp this up with proven ads.