I audit a lot of FB ad accounts, which means I get to see what is and isn’t working, and I can also spot patterns in terms of common mistakes made.
One of the least sexy bits of advertising is testing – but it’s also pretty much the sure-fire road to consistent improvement.
One of my very first clients initially signed up with me just to ‘freshen up the ads’ – they ran their ads in-house, but fancied getting an expert opinion on how they could step them up a notch.
In the 1 month I had access to their account, I set up a bunch of tests. Later, when I was retained long-term, I could see the result of these tests.
In one campaign, the single tweak I made (using lead forms instead of a conversion ad) had brought in an extra £4200 from a £25 per day ad spend.
That was without long-term optimisation, and just with one change.
Here are the 9 most common mistakes I’ve seen businesses make when it comes to testing, and then tips on how I would recommend you should be testing.
Most common mistakes made:
1. Not testing – just putting stuff out there and leaving it
Clearly, if you never test stuff, you’re never going to improve your results. This typically comes down to business owners just not having time, but even big clients – such as an influencer brand I worked with who were spending £200+ a day – are guilty of this.
2. Not being systematic – throwing out loads of different stuff without knowing how to analyse it or what to do with the info
No better than not testing, really. A scattergun approach might get a win or two, but if you don’t understand why or how you did it, then you can’t replicate the win. An e-commerce store owner I worked with had this problem, he would get great results from one campaign out of every 10, and never know which one it would be.
3. Sending the traffic somewhere useless – i.e. optimising for lower link clicks, but then sending people to your homepage
I’ve seen people optimise their ads endlessly, but they continued to send them to a boring, unfocussed landing page. I even audited a charity whose ads were being run by an agency. They were getting nice low numbers for their reports, and then they blamed the lack of conversions on the charity’s follow-up team.
4. Not testing towards the final objective
Conversion ads are really powerful, but people often stick with traffic ads, focussing on click-through-rates rather than opt-ins or sales.
5. Running tests at the ad level, not the ad set level
I don’t know why, but gyms seem prone to this mistake. They whack a whole bunch of different ad images at the ad level, but since the way that FB chooses a winner is flawed, you’ll end up with one or two images getting all the impressions, and others are seemingly ignored.
6. Using a saved audience, and sticking with it
I audited an SEO company’s ads. They had a clear picture of their avatar, so they created a saved audience of interests that were all business/marketing/SEO-related. All their ads went to this same audience. This meant that they had no idea which of those interests actually worked best.
7. Not testing a content-first funnel
A lot of B2B clients are guilty of this. They go straight for a consult call, rather than using content at the ‘Top Of Funnel’ to build a relationship first.
8. Short-term thinking
Testing is the slow, boring route to guaranteed success. I had an e-commerce client for a good few months, but every time we started making any progress in terms of optimisation, he would want to throw a big sale, or try a whole new product range. It killed all the progress we were making. Don’t rely on discounts, and don’t expect to have it all ticking perfectly after 2 weeks.
9. Not turning off the worst performers
This is a weirdly common thing, and a driving instructor training business I worked with as very guilty of this. They would set up a bunch of different ad sets with different variables, but then they would leave them all running for the duration of the campaign. This defeats the whole purpose.
Tips for testing FB ads
When you are starting from scratch, I recommend aiming for the ‘best practice’ and running an initial 2v2v2 creative test. This is 8 ad sets, with 2 variations of headline, ad copy, and visual being tested.
Set them all up, then start killing off the worst performers one by one after an initial 48 hour period.
When you’ve done that, you have a winner, and can continue to test as below.
Here’s an example of how the variations could be set up:
- Headline 1 – Immediate benefit
- Headline 2 – Deeper benefit
- Copy 1 – Short – Who, What, Why, CTA
- Copy 2 – Long – Add testimonial.
- Visual 1 – Image of the product/service
- Visual 2 – Image of your target audience
Choose a winner, then try one tweak at a time. Test the things that make the biggest impact first.
- video VS image VS slideshow VS carousel
- image type
- pattern interrupt
- text on image
- The offer
- The headline
- question vs feature vs benefit
- short vs long
- positive vs negative
- Body copy
- long vs short
- with/without testimonial
- with/without scarcity
- with/without CTA
- different Interests
- different custom audiences
- website VS page likers VS engagers
- different lookalikes
- based on different custom audiences
- different %’s
- A landing page VS a lead form
- Different objectives (traffic vs conversion vs engagement)
Always test at the ad set level.
Know what metrics to focus on:
- Primary objective – i.e. cost per sale/lead/sign up
- Then look at CTR & CPM, if your lead cost is too high to get good data
Give the ads 48 hours first before making ANY changes (unless you spot an error, or it is massively clear that one variable is utterly shit).
If you have a short timeframe, start with lots of variables at a high budget, and remove them in quick succession. For example, if you have a £400 budget, and only a week to run the ads, start with 8 ad sets, with each one set to £20 a day.
This sounds stupid, because £20 a day X 8 ad sets X 7 days = £1120. But what you must do is wait for an initial 36/48 hours, then aggressively kill off each worst performer, leaving you with just the best one or two pretty early on.