Facebook and Instagram are hugely popular platforms in Greece, and a large share of the advertising budget of e-shop and of businesses in general is invested in them. We often hear about targeted advertising on Facebook / Instagram and about how we should target an audience based on interests or on demographics. What nobody tells you, though, is that in recent years all of this has been changing, and the most successful advertising now usually runs on AI (artificial intelligence).
How does audience targeting work in Facebook & Instagram ads?
In the past, before the data protection law (GDPR), Facebook knew very well what you wanted to buy and when, even before you wanted it yourself. The platform knew your interests exactly and, based on them, knew your gender, your age and pretty much everything about you. Back then, and still today, a common process was to make 2 or 3 creatives, then create different ad sets based on single interests, see within a month which one performed best, and then test and test again, and so on.
For example, if we had a wooden gift for couples in love (linked here so you can see exactly what the product is), we would create one ad for an audience interested in love, another for people in a relationship, another targeting emotions, and so on. That way we would find which audience works best, and then we would run more tests, e.g. only men with the “winning” interest or only women, and then perhaps narrow down the age range. All this was done to get the best possible result.
What do we do today for better audience targeting?
The process above is a good one that can still work in some cases and may even still be the best. What can go wrong with it, though, is that we might not think of a particular interest. In our example, we might never discover that people who eat junk food are a bigger audience and much more likely to buy from our e-shop. You may ask what junk food has to do with a gift for couples, and our answer is this: the junk food audience could be at the ideal age for our buyer persona, and because these people buy junk food when they want something cheap, most of them could end up choosing our gifts.
Even better, we could build a combination: users who eat junk food, narrowed down to those interested in personalisation or love. Naturally, we can’t test every possible combination, or we may not have the budget to do so. Let’s be clear that at this point we are talking about the “top” of our funnel, and of course awareness campaigns should already have run so that the customer has got to know our business.
Facebook & Instagram ads with AI (Advantage+ ads)
Luckily for us, Meta has created an ad type that uses AI, and it is now up to the platform to find the ideal audience and the ideal buyer persona. All we then have to do is a few simple things:
- Have the right, good-looking creative
- Get the right message across with our creative
- Advertise similar products together
- Send people to the right pages, based on the message of the ad.
- Have at least 4–5 creatives.
Another important thing the AI does is decide the placements where the ad will appear, based on the results we have set as our goal.
What the algorithm of Facebook / Instagram does is run a huge number of A/B tests and, after a period of 1 to 2 months, reach the best possible result.
Something else that matters a lot is that campaigns like these need a big budget. Depending on the audience you are targeting, the daily budget can be anywhere from 60 to 160 euros, so that the algorithm can collect data. Be careful: if we have an e-shop that sells tools and also sells garden furniture, we need two different campaigns, one for each product group. The same goes for websites selling services, for example digital marketing and website development.
Targeting the right audience through your creative on Facebook & Instagram
The audience that Facebook and Instagram will find, and the success of either AI or interest-based targeting, depends mainly on having the right creative, whatever the case. That means your creative fits the job and you have paid real attention to it, for example by investing in a video instead of a simple photo, or by using a photo with high production values that gets the right message across, compared with a poorly chosen one.
Let’s look at an example, using the gift for couples again. Image 1 and image 2 show the same product, but image 2 will be far more successful and bring in many more people (it will grab attention), so Facebook will get better data and target better. And below you can see a video that was made for the same product, which of course works far better than any image!
This applies to more than just the image in your ad. It also applies to the images in your product feed, which should be high-resolution and look good, and to the images on your website. If all of these have the right look and the right message, the customer will buy the service or product more easily, and we will get more and better data. As we have said in an earlier article, a huge role is also played by website speed and the overall look of the site, so that users feel they can trust it.
Advantage+ ad results
Below we will show you an AI-driven ad case so you can see how powerful it is. To repeat: nothing works out of the box, and something that works for one client may not work for another. What we describe above comes from thousands of tests on our clients’ accounts and hundreds of thousands in ad spend.
As we have said before, numbers may not tell the whole truth: even with results like these (ROAS), an ad might not be profitable, while another ad with a cost per result of 100 euros could be. So we need to learn to read what lies behind the numbers.
What kind of ads should I run on Facebook / Instagram?
If you have read the whole article, you will have understood that Advantage+ is a very good option when the essentials are in place (budget, creative, website), and that interest-based ads with case-by-case testing can also work well. Many people will say that interest-based targeting gives you more control, but the question is whether you can really run all the tests a machine would, in the same time and just as accurately.
Leave a comment