Chapter 7: Ads Strategy
After running your wide target and detailed suggestion ASC’s for a bit, you should have built enough data to start using custom and lookalike audiences. It’s well known that Purchase event lookalikes are by far the most powerful, but it will also take you a while to build enough purchase data to effectively run these campaigns. I mentioned that you want to get at least 1000 data points for LLA’s to work properly - and the more data you have, the better (although it’s worth starting Purchase LLA’s when you have 100-500 purchases tracked). There are however some other LLA options with less barriers to building data that you can start using sooner in order to point the Advantage+ AI in a better direction.
As I mentioned when going over custom audiences, the Top 25% Time Spent on Website customer group and 95% Video Viewed group tend to be pretty solid markers for people that are actually interested in your products. If they’re watching your video ads all the way through, or spending more time browsing your website than 75% of people, chances are they like what they see. For this reason, these are the first custom audiences and LLA’s that we should create.
At this point, you should also have identified what your top performing creative angles are. From this point on, we’re going to focus in on your top 2 (or 3), and try to create more ads that are similar to the best performers. Ideally, you should be rolling out ~5 new ad variations per week that are similar to your best performers. This can be done in one ad using the Flexible format, and can be duplicated into each of your ad sets, or as multiple new ads (or a combination of both).
Let’s take a look at what your first data-based scale campaigns should look like:
ASC Campaign ($500+/day budget)
5x Ad Set Targeting 0-5% Top 25% TS LLA (1% per ad set)
2-3x Best Performing Ads so far (each ad has flexible creative/copy)
1x New Creative Testing Ad (test variations with flexible format)
5x Ad Set Targeting 0-5% 95% VV LLA (1% per ad set)
2-3x Best Performing Ads so far (each ad has flexible creative/copy)
1x New Creative Testing Ad (test variations with flexible format)
Note: Each ad set has all of the ads
You don’t have to split up the 0-5% LLA into 5 separate audiences and ad sets if you don’t want to. You can group 1-3% in one audience and ad set and roll with it, or only do the top 2-3% as individual ad sets so you end up with only 4-6 instead of 10 ad sets in this campaign. It’s really up to you.
As you build more purchase data, we want to create a new ASC campaign with basically the same structure, except now we are using the Purchase LLA:
ASC Campaign ($500+/day budget)
5x Ad Set Targeting 0-5% Purchase LLA (1% per ad set)
2-3x Best Performing Ads so far (each ad has flexible creative/copy)
1x New Creative Testing Ad (test variations with flexible format)
At this point, it’s all about optimizing and increasing budgets. Start in the ads tab - if any ads have spent more than 2-3x your average order value on your site, or have been running for 4-5 days but are performing lower than your break even ROAS, turn them off. Once you’ve filtered the ads, take a look at the ad sets tab, and do the same. Don’t do this at the same time however - if you just turned off some unprofitable ads, give the ad set another 4-5 days to try spending on the more profitable ads it has, and then decide whether to turn off the entire ad set or not. Clearly, if none of the ads in a given ad set are performing, then turn off the whole ad set.
Moving forward, this is basically the whole process. Constantly introduce new creatives to your ad sets via a testing ad, and then rebalance and turn off any ads you have that are unprofitable. If you’re lucky, you’ll lock into a few winners and be able to ramp up the budget, basically just printing money. More likely, you’ll have to continuously be testing new ads and finding new winners, since sometimes good ads can die out of nowhere. Changes to the Meta ads algorithm can also break ads that have been performing good for months, so keep an eye out for sudden drops in performance, and always, always, always keep testing. You don’t want to be fully reliant on one good ad that’s printing like crazy, because it could decide to break the next day. Always keep testing.