Chapter 7: Ads Strategy
Now that we know the technicals that go into setting up campaigns and ads, let’s go over the current best practice strategy for starting from zero. At a high level, the overall strategy is pretty straight-forward and intuitive: Step 1, build data; Step 2, scale with data.
When I say “Build Data” for Step 1, I’m talking about populating your Meta Pixel with events data. Until we have data in our pixel, we can’t utilize any of the higher tier features for scaling and optimization like custom and lookalike audiences. So in the meantime, we’re stuck running ads using detailed targeting of interests and the Advantage+ AI, and trying to drive as many purchases on our website as possible (that the pixel can track).
Before we start running ads, it’s important that we have our creatives in order. I like to have at least 5-10 distinct angles for creatives. This means different messaging, different offers, different styles, etc. Each ad should be saying something different to the viewer. Each of those angles should also have 5-10 different variations, meaning video vs. image, colors, formatting, different video hooks, etc. Luckily, it’s easier than ever to do this with the help of AI, and if done properly, should yield us somewhere between 25 and 100 ad creatives to get started with our testing.
For initial testing, you should plan to spend between $50-100/day minimum, but you can speed up the process by spending more if you have the budget. The goal here in the testing phase is not to be super profitable, but to simply break even. If we can break even with our initial creatives and no data, that means we can surge into profitability with access to custom audiences, LLA’s and refining of our top performing creatives. Hopefully this testing period should only last ~16-24 days.
So what does our campaign structure actually look like? We can get away with doing most of the testing in one campaign to start:
1x Test Campaign
Sales optimization event
Budget set at ad set level to start
1x Broad Target Ad Set
No detailed targeting, let Advantage+ AI do all the work
$20/day budget (minimum)
3x Suggested Interest Group Ad Sets
Add detailed targeting suggestions for buckets of interest groups you think will work well. For example, if I’m selling luxury watches, my 3 suggested interest ad sets targeting would look like:
Rolex, Breitling, Audemars Piguet, Patek Phillippe, Hublot etc. (competitor interests)
Luxury, Luxury Lifestyle, Luxury Real Estate, First Class travel, Luxury goods etc. (lifestyle based interests)
Ferrari, Supercars, Yachts, Lamborghini, Fine dining etc. (secondary interests)
$20/day budget per ad set (minimum)
5-10x Ads (per ad set)
Each ad is focused on one of my different “angles” for my creative
Each ad includes all 5-10 formatting variations for each of my angles (using “Flexible” delivery)
Each ad includes 5 captions and headlines
This structure should ensure adequate coverage for our testing, and will allow us to get feedback on which of our creative angles is performing best (and with each audience). Since the budget is set at the ad set level, you can determine your best performing creatives by which ad gets allocated the most spend (and best results) via the Meta AI (this also holds true for campaign level budget CBO/ASC). Obviously, the more budget you can put towards each of these ad sets, the quicker you are going to get results, and the faster your ads will optimize. If you could do $50/day per ad set, it would be even better. However you don’t need to go any higher than that. If you have the budget, duplicate the entire campaign, but switch the budget to CBO/ASC instead of at the ad set level, and keep everything else the same.
The goal is to let this run for 4-5 days, then take a look at how things are going. Likely, the Meta AI will have picked a couple ads to put most of the budget into, while the others only spent pennies. If the ads with spend are performing well (breakeven or profitable ROAS), then simply turn off the other ads that aren’t getting spend. If an ad IS getting spend and isn’t performing well, turn off those ads, and force the Meta AI to start spending on the other ads you have. If any of your ad sets are performing extremely poorly, go ahead and turn those off too.
Once you have some combination of ad sets and ads that are working well, it’s time to start duplicating into CBO’s/ASC’s. CBO’s refer to the Meta Platform’s “Campaign Budget Optimization” or “Advantage+ Sales Campaign” option. So far, daily spend budgets have been set at the ad set level. However, with CBO/ASC turned on, your daily budget will be determined at the Campaign level. Ad spend will then be allocated to the ad sets in your campaign based on the algorithm’s discretion, with a goal of achieving the best bang for your buck. As we’ve mentioned, the Meta Ads algorithm is pretty damn good, and CBO/ASC’s can yield amazing results.
Typically, CBO campaigns should be utilized when spending >$100/day on a handful of ad sets. Since the budget is coming from the campaign level, you want to make sure you’re giving the algorithm enough ad spend to work with properly. Your first CBO’s should consist of 2-5 of your previously top performing ad sets, with a $100-200/day budget. CBO’s open the door to the upper echelon of Meta Ads, with profitable CBO/ASC campaigns comfortably spending $2,000/day+ each. This level of profitable spending, for all practical purposes, is only possible using CBO’s. Create a new ASC campaign, and duplicate out your best performing ad sets and ads. Since the budget for this campaign is CBO, we can ramp up the spend on our best performers to accelerate our data collection. Hopefully before long, we can start rolling out some data-based campaigns.
If none of your ad sets and ads are performing at breakeven, the problem is likely your creative. Take another look at the Meta Ads Library, and figure out what’s working for your competitors. The idea “if at first you don’t succeed, try try again” is completely applicable to running meta ads. Try as many different creatives with the above structure as possible. I promise something will work eventually.