You will be able to build audiences from your own customers, site visitors and engaged followers.
Mei Ling noticed something about her best month. Many of the orders came from people who had ordered before, or who had messaged her weeks earlier and not booked, or who had liked her Instagram posts for months before finally asking for a quote. Very few came from total strangers. Her ads, meanwhile, were aimed entirely at strangers.
The people who already know you are often the cheapest to win, and the people who look like your best customers are often the next cheapest. The platforms let you build audiences from both. This lesson covers how, and the rules that keep it useful.
A custom audience is a group you build from data you already hold or from people who already interacted with you, rather than from the platform's interest labels. Meta uses that name. Google and TikTok use their own terms for the same idea, and the names change, so check the current ones in each help centre.
There are four common sources. A customer list is a file of emails or phone numbers that you upload, which the platform hashes and matches against its users. Website visitors are people your pixel or tag recorded, which is one more reason module three came first; you can choose visitors to particular pages, such as a product page or the checkout. App users work the same way if you have an app. Engaged people are those who watched your videos, followed your account, opened a lead form or messaged you on the platform itself.
Mei Ling has strong sources even without a website shop. People who messaged her Instagram account, watched most of her cake videos or followed her page can all be built into audiences on Meta. A customer list is possible too, but only once her collection of contact details is on the right footing, as lesson 3.3, Consent, privacy and the PDPA, explained.
A lookalike audience starts from one of your custom audiences, called the source, and asks the platform to find other people who resemble it. Meta and TikTok both offer lookalikes under that name. Google retired its similar audiences feature and now uses your data more as a signal to guide its automated targeting, so on Google the idea survives in a different form.
The result depends heavily on the source. A lookalike built from everyone who ever visited your website copies people who bounced after two seconds. A lookalike built from people who actually bought, or better, from your repeat customers, copies the behaviour you want more of. Use the best-quality source you have that is large enough for the platform to work with; each platform publishes its own minimum.
Lookalikes are less central than they were. With broad targeting and good tracking, as lesson 4.2, Broad targeting and why platforms now push it, explained, the platform is already doing something similar on its own. A lookalike is still a useful starting point when your conversion data is thin or when you want to test a defined group against broad.
Retargeting means showing ads to people who visited or engaged but did not take the action you wanted. Someone looked at Farah's work dresses, added one to the cart and left. Someone watched most of Priya's video about P5 problem sums and did not book a trial.
These people are much warmer than strangers, so retargeting often brings results at a lower cost. But it works best with a different message from the first ad. They have already seen the first one, and it did not make them act. Something stopped them, and the retargeting ad should try to answer it.
For Farah's cart abandoners, the common worry is fit, so the retargeting ad might show the size chart and her exchange policy. For Priya's video viewers, it might be a parent's quote about the first trial class, or the timetable with evening slots. For Mei Ling's people who messaged but did not book, it might be photos of recent cakes with the order lead time.
Keep retargeting windows sensible. Someone who visited yesterday is more likely to act than someone who visited three months ago. And cap how often people see the ad where the platform allows it, because being followed around the internet by the same ad annoys people and earns the negative feedback that lesson 1.1 showed raises your costs.
The last use of custom audiences is to keep people out. If your campaign aims to win new customers, people who already bought from you should not see it. You would be paying to reach someone you already won, and often showing them a first-order discount they are not entitled to.
Upload your customer list, or build an audience of people who completed a purchase on your site, and exclude it from acquisition campaigns. Exclude people who converted recently from retargeting too. Priya excludes current students' parents from her trial class ads, and runs a separate, small campaign to them about holiday workshops instead.
Exclusions are easy to forget because they do not show up as a cost. They show up as money spent on people who were never going to become new customers.
Look at what you could build today. In the activity below you will list your possible custom audiences and the exclusions you would add to an acquisition campaign.
List the custom audiences you could build today from your own data, and the exclusions you would add to an acquisition campaign.
Junxiong-WFG Organisation is an authorised representative of AIA Financial Advisers Private Limited (Reg. No. 201715016G).