Decide when to let the platform find your audience and when to narrow it yourself.
Open the ad set you're running now, or the one you plan to build, and list every targeting choice in it: each interest, the age band, any behaviours or job titles. Expect to remove most of them. Keep only the ones you can give a reason for.
When you pick interests, you're working from interest labels. These are tags the platform gives people based on pages they followed or ads they clicked, sometimes long ago, so they're rough. Many real customers never get the relevant label, and many people who have it never buy.
The platform's delivery system sees how millions of people behave, in close to real time. With conversion tracking, which you set up in module three, it also sees which people actually bought from you. It then looks for others who behave like your buyers, whatever labels they carry. The prediction doing this work is the estimated action rate, which you met in lesson 1.1, "The highest bid does not always win".
A few years ago, telling the platform exactly who you wanted often worked better. Today the platforms argue that, given good data, they usually guess better than you can. Meta, Google and TikTok all now encourage broad targeting, and their automated options lean on it. Meta's audience settings under the Advantage+ name treat your choices as suggestions and let the system go beyond them. Feature names and defaults change, so check them when you build.
Broad targeting means audience settings with only real limits, so the platform finds buyers from behaviour instead of labels. It amplifies whatever you told it to find, so it needs good conversion data. If your tracking is broken or you chose a shallow objective, the system will look for the wrong people very efficiently.
Farah runs an online fashion business selling blouses and office wear, and she ships anywhere in Singapore (an invented example). Her first Meta ad set targeted women in Singapore aged 24 to 34 who were interested in modest fashion, office wear, two particular fashion magazines and LinkedIn. She was proud of how precise it felt. It reached a small audience and cost a lot per sale, and the platform kept warning that the audience was narrow.
A friend who runs ads for a living reviewed it and suggested deleting almost all of it. Because Farah ships anywhere in Singapore, her location limit is the whole country and nothing beyond. She rebuilt the ad set with two limits: Singapore, and women. Her cost per sale fell over the following weeks. She also changed her creative at the same time, so she can't credit the targeting alone. If you change creative and targeting together, you can't tell which one produced the result.
Broad doesn't mean no limits. Some limits are real business constraints, and the platform can't know about them unless you set them. Location is the most obvious one. Priya runs a maths centre in Tampines (an invented example). Parents in Jurong are unlikely to send a child across the island twice a week, so she limits her ads to the east, or to a set distance around the centre.
Language is a real limit when your ad and service are in a language many people in your area don't use. Age is a real limit when the law or the product requires it, or when the product clearly doesn't fit a group, such as a retirement product for people in their twenties. Anything beyond these limits is a preference, and broad targeting asks you to let preferences go.
With broad audience settings, the ad itself largely decides who sees it. The platform shows the ad to a range of people, watches who stops, who clicks and who buys, and then shifts delivery towards people like them. This is creative as targeting: what the ad shows and says steers delivery towards the people who respond to it.
Say an ad opens on a woman getting ready for her first day at an office job, with the words "what to wear in your first week". Mostly women starting a new job watch it and act on it, so the platform finds more of them. A specific ad like this gives the system a signal. A generic ad, with a generic model in a generic blouse, gives it nothing to go on. Module five, on making ads people stop for, matters as much as this module.
If you want to reach two kinds of customer, your strongest lever is often two different ads, each speaking clearly to one of them. Two hand-built audiences shown the same ad usually achieve less than two distinct ads.
A narrow audience means fewer people, fewer auctions to enter and fewer cheap opportunities, as lesson 1.2, "What you pay and why it moves", explains. On a small budget it also means the system sees few conversions and learns slowly. Platforms warn you when an audience is narrow.
There are still good reasons to narrow. Retargeting means showing ads to people who already visited your site. It's narrow by design, and lesson 4.3 covers it. A business with almost no conversion data may get better early results from a sensible starting audience. A test may also need a narrow audience to answer a specific question.
The default has shifted: start broad within your real limits, let good creative and good tracking do the work, and narrow only for a reason you can name. When you build an ad set, work in this order:
Check that conversion tracking works and your objective isn't shallow (module three). Set only real limits: location, language if relevant, and age if the law or product requires it or the product clearly doesn't fit a group. Drop interests that are only preferences, and let the platform go broad. Make creative that speaks clearly to the customer you want, so the ad does the targeting. To reach different types of customer, make a separate ad for each instead of a separate hand-built audience. Narrow only for a reason you can name: retargeting, almost no conversion data, or a specific test.
Go back to the list you made at the start and cross out every item that's a preference. In the activity below you will write the minimum limits your own business genuinely needs, and nothing more.
Write down the few targeting limits your business really needs, such as location or language. Leave out everything else.
Junxiong-WFG Organisation is an authorised representative of AIA Financial Advisers Private Limited (Reg. No. 201715016G).