Part 3: the story of websites and advertising
It is 10.20 in the morning. Rafid is looking at a report on the previous day’s campaign. Suddenly Tanvir, at the next desk, says:
— I have been wondering about something for a long time. The advertisements I see at home — do you see the same ones sitting here in the office?
— Not always, Rafid says.
— Why not? We go to the same websites.
Rafid turns his chair round.
— The website may be the same, but at the moment an advertisement is shown, the situation for two users may not be the same.
Tanvir’s curiosity grows. Rafid opens the settings of a campaign.
— Suppose a company has brought a new motorcycle to market. They want the advertisement shown mainly to people aged 25 to 40 in particular areas. The question now is: will everyone who enters the website be shown the same advertisement?
— I don’t think so.
— Exactly. This is where targeting comes in.
What lies behind targeting
Various kinds of permitted signals may be used to understand a user in digital advertising: the type of device, geographical information, the context of the web page, language, time, or the audience signals set for the campaign.
— So does the advertiser know every person’s private information? Tanvir asks.
Rafid shakes his head.
— It is not that simple. There are specific rules on personal data, consent, privacy and which data may be used in what way. Not every kind of data can be collected or used as one likes.
Rafid gives another example.
— Suppose you are reading a cooking recipe while I am reading football news. Even on the same website, the context of our pages is different. If the advertiser takes context into account, different advertisements may appear in front of us.
— So what I am reading can also play a part in choosing the advertisement?
— In some cases, yes. You can think of that as an example of contextual targeting.
As they talk, Rafid opens the settings of a campaign. There the audience has been narrowed very tightly according to the advertiser’s requirements.
— That’s good, isn’t it? Tanvir says. The more specific the better, surely?
Rafid laughs.
— Not always.
If targeting is too narrow, the size of the desired audience can fall so far that the campaign cannot deliver properly. If targeting is too broad, on the other hand, the advertisement can reach people to whom it is not particularly relevant. A balance is needed here.
The frequency of advertising
— The advertiser’s real question should be: whom do I want to reach? Rafid says. And the AdOps question is: how reliably can that audience actually be found?
But there is another matter here — frequency. The same person is shown the advertisement once. Then again. Then again.
— Showing a user the advertisement over and over will not always produce a good result, Rafid says. Excessive exposure can create irritation instead.
Controls such as a frequency cap are therefore used in many campaigns, to set a limit on how many times a user can see an advertisement in a given period.
— So showing an advertisement is a calculation too? Tanvir says.
— Absolutely. Whom to show it to, how many times to show it, in what environment to show it — all of it is calculation.
Just then a new brief arrives in Rafid’s mail. The campaign starts tomorrow.
— Now we have to see whether there is actually enough inventory for the audience the advertiser wants to reach, he says.
— So the job isn’t finished once the targeting is set? Tanvir says.
Rafid smiles.
The user as a golden deer
— Targeting is choosing the door. But you also have to see whether the people you want are really on the other side of it.
The settings of the new campaign open on the computer screen. A thousand people will enter the same website. Some will read sports news. Some the economy. Some entertainment. And every advertiser’s aim will be to find the person who is relevant to them.
The user may see only a banner. But behind that banner an invisible web of audience, context, data, rules and decisions is at work.
— The real question now, Rafid says, is whether the person the advertisement reached saw it, whether they clicked it, and whether they finally did anything. To know that, we have to start up another machine.
— Which machine? Tanvir asks.
— Tracking and measurement, Rafid says.
(To be continued)