This week’s guest blog post is an abstract from the original published in The Data Source by Audrey Thompson, Director, Data Science, Oracle Data Cloud and Christa Hammond, Data Scientist, Oracle Data Cloud.
Audience matters in advertising: even the most clever, compelling and timely placed dog food ad is wasted if shown to a person who doesn’t have any pets and has never purchased pet food.
Money spent showing this person creative is money wasted by the advertiser because there is very little chance the ad will influence their behavior.
In marketing, targeting helps by identifying the right audience for a given campaign objective — those who are likely to buy, or buy more, because of advertising.
By knowing what people buy, what they like to do and what they see online we can create purchase, interest and demographic-based audiences.
However, in digital advertising, even the most perfect offline-based audience can be ineffective if you aren’t able to find the users online.
So, how can we find these online users? How do we know the devices we are targeting are actually linked to the users we care about?
And what difference does it make if those links are incorrect?
In order to reach an offline-identified audience in an online space, an identity (ID) graph is used.
An ID graph, like The Oracle ID Graph™ from Oracle Data Cloud, enables an audience from one ID space to be translated or converted into another ID space.
ID spaces include individuals, cookies (unique identifiers for a web-browsing session), email addresses and mobile ad IDs (unique identifiers for mobile devices).
By traversing connections within the ID graph, for example, you can figure out which mobile ad IDs belong to an audience of individual dog owners then serve them advertising as part of a campaign.
When a mobile campaign is executed with a target audience that has been converted to the mobile ad ID space via an ID graph, the key assumption is that these links are correct.
That is, the devices we are targeting are, in fact, 100 percent accurately associated with the users we want to reach. Unfortunately, this assumption is often wrong.
The reality is that all links within an ID graph are not created equal and no link can be assumed to be 100 percent correct.
Even the so-called “deterministic” link of a mobile ad ID to an email address can prove incorrect.
Sure, a given mobile ad ID and email address may be observed during an event (an observed “fact”), but how do you know the email address entered is for the owner of the phone?
What if they logged into an application with their friend’s email address?
Or what if they made up the email address? “Deterministic” cannot be assumed to be correct all of the time.
This means that all linkages are “probabilistic” and while some connections are highly likely to be correct, others are lower quality and will be wrong more often.
So, what happens when an ID graph compromises on quality?
What difference does it make to a campaign’s success?