The delivery of advertising has become ever-more targeted. The kind of ads a consumer sees used to be based on fairly simple criteria: media (print, radio, television); content and demographic information about the audience consuming that content (type of print publication, radio or television program) and geographical location of the audience. Later on, a few more targeting variables came into play like time-of-day “day-parting”), audience segmentation and past purchasing behavior.
Digital Advertising’s Early Days
When the Internet came along, it introduced whole new forms of ad targeting. First of all, the Internet wasn’t just one medium; it was an amalgam of several, each lending themselves to different ways to target:
- Websites (banners/buttons [“display ads”]; sponsorships; paid content; paid links)
- Email (display ads; paid content; paid links)
- Search engines (keyword targeting)
- Computer games (sponsorships)
- Mobile (mobile display ads; sponsorships; apps; text ads; audio ads)
- Social (display ads; sponsorships; paid content; paid links; hybrids)
- Video (video ads; sponsorships; paid content; paid links)
But the Internet enabled targeting far beyond content, which for so long was the basis that drove the advertising planning and buying process. We began to hear about other forms of targeting like…
- IP Address
- Operating System
- By Device
And more recently still, Data Targeting.
Big Data Refines Ad Targeting
In order to understand data targeting, you first need to understand how much of this data gets generated and the diversity of the data sources.
The amount of data we continue to generate is staggering, much of which is related to our digital interactions. A 2013 report issued by the University of Southern California Marshall School of Business predicts that by 2015 the average American will consume 15.5 hours of digital media alone per day. By 2012 this consumption amounted to 6.9 zettabytes (6.9 million million gigabytes) of media data alone generated per year. For perspective, check out the great comprehensive list of big data stats put together by Wikibon.
All the data being collected and computer processing power to store and crunch it makes ad targeting easier than ever. Practically everything that anyone does these days leaves a digital trail, and each of those actions and transactions reveals a little bit about ourselves. Where we are. Where we shop and what we buy. What we like. What we watch. Even if anonymized, these parcels of data or “data packets” are being sliced, diced and sold to companies who use the data for targeting purposes.
Methods of Data-Targeted Advertising
In addition to the advancements made in digital ad targeting I mentioned earlier, data-targeting enables a whole further degree of refinement.
- Audience Targeting – ads specifically served to anonymous users based on their shared behavioral, demographic, geographic and/or technographic attributes (IAB Wiki)
- Predictive Modeling – ad delivery models based on probabilites
- Time-Since Targeting – using duration of time between one or more past actions to predict possible or likely future action
- Lookalike Modeling – building larger audiences from smaller audience segments whereby the lookalike audiences reflect similar characteristics to a benchmark set of characteristics the original audience segment represents
- Actalike Modeling (also known as Presumed Intent Targeting) – using certain past online behavior as a prediction of future intent
- Interest-Based Targeting – targeting users based on a similar set of interests; targeting can be done on a predictive, lookalike or actalike (or a combination thereof) basis
- First-Party Data Targeting – employing direct data from an advertiser’s existing customer and/or prospect databases to help refine targeting
- Personally Identifiable Information (PII) Targeting – Non-anonymized targeting based on accounts where users have already created logins and profiles about themselves that allow ad targeting based on that information
- Social Share Targeting – targeting triggered by what a user specifically shares in social networks or the type of content they have a propensity to share
- Location Targeting & Triangulation – serving ads based on combining geo- and location targeting from device GPS coordinates with known variables like travel plans or anonymous information like changes in location after various types of media exposure
Between proprietary nomenclature and new advancements in data targeting, I’m sure my list will be considered incomplete or inaccurate as soon as it’s published. What I hope this points out, however, is the precise extent to which a consumer can be targeted through the marriage of data and technology.
The Death of Random Circumstance?
This kind of precise ad targeting is great for advertisers – what advertiser wouldn’t want to eliminate wasteful or ineffective ad spending? – but it does make me wonder about the special something that life’s randomness often leads to. Sure, I prefer to be shown ads that are more relevant to me, but what will happen if every ad we take in is a result of something we did, saw or interacted with in the past? Is past behavior alone an indicator of future intent? Will all impulsiveness be squashed because our ad experiences have been pre-determined for us?
I think about all the random events in my life – ad-related and otherwise – that yielded more positive results than I could have ever imagined, and it’s the thought of the loss of this unpredictability that I rue.
Enter the “Happenstance Variable”
How to solve for this problem, for the desire to maintain some chance – and its positive benefits – in modern advertising? The way I see it, since digital ad technology has been built on or utilizes different algorithms to automate and deliver targeted advertising to the masses, why not just add another variable into the algorithmic equation? I’m calling it the “Happenstance Variable.”
The Happenstance Variable can be dialed-up or dialed-down by advertiser and consumer alike. Just as AdChoices was created to give the user more control over what information from a site will be stored, collected and how it will be used, the Happenstance Variable could be modified by the user. Perhaps a particular user doesn’t really like to leave things up to chance – this person would leave their Happenstance Variable setting low, allowing for ad technologies to do their usual ad targeting thing. On the other hand, someone else who enjoys a measure of random chance in her life could dictate the percent of happenstance she prefers. And because the Happenstance Variable could be modified at any time, if the user’s ad experience becomes so random as to once again be nonconstructive or unpleasant, that user could log back into the dashboard and reduce the rate of happenstance.
That’s not to say that how the user reacts to randomly served ads would not be tracked or analyzed any differently than targeted ones. In fact, reactions to random ads could be even more telling than targeted ones. It would give the advertiser even more insight into the psyche of that individual and what makes him tick.
Perhaps, “Reaction to Randomness” becomes yet another piece of data used in predictive modeling and the future targeted delivery of advertising??