Premium inventory is a term used by publishers and networks to generate higher demand and higher fees for advertising on portions of their Web properties.
Often, premium inventory is defined as the pages that see the most traffic. Sometimes it is the degree to which the advertiser can customize the advertising space for their needs, or the specific content designed attract a specific demographic. There are other benefits, but this inventory is usually designated as premium using simple attributes such as these.
Paying more for banner placement on a page that receives a higher amount of traffic or that attracts a particular demographic, makes complete sense if you know nothing else about the likes, dislikes, or intent of the audience.
However, in recent years there have been major advancements in acquiring audience data — data that, when combined with real time bidding (RTB), allows the marketer to define the banner placements’ value in real time based on data about the consumer.
One of the most common examples is retargeting. For instance, a consumer places three Christmas gift items in their shopping cart, but decides to leave the website without completing the purchase.
Two hours later, they may be reading a four-year-old article on an obscure page of some online publisher. At that moment, that potential customer is worth a great deal to the abandoned retailer, which is willing to pay a premium to get its message seen. In this instance, the person, not the page, defines premium.
So Much Data, So Much Potential
Without sacrificing individual privacy, the depth of information that is known about consumer browsing habits today is significant. Common data elements collected for use in display campaigns include: the device, the operating system, the browser, IP address, sites visited, keywords recently searched, subject matter viewed and more.
With so much data at our fingertips, one would predict display campaigns focused on new audiences would perform as well or better than retargeting campaigns focused on existing site visitors. After all, there is now a wealth of known data to support the decision to buy an ad and to influence the content contained in the message or offer.
In addition, publishers would benefit from the fact that the audiences visiting their websites have increased in worth as their history prior to visiting increases the value of that consumer when on their website — any page of their website.
Unfortunately, the vast majority of new-audience-focused display campaigns through networks, first-generation demand side platforms and behavioral targeting companies have failed to fully capitalize on this massive wave of audience data.
What Went Wrong?
The truth is that most of the companies built in the last seven years are ill prepared for the era of big data. This means that legacy technology incapable of accepting raw data is instead manipulating valuable individual audience attributes into a format that decreases the value and effectiveness of the data tenfold. This parlor trick is known as custom audience segments.
Allow me to explain. First you take the raw data: individual unique IDs or cookies that tell you specifically what keywords a user searched on, a website they visited, the subject matter they read, the device type they are on, etc.
These individual data attributes are then used to opt the individual user into a custom audience segment. The campaign is launched and audience segment optimization begins: blacklisting low-performing domains, day parting, adjusting the bid, etc. The key problem is that campaign performance data and the subsequent optimization and bid decisions are based either on post-impression data such as domain served and creative location or group-level data such as total impressions, clicks, conversions, performance by hour, etc.
What Is Missing?
How about the knowledge of what specific data element was associated with the specific user who was served a specific impression?
This is the core attribute used to opt the user into the segment. For 99% of ad tech companies, this information is lost. They can’t tell you what data triggered the impression; they can’t tell you if the data element is five minutes old or twenty-eight days old.
Even if they could identify it, their technology groups all of the users into a fixed segment, forcing bid and optimization changes to be spread across the entire group. Today, enormous dollars are poured into campaigns built around custom audience segments stuffed with users loosely associated with the campaign goals to generate more dollars and increase the odds of hitting some conversions.
The latest vogue marketing spin is algorithmic black magic that predicts and rates the value of impressions or audience segments so that the bid can be reduced for less desirable impressions and increased for those more likely to convert.
The reality is that this is simply programmatic buying at the group level that still can’t identify, bid and report at the data-attribute level. Most savvy marketers I know are not interested in buying low-performing impressions even if offered at a discount.
Unstructured Data Is A Game Changer
Today, there is newer technology in the demand-side platform (DSP) and data-management platform (DMP) space that has cracked the code on leveraging individual audience attributes for buying, optimizing and reporting on display campaigns. Client first party data, offline customer relationship management (CRM) data, and 3rd party data are accessed in their raw form — unstructured.
Every individual prospect being served an ad in real time is a single, stand-alone segment. The specific data that triggered the impression is revealed along with the age of that data — known as recency.
Once identified, real-time bidding decisions can be made around the performance of originating data elements in addition to traditional optimization metrics.
Example Scenario: Suppose your performance reports indicated that 5,000 impressions have been served on a specific website and so far, there have only been ten clicks and two conversions.
Traditional Segment-Based Optimization: With no insight into the data behind the specific users who received an ad, the algorithms are forced to look at the performance of the entire segment at the lowest point available post-impression, which is the domain. The data suggest that the click-through rate is .002%, and the most common optimization decision would be to blacklist the domain.
Unstructured Data-Driven Optimization: With the ability to see what data was behind the decision to fire the impression, there is a new level of insight and control that might lead to a completely different optimization decision.
The data may tell you that of the ten clicks and two conversions on this domain, the consumers behind five of the clicks and one of the conversions had searched on specific keywords in common within 15 minutes of coming to that website.
Rather than blacklist the site, perhaps the decision is made to restrict bidding for impressions on this domain to users who searched on a specific set of keywords and only if the search has occurred in the last 15 minutes. A domain that was thrown away previously may turn out to be a high converting website.
Traditional Segment-Based Optimization: You could be a Caribbean cruise line which has been sold a custom segment for audiences likely seeking new vacation travel. In the black box segment, your vendor is opting-in users who are on pages about travel and cruises. It may or may not be performing well, and therefore, they will continue this tactic, remove this contextual category or increase or lower bid.
Either way, there is very little insight into why the impressions are being served on these sites or why some impressions convert well and why others do not. Even worse, the success or failure may be attributed to something that has very little influence in reality.
Unstructured Data-Driven Optimization: With transparent data, you would absolutely know that your vendor is serving impressions to users on travel sites about cruises. Further, they may find some interesting data elements invisible to you previously.
For instance, perhaps the only impressions that truly garner any success are those pages served on travel sites about cruises that contain specific exact-match keywords on the page such as: specific ports, specific Caribbean islands or perhaps the brand name of competing cruises.
In a world where data is unstructured, the decision may be made to restrict the campaign contextually to pages that contain these exact-match phrases while increasing the bid at the moment the page loads to increase share-of-voice in this moment.
If you truly know who you are targeting, why they are being targeted, and on what device they are being targeted, then you can reach the ideal consumer armed with more data when they land on a popular website, or a rarely visited niche blog.
This is your customer, the data says the influence window is now, and in a world where data isn’t lost inside an inflexible segment, you can make a new bidding decision at the impression-level based on the data elements without applying the bid change to the entire audience group.
Today, the ad tech industry is flooded with DSPs, DMPx and traditional ad networks built on legacy technology frozen in time.
With few exceptions, they share access to common inventory and inflexible audience segments sourced from similar data brokers.
Technology that can manage unstructured data will change the industry for the better. Brands will unlock the full potential of online display advertising through the elimination of traditional segment-based targeting in favor of real-time media buying, optimization and reporting technology fueled by individual data signals.
This evolution empowers display campaigns to scale faster, perform better and provides deeper audience insights than previously thought possible. Unstructured data redefines premium officially as the person and no longer the page!
Opinions expressed in the article are those of the guest author and not necessarily Marketing Land.