There are literally thousands of online publishers seeking to monetize the “unique” traffic visiting their websites. Concurrently, there is a thirsty crop of data brokers lining up to assist publishers in connecting their data with the ever-growing advertising demand online.
We live in an era of increasing concern about privacy, which makes sense, given there is simply more non-personally identifiable information (non-PII) than ever before on virtually every consumer online — regardless of device, age, race or geography.
This non-PII information includes where you browse, what you view, what you search, device used, operating system, browser type and more. In fact, the data used online now includes the ability to leverage the mountains of data collected over time on user behavior offline. Through the miracle of technology, the offline data elements are merged with the consumer’s known online behavior encased in a non-PII user ID.
In short, more and more online display impressions are won and lost in a real time bidding environment that, in theory, rewards the advertiser that selects the vendor with the most data available in the milliseconds of a single ad call. The data determines if you should bid, how much you are willing to pay to win the impression and, frankly, what message to serve up.
So, the company with the most data wins, no contest — game, set and match!
Unfortunately, this is a myth being passed down and repeated as a fact, company after company, sales rep after sales rep.It’s not unlike your mom telling you that Christmas poinsettias are poisonous or that you can’t swim until an hour after you have eaten. Big thanks to Ken Jennings for finally debunking those myths and others in his new book, Because I Said So.
Watch The Marketing And Public Relations Machines Go To Work
“Big Data” is perhaps the most vogue marketing tagline of 2012 and the heading by which legions of companies are anchoring their value. Like most buzzwords, this term is misunderstood on a grand scale. In the world of public relations and marketing, this creates an influenceable moment; and anyone with a pulse and some marketing decision-making power should prepare for the data-themed sales pitches that will flood through their door from online display vendors in 2013:
- “We have more search data than Yahoo and Bing combined”
- “We have a 200 million-plus consumer panel to work with”
- “We have access to site visitor data from ___________ mega site, and no one else has it”
- “Artificial intelligence, big data, machine learning, and decision science”
I personally work for a Demand Side Platform (DSP), Simpli.fi, that is recognized as an authority in search retargeting, and we proudly report that today we see more than 10 billion search events on a per-month basis, including searches performed on Google, Yahoo, Bing and a vast network of publishers.
So does size matter? Of course it does, but in the end, it means very little unless you are able to access the right data for the right consumer at the right time with the right kind of controls… and make the right decision about what ad to serve. The real battle is won with the complex technology that is difficult for PR firms to articulate and even more difficult for the average media buyer to comprehend in our sound-bite-driven world.
Is All That Data Accessible When You Need It Most?
It turns out that data is just data. There are a ton of variables related to the use of data that define just how valuable the information is at the moment when you need to access it. Consider these two scenarios:
You are in a speeding car rounding the curve of a mountainside road when you hit a patch of ice. At that moment your car begins to spin as it drifts toward the edge. At that moment you have some data: you know that you are on a road with ice and that you just hit that ice going too fast, you know your car is spinning, you know you’re headed for the edge of the mountain. Unfortunately, all that data will not prevent a wreck. You may, however, brace for impact!
You are in a speeding car rounding the curve of a mountainside road when you encounter a series of signs warning of severe and dangerous ice ahead. With full knowledge of this data you proceed to reduce your speed to a safe level and manage to avoid a wreck completely.
In both instances, you received data regarding speed, ice and location. The only difference is that in one scenario the timing of the data and the ability to control the associated variables resulted in a completely different (and for most of us, a preferable) outcome.
My friends, I can tell you that your data-driven display campaigns can have two completely different outcomes from two vendors accessing the same data serving across the same inventory with the same creative. The secret is in how each company has learned to use the data, big or small.
To prove my point, I will stick with the example of search retargeting, aka keyword-driven display, or intent display. The data element in this case is the knowledge that an individual has performed a search for a specific keyword, thus sending up a flare that they intend to take some action online.
Where The Real Battle Is Won — User Match Rates
Once your display vendor has obtained information that an online consumer has performed a search, they store this data through one of several means. The data could be stored in a cookie or some other unique ID stored on a server in a cloud.
The display inventory sources, primarily exchanges, have their own separate ID that identifies a user as they browse through websites in their publisher network. As most retargeting companies and demand side platforms are accessing multiple exchanges, it turns out that each exchange ID is unique to their specific exchange, and the rate at which they update and purge their data on consumers varies from source to source.
In search retargeting, the number of keywords you have access to matters. The sources of the keywords you have access to matter. How many queries-per-second your platform can bid on across the exchanges matters. However, none of these often-touted vendor strengths mean very much if, in the moment of an individual bid decision, the vendor fails to connect their internal ID with the exchange ID. The rate at which you can successfully merge these two IDs on an impression-by-impression basis is vital!
In short, two vendors may have the same data on the same users while accessing the same inventory source. However, one vendor may have a match rate of 20% while another has a match rate of 75% or higher. Translation: greater reach and bidding opportunities within the same inventory source is a reality.
Data Element Level Recency
Keeping with the theme of keyword data used in search retargeting, it doesn’t take a rocket scientist to tell you that all keywords are not created equal. One keyword may send a signal that a consumer intends to take some sort of action online immediately. A completely different keyword may indicate that a consumer is going to take action within the next two weeks. Every search marketer in the country just nodded his or her head in collective agreement.
The obvious lesson is that data derives a significant amount of its value from the variable of time. In search retargeting we call this recency.
Suppose you wanted to target people who are searching for “Cheap Airline Tickets.” Wouldn’t it make sense that you would be willing to bid to put an ad in front of a consumer if they performed the search 2 minutes ago? Now the question is, do you feel the same way if you knew the consumer performed the search 30 days ago?
My warning to the average media buyer blinded by the latest big data pitch is that almost every DSP and retargeting company in America with search retargeting on their product menu has no control over this variable.
Sure, they can dynamically customize the creative, but the data used to build the audience pool is placed inside a segment where the data may be fresh or it may be stale.
Data Element Level Optimization
The typical display campaign in today’s ecosystem often extends way beyond keywords that individuals searched for when browsing the Internet. As I noted in the opening of this article, to achieve scale, additional data elements are used to opt a user into a campaign.
For example, if you are a car dealership seeking “auto intenders” then many vendors will build some magic audience pool that includes people who have visited a national auto site, consumers who have searched on “certified used Toyota Phoenix,” and even some who completed a survey offline.
Some of the data is good data and some of the data is poor data. Some of the data is recent and some of the data is old. If the entire audience is managed as a segment with traditional optimization metrics, then all of the data is being treated as if it has the same value.
This is a dramatic devaluation of the full potential of big data. It is vital that any data-driven campaign never loses the ability to report transparently on what data element triggered the impression.
More importantly, can you bid differently based on the performance of each data element with this segment approach? Can you adjust recency, ad frequency, messaging and more at the data element level? Until you can access the big data in an unstructured format free from the shackles of segments, then your success is handicapped.
A Complex Marriage Of Variables
There is no question that a company and a campaign with more data are better equipped than the alternative. However, it is a complex marriage of variables that define just how effective the data truly is.
In an industry where an increasing number of display partners are seeing similar data and similar inventory, it is more important than ever that media buyers seek solutions with transparency and control over the variables of big data that are rarely discussed and often overlooked.
Ask anyone who is truly in the know if size matters, and they will be quick to let you know it’s how you use it that counts.
Opinions expressed in the article are those of the guest author and not necessarily Marketing Land.