How To Make Unstructured Data Actionable In Display
I wrote recently about the new era of display marketing, and how real time bidding and other advances are bringing the precision and performance of search advertising to display. Well…a new era of unstructured data is also upon us.
A widely cited 2011 IDC study found that the amount of data in the world is now doubling every 2 years. This trend is especially apparent in marketing, as massive amounts of data are being created by ever-increasing numbers of search queries, SKU views, social signals, page visits, and more.
Unstructured Data Volumes Exploding
Traditional paradigms for managing data are struggling under this tsunami. In addition to the sheer volume of data, the fact that so much of it is “unstructured” creates special challenges.
AdExchanger.com recently quoted John Iwata of IBM as stating that 80% this data is “unstructured”. This means that it doesn’t come pre-packaged in neat segments, fitted into a cascading taxonomy of some sort.
Instead, the data is hugely varied and constantly evolving. As an example consider search, where new terms and groups of terms are continually being formed as new artists, politicians, products, companies, and other evolutions take place.
The Old Way: Creating Static Segments
A traditional way of dealing with large amounts of unstructured data is to…you guessed it…add structure.
This typically involves spending lots of time analyzing data and then grouping various data elements (e.g., search terms, SKUs, pages visited, etc.) into segments that are then used for analysis, targeting, and other marketing activities.
This approach has drawbacks, however. For starters, it is time consuming and often requires the attention of experts in the data type. In addition, creating opaque segments often reduces the effectiveness of the data. This is because once data elements are grouped into a segment, the individual data elements within the segment are treated as if they are the same.
Search Marketers Know Better
Search marketers would not group dozens, hundreds, or thousands of keywords into a segment, and then be content with not knowing the individual impression volumes, click through rates, cost per clicks, and conversion rates on each keyword.
Search marketers know that keeping data at its elemental level (in the case of search, at the keyword level) is necessary to optimize and achieve the best performance.
Search marketers know from experience that some data elements (again, keywords) are far more effective than others, and they want the ability to pay more for the keywords that perform well and less for those that do not.
We also see this in search retargeting, where performance often varies greatly between similar keywords that would seem to belong in the same segment.
Elementary, My Dear Marketers
So how can display marketers take advantage of the rising tide of unstructured data, and leverage what search marketers already know?
The solution is elementary. Targeting at the data element level not only reduces the time and effort required to create and populate segments, it also provides improved performance, and deeper insights.
The improved performance comes from the ability to allocate budget to the best performing data elements. This can be done either manually or automatically.
In fact, the automated algorithms that make decisions on how much to bid on each impression are more effective when they are fed more granular, element-level data.
Element level targeting provides deeper insights by enabling marketers to analyze the performance of each keyword, SKU, page, or other data element targeted. From these insights marketers can develop future creatives, offers, and campaign criteria.
Making Unstructured Data Actionable In Display
Unstructured data is already actionable in search, which provides keyword level bidding, optimization, and reporting.
For marketers looking to take advantage of unstructured data in display, the latest generation of Demand Side Platforms (DSPs) provide several options that enable managing, bidding, and optimizing to data at the element level. These include:
- Keyword Level Search Retargeting – Keywords are not grouped into segments, but instead targeted at the individual keyword level. Campaigns can target over 100K individual keywords and maintain bidding, reporting, and optimizing at the keyword level.
- Keyword Based Contextual Targeting – Instead of targeting ads to pages about fixed contextual categories, custom contextual categories are defined by a list of keywords. Bidding, reporting, and optimizing is then done at the keyword (element) level based on which pages contain which word.
- Element Level Site Retargeting – Instead of grouping visitors to a site into just a few segments, ads are targeted based on the individual incoming search terms, individual pages visited, individual SKUs viewed, and/or the products that have been put into shopping carts.
- Element Level Behavioral Targeting – Campaigns target a set of behaviors while maintaining transparency into the volume, pricing, and performance of each individual behavior and/or site where the behavior is measured. This enables more spend to be allocated to the best performing behaviors.
This list will surely grow as the display advertising eco-system evolves. In the meantime, online advertisers can look forward to a day when opaque data segments are nothing but memories of a bygone era.
Opinions expressed in this article are those of the guest author and not necessarily Search Engine Land. Staff authors are listed here.