A Peek into the Ad Measurement Crystal Ball
What academia is thinking about online measurement.
What academia is thinking about online measurement.
While browsing blogs last week, I came across a post on Gary Price’s priceless site Resource Shelf. Gary found a list of papers that had been submitted for an upcoming workshop called “Data Mining and Audience Intelligence for Advertising.” I’ve never heard of this before, but evidently this year will be the 13th edition of the workshop. The point is to focus heavy-lifting data researchers on problems posed and opportunities presented by the enormous amount of data generated by advertising.
I’ve written here about how the accessibility and clever manipulation of data will have a bigger impact on advertising than anything we come up with on the front end. This conference seems to be deeply focused on the topic. While the actual content of the papers to be presented hasn’t yet been posted, the titles themselves seem to suggest a few areas are being explored.
So let’s take a look at this list and see if there’s anything we can guess at or try to predict, based on what appears to be happening down in the data-mining labs.
Contextual for Broadcast
One paper is rather clumsily titled “Finding Keyword from Online Broadcasting for Targeted Advertising.” The critical words here are “keyword,” “broadcasting,” and “targeted.”
Online video has been amazing. ClickZ’s conference (speaking of conferences) on the topic has been examining video’s growing influence, including events that certainly would qualify as “broadcasting.” Consider the Live Earth concert that took place in July. According the site, 10 million people watched the event, live online. OK, 90 million people watched the Super Bowl and 11 million people watched the premier of the “The Office” a few years ago. But a single event that size, streamed live, is clearly a major broadcast.
But how do you sell/place ads alongside a live event? These researchers are on the problem, it seems. A very fast speech recognition system may be able to capture the stream of words flowing out of a live event and re-order commercials in an upcoming pod based on the content of the show. This sort of fluidity could be extremely valuable for an online broadcast that may have a user-controlled component. If the audience votes for an ending where the couple fall in love, show the ad for chocolates. If the audience votes to have them breakup…well, show the chocolates anyway. But change the message a bit.
Another paper is titled “Sentiment Classification with Interpolated Information Diffusion Kernels.”
Sounds good and academic. Let’s just pull three words out here: “Sentiment,” “Classification” and “Interpolated.”
The challenge presented by user-generated media has always been trying to sort it all out. That is, you can find tons and tons of the stuff for any given topic (or brand), but how do you decide what’s good for your brand and what’s bad? Then, how do you come to any strategic decisions, based on this big pile of unsorted content?
Sentiment, analysis, and classification are the goals here. Companies like Buzzmetrics have become experts at it, dedicating lots of really clever people and resources toward the challenge. But there’s always more to do, just as search algorithms can always get better.
As researchers get better at understanding how information is created and shared among consumers, the better their chances at turning all this content into real, usable data. “Interpolated” (in a mathematical sense) means to create a new number in a set, based on the other numbers that are around it. Who knows what “Information Diffusion Kernels” even are. But assumedly, if you know enough about a set of kernels, you can fill in a pattern, and that pattern can reveal how a group of consumers in conversation feel about a brand or a category.
This is pure gold to an advertiser either seeking to understand his consumers or trying to find a good opportunity to present them with an offer.
Two papers that bear calling out: “Extracting Opinion Topics for Chinese Opinions using Dependence Grammar” and “From TV to Online Advertising: Recent Experiences from the Spanish Media.”
The cross-cultural mandate has been in effect online for a good long time now. We’ve been hearing about both opportunities in the titles of these papers: Chinese and Spanish. It’s worth noting as well that the titles focus on the language and the culture, not the places. In other words, they’re not necessarily talking about China, Spain, or Latin America.
The great, rich opportunity is often right in the country you lve in. Multicultural populations are moving online at extremely rapid rates. We have a tendency to think of the World Wide Web as an English-speaking medium, but that’s far from true. Savvy marketers are seeking to find ways to take the technologies that are making advertising more effective and looking for ways to leverage them with consumers who communicate in a language other than English.
Following the Message
Finally, there’s “Discovering Information Diffusion Paths from Blogosphere for Online Advertising.” What a title!
Information diffuses online like dye in a bucket of water. Along the way, the information sometimes changes, often is commented upon, and discussion pathways of a community are uncovered. The ability to follow this path, whether it’s live or after the fact, can be extremely valuable to an advertiser. Consider the case of an advertiser who wants to control a false and damaging rumor. Or on the flip side, he may need to locate the source of a positive product review.
If the paths that information follow can be charted this way, potentially they can also be leveraged. This is really the underlying notion behind much of this exploration: how can we begin to get a grasp on the data flowing around our space and use it to be much better at achieving our goals?
For More Navel Gazing
This was just a sample of the titles submitted to the conference. Again, we don’t even have the content. But when we’re asked to look into the future, sometimes it makes the most sense to look into the present — in unique areas. If you want to continue this thread, I suggest periodic searches on scholar.google.com. Take whatever you were going to search for (something like “advertising to teens”) and type into an entirely different kind of index.
You’ll be surprised by what some people are thinking!
Nominate your choice of technologies, companies, and campaigns that made a positive difference in the online marketing industry in the last decade. Nominations end August 3 at 5:00 pm (EDT).