ITBusiness.ca

What’s the most valuable information to get from a cookie?

Cookie Bite

Image courtesy of Shutterstock.com

Does advertising really work? It’s the question that marketers couldn’t clearly answer for a long time, but the wonders of online measurement and tracking have finally delivered answers.

In the days of traditional advertising – you know, broad channel, one-directional, and without data on performance aside from the odd survey – marketers didn’t know exactly what ads they put out into the market were working, only they needed to be out in the market to win attention away from competitors. With the wonders of web advertising, marketers not only know which of their ads are delivering results, but they can get a cost analysis of exactly what investment is required to make a sale on any given channel. Further, advertising networks that tap user information and deliver ads programmatically based on demographic and behavioural information can further tweak the dial on precise information, so marketers know not only what ads are most effective, but who is more receptive to them.

In a research paper released in March by Carnegie Mellon University researchers Arslan Aziz and Rahul Telang, a different and more modern-appropriate question is asked about advertising. Rather than “does advertising work?” the researchers are asking “What is a cookie worth?”

It’s a slightly more complicated question and to go about it, the researchers dig into the data of 30 million bid requests for 586,909 individuals, provided by a digital advertising firm that has more than 200 clients and operates in 20 countries. The researchers separated the information that was tracked in the cookies stored with the individuals into six different categories, on a spectrum from least detailed information to most detailed information.

  1. Web demographics such as browser, operating system.
  2. Website details such as if the visitor is returning to the website in a certain time period, if they made a purchase or added a product to a shopping cart, or if they logged in or peformed a site search.
  3. Product-level details, such as how many products a user looked at, added to shopping cart, for each session over the last 15 days.
  4. Temporal website details like how much time the user spent on the site and the amount of time since the last visit.
  5. Temporal product level details like how long a user spends on each page, including product pages.
  6. Product and category frequency such as the number of times the most-often visited and most-recently visited product categories were visited by a user.

Some serious multivariate regression analysis ensues on the data set and after data crunching, the researchers put forward how each level of data affects the impact an ad could have on purchase intent.

Level one information is used as the baseline, with researchers finding that ads randomly shown to a random sample of 6,427 people would result in 59.91 sales attributable to the ad. Adding in levels two and three information, the ads become more effective, bumping up to 60.75 sales.

The big jump comes with level four, where the behavioural information about how long a user spent on the site and the time since their last visit is known. That takes the number of attributable sales all the way up to 78.09.

Adding other behavioural information helped, but not as dramatically. With level six information, sales reached 78.18.

Marketers that are wondering just how much private information they should be collecting about their targeted audience now have something to work with. The bottom line is that site-level time spent has a big impact, while page-level time spent has much less impact.

If you’re trying to hone an effective cookie-based ad targeting program while balancing user privacy, this study offers a good starting point for quantifying where to draw the line.

Exit mobile version