Analytics completely revolutionized digital marketing. For the first time, it became possible to quantify the value of clicks, leads, and advertising without resorting to archaic techniques that print advertisers used, like coupon codes or landing pages, specifically designed for different campaigns. It saved time, energy, and money to be able to track everything in one piece of software, but more than that, it showed what pages users visit most, what pages contribute the most conversions, and helped quantify traffic statistics from multiple traffic streams: organic, social, and campaigns. This type of breakdown of statistics let marketing professionals strategically spend more money on channels with the highest profit, making more money for businesses in the processes.
The fault of analytics lies in the fact that it only tracks activity in one session on one site, so any kind of engagement or branding that happens outside of that session where the conversion happen is discounted. Intuitively, we know that building customer engagement is key to conversions. Often, customers visit the site multiple times before making a purchase. Unfortunately, once a user visits and leaves without converting, under the current Analytics modeling method, their visit is valued at $0. Once they come back and make a purchase, that traffic receives the full value of the conversion, say $1000. This type of tracking only makes sense due to the technological limitations in the last generation of tracking. If it’s the best we can do, it’s better than nothing, but it holds no water in a theoretical discussion of how tracking should actually work.
This type of tracking creates an issue for advertisers who rely on data to build campaigns. Consider banner ads that are cookie-based, for instance, so that they follow the user around the Web and display on many websites for the user every day. If a user finds the site through organic search, browses for a while, leaves, and sees a dozen banner ads reminding them of the product offerings over the next week, then goes directly to the site to make a purchase, the direct visit will receive 100% of the conversion value. The banner ads will be seen as worthless, and the organic hit will display as a user seeking information but not interested in purchasing. This is simply not an accurate method of modeling traffic value.
Attribution modeling, on the other hand, assigns a value to each of these actions. A user who finds the site organically and goes on to make a conversion gives 100% of the conversion to organic, split equally between the pages the user visited. A user who is a regular visitor, signing up for the email list, engaging via social media, and viewing banner ads, on the other hand, assigns a small value for every page the user visits, some for every email they open, some for the social channel, and a percent or two for banner advertising, which encourages conversions. It’s a more holistic way of looking at advertising and the way your brand interacts with and retains customers.
Google Analytics doesn’t offer easy-to-use attribution modeling by default, but it is included in their Google Analytics Premium package which retails (currently) for $150,000 annually. Other companies, like the emerging startup saletrail.com, which just received $200K in venture capital funding, promise to bring attribution modeling to users at a fraction of that cost.
Fortunately, we can get very close to true attribution modeling using Google Analytics, so long as we configure our sales and goal funnels accurately and are willing to put in a good deal of work into making a system not designed for attribution modeling support it.
Setting Up Google Analytics To Use Attribution Modeling
The easiest way to start getting closer to accurate attribution is embracing the idea that the most important connection you make with the customers is the first one, not the last one. This is relatively simple to do. Google supports a “utm_nooverride” tag in links, so by adding a simple parameter, “&utm_nooverride=1″, to the end of your PPC and Display Ad click URLs, the conversion value will pass to the method the individual used first, like organic search, rather than their last click.
Delving deeper into attribution modeling, setting up values for multi-touch attribution modeling, we can actually give “assists” to touch-points the user follows along the path to conversion. For example, a common path might be organic via a long tail keyword, organic via a branded search, and then a direct visit.
To do this, we can use the Google Analytics superSetVar user-defined variable functionality to create a custom variable that follows the user as they interact with the site. Into this variable, we’ll stuff every user interaction that the individual has with a site. Every page they visit. How they found the site initially and on repeat visits.
attribution = 2374082340.011.012.013.014.015
Where the first number is the user’s unique visitor ID, which is cookied into the user’s browser by Google Analytics and is persistent between sessions and visits, the second number is a description of how they initially found the site. For example, if they found the site via PPC, which is designated 011, we will append 011. If they found it through organic search, which is designated 012, the URL will begin:
attribution = 2374082340.012… and so on.
Once this variable is defined and is picking up user identification and interaction information on every visit, analyzing it is as simple as opening up your traffic report in Excel and setting up a Pivot Table to analyze the data value. Through Excel, you can analyze which types of traffic are most likely to lead to conversions, calculate what percentage of users who visit via a certain method are likely to convert, and create an average goal value for each hit for type of traffic.
While this is still not a perfect way to model multi-touch attribution, it’s far cheaper than $150,000 a year and it does provide more data about what paths a user takes on your site before converting, so you can get a better understanding of the value of different conversion paths.
Image – © maotun – Fotolia.com