As I mentioned a while back, I keynoted at Search Insider Summit in Key Largo earlier this year. One of the main threads of that conference was on attribution, which attempts to give the proper credit to marketing tactics.We discuss this at some length in Outside-In Marketing: Using Big Data to Guide Your Content Marketing. The object of this article is to give you a taste of what’s in the book.
Most of you are familiar with the challenge: your boss wants to focus on getting more leads and, based on his reading of the data, the way to do that is to improve conversations at the last step of the buyer journey. You know so-called last touch attribution is not the only way to improve the volume and quality of leads. It is probably not the best way. But how do you prove it?
Until you prove it, your marketing budgets will focus on the bottom of the funnel at the expense of the top of the funnel. The result is predictable: a lack of focus on the experiences that could attract new buyers, and only incremental improvement in leads and wins.
If you can’t prove it at scale, try proving it in a small test case. You can usually get funding to at least try a new way of giving the proper credit to top-of-funnel activities. When you do, your trial can turn the religion of last-touch attribution into the science of an attribution model that works for your business and its digital buyers.
Attribution science
The first step towards attribution science is to decide not to test any preconceived attribution model–last touch, first touch, all touches, etc. All your current models could be inadequate to your buyers and their journeys. Rather than starting with a preconceived model, look at the data you have, and see if you can discover any patterns. Focus on the most successful experiences and try to find common best practices. Then test these practices in your trial. The idea is to let the data guide you in what attribution model works best.
One thing your data might tell you is that you have a measurement problem. You could, for example, only measure the quality of your assets (white papers, case studies, videos demos, etc.) based on the number of quality responses they generate. But, try as you like, you can’t find anything distinctive about the assets that are performing the best. A common reason is your digital journey is complex, and you are not measuring all the variables.
Before you begin your test, make sure you are at least measuring these things:
- Tracking: Cookie every user and measure each users’ entire journey. A common thing that digital marketers miss is the users who enter your website through natural search. Many pixel-based measuring systems only measure visitors who originally arrive from paid media clicks. When you start tracking organic search visitors in the same way, you can see the whole customer journey for all of your potential buyers.One thing you will find is that every direct referral (those that come from direct URL typing or bookmarks) has its origins in another kind of referral, either paid or organic. You can’t attribute anything to direct referrals unless you can track them back to their origin–a visit that precedes the direct load. When you start attributing direct traffic to their ultimate source, it will be a revelation.
- Ungated responses: Buyers don’t download assets they can’t find. Google doesn’t index assets that require a registration. So the only way they find your assets if you put a registration in front of them is through your digital experiences. That severely limits potential buyer interactions with your key assets.The other main problem with requiring registrations is abandonment. Look at the abandonment rates on your registrations. If they are high, it’s because you are not giving enough value in exchange for the effort required to fill out the form. How do potential buyers know they will get equal value? When in doubt, they abandon.If you put your registration in the middle of the asset, the asset can be available to every potential buyer and it can generate more quality responses. For example, if you have an e-book, give the first chapter away and require registration for the rest. When you give your potential buyers a taste of what they will get when they register, you get a much more accurate picture of the value of your assets.
- Site architecture: Besides over gating, the main way sites challenge users is through confusing user journeys. The way to measure to what extent your site provides elegant user journeys is by analyzing the links. Numerous link analysis tools are available that can measure the relevance of a call to action to the experience it sends people, and can replicate this measurement at scale. When you correlate these results to the bounce rates that occur when a user lands on one experience from another, you can get a good sense of the choke points.
If you are confident you can measure user interactions accurately, it’s time for marketing science.
Testing Hypotheses
All science starts with a hypothesis. In the case of digital marketing, your digital experiences can be so complex it is tough to know where to start. This is where looking for patterns in your current user behavior data can help. You might find that certain content types don’t seem to work at all in certain places in the buy cycle. For example, white papers are particularly bad at helping people figure out if a product is right for them. You might find that certain experiences seem to work pretty well for certain stages in the buyer journey. For example, demo videos have a low abandonment rate and a high response rate in the consideration phase.
These discoveries could be starting points of hypotheses. For example, “If every potential buyer could see a demo video when they are considering different products in a category, quality response rates will go up across the board.” That’s a hypothesis. To see whether it is worth doing a proper test, compare the buyer journeys for products that have the demo video and those that don’t. If you see a pattern, start building a test.
Sticking to our hypothesis, we could conduct an A/B test where the same experiences either contain a demo video or don’t. Perhaps in one version, the call to action is a demo video and the other is a white paper on the same topic. If you replicate this test in a representative sample, you have scientific proof that your hypothesis is right or not.
What does this have to do about attribution? The idea is to string these hypotheses together for at least one buyer journey and product family and see what’s working and what isn’t. If you measure everything and build your tests well, the correct attribution model will emerge. Then replicate it for other buyers and products. Not all buyers are alike. And not all products require the same journey steps. So expect some variation.
Device tracking
What I have outlined above is a good approach to attribution if you can track a user on a single device. If your buyers use multiple devices for the same journey, tracking is challenging, to say the least. But we will leave that kind of attribution to another article. For now, you should have enough to start building a scientific attribution model that is tuned to your buyers and your business.
James Mathewson is co-author of Outside-In Marketing: Using Big Data to Guide Your Content Marketing and Audience, Relevance, and Search: Targeting Web Audiences With Relevant Content. He is an IBM Distinguished Technical Marketer.
