Conversion is often thought about too simply. We go from an impression, to a click, and then a conversion.
But what if that click doesn’t lead to a conversion? People are not inclined to buy after one interaction with your ad.
Finding out what is going on in between the impression and conversion is not easy. Your audience could click on an ad, email, or an organic search.
How do we quantify the value of an impression or a click that don’t lead to a conversion? These six attribution models will help you understand.
The Last Click Model
Last Click gives all the credit to the last action before a conversion. It strips credit from anything else that happened.
There is no funnel, because there is only one goal, one content, and one target. This gives a very simplistic view of Goals, Content and Targeting (GCT) because we only have the goal of driving sales.
This makes calculating your ROAS difficult. If you have a 5% click through rate, 4% conversion rate, $10 spend and $100 in revenue. Your ROAS would be 10x right?
But what if you had twice as many clicks? This means that your click value is half of what it was before. This is not very accurate.
People rarely buy after the first touch. Many times they have to see a variety of content before feeling comfortable to make a buying decision.
That is why it is important to break down the in-between process from the first impression to the final conversion.
The Overlap Model
An overlap uses multiple touches. It says that somebody will interact with multiple content before deciding to buy.
We use this model to determine how many people we can bring from one stage to the next. This is referred to as Infinity Analysis. The more data we have along the way, the more power we hold.
They could click on an ad, watch a video, sign up for an email list or talk to a friend about your product. After seeing, hearing, and doing, your audience will be motivated to buy.
In this scenario, we could say all four touches are worth $50 combined. If we divide $50 into those 4 steps we learn that each click is worth $12.50. Where as in the last click model the data would show that the click was worth $50.
The Viral Spiral allows you to cycle your audience through different combinations of Goals, Content and Targeting. You will feed them different content based on their position in your funnel, Audience, Engagement or Conversion.
Continually optimizing your content is vital. Calculating how many people have moved from point to point in your sequence gives you power. This will determine what your most valuable piece of content is and guide you to what you should create next.
With enough data indicating where the overlaps are happening, you can assemble these overlaps into a particular sequence. Sequences are if/then statements that help automate your funnels.
A sequence could be as simple as, if they watched this video, then they receive a 10% off your next purchase. If they’re a mom that bought tickets to tomorrow’s game, then recommend they buy a jersey.
Assembling complex if/then conditions is how you build a funnel over time and that’s why attribution, analytics and campaign optimization are actually one and the same thing. This can’t be understand when using the last click model.
Assigning fraction credits will help you understand that there are multiple steps through conversion. If you are only giving credit to the bottom of the funnel, you’re going to overlook engagement.
Let’s say you’ve been looking to get married and your applying the marketing principles we have been discussing. You first, start a conversion path from when you first meet your significant other. Your first date, to the first kiss, to getting engaged, to eventually have a wedding ceremony.
You determine you are 50% of the way to marriage after certain milestones have passed. After more time passes you find yourself 75% of the way there. This progression continues to happen until you are at the wedding ceremony, reaching 100%.
The more complex the sequences are, then the more you can do lift testings. Lift testing is like adding a variable a scientific experiment, to see how it affects the results.
The same tests are performed when testing for a new drug. They have a test and placebo group. Afterward the two groups are tested separately, to see what difference they can find between the two.
Let’s say you have a one click sequence from Point A to B. What would happen if you were to lead people through an intermediate Point C. Would they be more or less likely to buy?
What we have found is different people need different amounts of information before they will convert. If you know that by exposing your audience additional information you can triple the conversion rate, from 5% to 15% then that’s called lift.
Lift testing is not the same as split testing, its more than changing a headline of a landing page. It proves that by running a certain piece of content you saw a direct increase to Audience A You know this because Audience B was not exposed to that content.
The Media Mix Model
We know that any social media platform affects one another along the sequences. The Media Mix Model, determines what effect running these different platforms together has on our results.
You could be running ads across AdWords, Facebook, YouTube, Email, in-store and on TV. If you want to see what impact Facebook, YouTube and TV ads have together, segment your audience and only run ads on those platforms for that group.
Does that cause more people to search on Google where we run TV? If people are seeing YouTube videos, are they more likely to respond to our Facebook ads?
That’s the beauty of Marketing Science, it’s the same thing as sampling or polling. Every single combination doesn’t need to be ran to know what works.
If you want to know whether Donald Trump is ahead or behind in the polls, you don’t need to poll everybody, you just need to have the 5 – 10% . Then you can establish your 95% confidence interval plus or minus 2%.
We’re looking at inputs and outputs, so instead of saying, “I’m trying to take this $10 and I need $100 of revenue right away”. I’m saying, “Look, I can put in $10 and get more than $10 of value from this point to this point in my stage along the way, then I’m profitable”.
It takes more than a click
We are moving away from the last click model, focusing on the bottom of our sales funnels. Keep in mind your attention to overlaps, because conversion is no longer as simple as impression, click, convert. With our plumbing in place we can set each stage and trigger.
The more action based triggers will lead to more if/then iteration, which helps build more complex sequences. Complex sequences allows experimenting with different channels to perform more lift test. The more overlap we find between these different channels the more we will move into a full media mix model.
Attribution, analytics and optimization are all the same thing. The idea of MAA is to understand why and what can we recommend as the next step. When we can figure that out, marketing automation becomes possible.
Which of these attribution models can you implement to calculate a more accurate conversion cost?
About the Author
Dennis Yu is the Chief Executive Officer of BlitzMetrics, a digital marketing company that partners with schools to train young adults. Dennis’s program centers around mentorship, helping students grow their expertise to manage social campaigns for enterprise clients like the Golden State Warriors, Nike, and Rosetta Stone.
He’s an internationally recognized lecturer in Facebook Marketing and has spoken in 17 countries, spanning 5 continents, including keynotes at L2E, Gultaggen, and Marketo Summit. Dennis has been featured in The Wall Street Journal, New York Times, LA Times, National Public Radio, TechCrunch, CNN, Fox News, and CBS Evening News.
He’s a regular contributor for Adweek’s SocialTimes column and has published in Social Media Examiner, Social Media Club, Tweak Your Biz, B2C, Social Fresh, and Heyo. He held leadership positions at Yahoo! and American Airlines and studied Finance and Economics at Southern Methodist University as well as the London School of Economics. He ran collegiate cross-country at SMU and has competed in over 20 marathons including a 70-mile ultramarathon.
Besides being a Facebook data and ad geek, you can find him eating chicken wings or playing Ultimate Frisbee in a city near you.
You can contact him at firstname.lastname@example.org