Introduction – Attribution Modeling
In the last two posts, I tried to explain the basic concept of Attribution Modeling and how to use Google Analytics to create custom Attribution Models. The previous two stages are relatively quite easy to execute. This last part is the point things get a bit trickier. If you are not yet clear on the concept of Attribution Modeling, make sure to go back and read the two previous parts of this series. To proceed reading without doing that will honestly be a waste of your time. So, make sure you understand the basic concepts before you read this final part of the series.
To read the previous parts of this series click here:
Let’s dive in. The whole purpose of understanding Attribution Modeling is so that as online retailers and marketers you can learn how to calculate and measure ROI (Return on Investment) precisely. It’s now time to dive in deeper, in order to truly understand the real impact that different channels have on your revenues.
For the sake of simplicity, I have classified all the commercial websites in the world in the following three main categories:
- Lead Generators
Since each one of these websites has its own unique characteristics, it is mandatory to deal with each differently. As a general rule of thumb, attribution-wise the e-commerce sites are the simplest to understand whereas SaaS and Lead Generators are more complex in nature and therefore require a different approach.
In the following lines, I am going to talk about new user acquisition. Let’s leave user retention and lead nurturing for some other time.
In e-commerce websites, all activities take place online and most of these activities are completed within the 90-day conversion window from the time of the first click to the first purchase. Google Analytics, therefore, provides a solid solution for Attribution Modeling if implemented the right way. However, you have to deal with a cross-device tracking issue (users who enter a website from different devices therefore not tracked properly). This can be easily resolved by making users register. This way each time they log in from a new device you can perform the cross-device tracking by placing a cookie on their device.
Lead Generator sites, on the other hand, struggle with different issues. Since the actual sales do not occur online, it becomes even harder to attribute real value to each lead; therefore, a rough guess is applied for overall averages. There is a separate process to address this issue. For that, I recommend reading this post on Offline Conversion Tracking in Google AdWords.
SaaS sites are the most difficult to crack when it comes to Attribution Modeling. This is because the acquisition funnels in SaaS sites are the longest ones and it takes a lot of time to progress through the funnel until a final sale is closed. On top of that, the buyers usually use different devices which increase the level of complexity.
In order to fully understand the complexity of SaaS systems, let’s use this fictitious example.
Think of a cyber security company that offers a SaaS product that provides cyber security defense to your site. The product offered is cloud based and is distributed by a SaaS product called ACS. As a customer, you enjoy a 60-day free trial period with some features. Once 60 days are over, you need to upgrade to a paid account or you do not get to enjoy these services anymore. Let’s have a look at the user’s journey.
|Registered for Trial
|No Action – Trial Ended
|Upgraded (120 days after the first click)
As you probably noticed, it’s hard to calculate how much credit each channel deserves and how to track the 120-day buying cycle when Google Analytics allows only a 90 day Lookback window. Under such circumstances, we have to find a connection between the marketing and sales (CRM).
In order to sort things out, let’s divide the buyer’s journey into two segments:
- Non-Registered Users
- Registered Users
As soon as a user registers for the first time, the magic begins.
If we can somehow track all the touchpoints up until he/she registers, and save the path to the CRM, we might be able to create a longer, more accurate time-bound tracking.
It is only when the user upgrades their account to a paid account, that the attribution can take place.
The question, however, remains – how much should we attribute for each channel. This question becomes more complex on occasions where the user upgrades their account to a monthly retainer making the customer life time value indefinite and accumulative.
There are two approaches to solve this issue:
- Using Customer LTV
Since by now you know how many customers to attribute to each channel, you can calculate your current average LTV, and deduce an average value to attribute
- Accumulate Revenue and Take a Longer Lookback
Data accumulates over time, take a long look back in order to get a clearer picture of what’s working better. Looking a year back, you can see how much revenue was generated within the first quarter. Compare that data to the rest of the year to analyze whether or not over time your estimations are in line with reality.
I recommend combining the two approaches. Customer LTV is good for the ongoing optimization. A longer Lookback will provide you with an understanding whether the data remains the same.
You may need to use some attribution software/tools in order to reach the most precise conclusion. Here is a list of some popular attribution tools you can use:
If you are a SaaS company and you practice Inbound Marketing, I personally suggest you use HubSpot as it offers a complete marketing cloud. Do not mix PPC with Inbound marketing, PPC is actually a part of Inbound marketing and is not a totally different ball game as some assume. You can get more insight into this topic in my post about Inbound Marketing v.s PPC.
Attribution Models for Mobile Apps
When it comes to mobile apps, most publishers and advertisers rely on the Last Click Attribution Model. There are a number of mobile-app attribution software programs that can easily track In-app conversions and installs and easily attribute them to the agency/channel that has created the last click. An example of these programs is:
Pitfalls to avoid
Using attribution modeling without the proper knowledge and experience can cause even more damage then not using them at all. Here are a few pitfalls to avoid:
- Not Testing/Testing Without a Coherent Theses
It goes without saying that like every digital activity, you should constantly test these attribution models. However, what to test and how to test sometimes remains answered. Keep in mind that different attribution models work differently for different businesses. At the end of the day, it all boils down to your understanding of these models. You should know which models work best for you and by changing these models you can analyze any increase or decrease in your profitability and by what percentage.
In order to avoid inaccurate assumptions and an unclear understanding of the tests results , it is recommended to avoid multi-testing different theses. It’s best to conduct each test as uniquely as possible and to measure the results before proceeding with another test.
- Data Leaking
If you are not able to handle Google Analytics correctly, you may experience breaking sessions. With incomplete data, you can easily lose track of your progress. Frequent session breakage can result in Google Analytics creating a new session for a user who is already on your website. This can lead to traffic source displayed as “not set”. As a result, you may not be able to attribute the revenue correctly.
- You should also be familiar with the factors that can cause these sessions to break:
- incorrect implementation
- sessions are taking too long
- cross domain tracking
- Trusting the System
You should never blindly trust the system, you need to consistently test the system and make sure that the data is accurate. Large websites with alot of data tend to loose information on the way. Try to avoid discrepancies as much as possible.
- Not Drilling Deep Enough
Make sure to use raw data and try to round up averages from different segments. You may think it will be easier, faster and better, but it is not the case. Sometimes a 5% difference can shift a campaign from a profiting one to a losing one. So, never use estimates or average data figures, dig deep to use accurate data.
- Do not Miss the Big Picture
You need to keep in mind that Attribution Models are models and by no means facts. They produce results based on the set of rules that you define. These rules are not the same for every other business, so you should test every model to see whether it generates more business or not. Never take your eyes off the big picture, i.e. making profits.
Conclusion – Attribution Modeling
Regardless of the type of business you own or you are advertising for, using the most appropriate Attribution Model can result in an increase in your marketing campaigns and ultimately your business profitability. If you miss out on estimating the valuable contribution each marketing channel provides to your business, you can find yourself facing failure. There is a good saying that truly reflects business growth “if you’re not moving forward, you’re moving backward”.
In the end, it all boils down to keeping yourself ahead of the game. Although your main focus should be on increasing ROI, you also need to keep on testing what to do to find new audiences and channels to add to your media mix.
If you found this useful, and you’re curious to learn more and find out how to improve your lead generation, we invite you to download our 30 greatest lead generation tips tricks and ideas eBook.