INTRODUCTION TO ATTRIBUTIONS
What if I told you that you’ve been measuring your digital channels wrong the entire time? Yes, you read correctly. I’m going to show you how different channels are assisting each other and how to really measure your multichannel digital marketing performance.
It is very common to mistake statistics in web analysis when measuring digital marketing performance and how to successfully approach measurement for multichannel marketing without being too shallow and misjudging the channels actual performance. It’s wise to have a holistic approach to digital marketing and to learn how to allocate your marketing budgets accordingly.
Most of the web analytics softwares use “last click” attribution by default. In this post I’m going to go through the most common attribution models, show you the Google Analytics “Model Comparison Tool” and “Multi-Channel Funnels”.
My goal with this article is to give you a wider perspective when tracking your full digital customer journey and when measuring your digital channels performance, Return On Investment (ROI) and Customer Acquisition Cost.
What is the Last Click Attribution?
Last-Click Attribution is the default way how most Web Analytics tools give your acquisition channels credit. For example if you run an Banner Campaign on an external website and someone clicks on that ad and comes to your website. In this case the ev. purchase credit (conversion) & traffic source will be assigned to that specific external website. This is pretty much a big part of Google’s business model.
The problem is, what if this customer first saw your website/brand on an advertisement in an online newspaper and after research online clicked on your ad on Google. Another issue is tracking organic brand keywords or direct traffic which usually stands for a majority of your traffic and conversions. Sure, these people might have typed “Your Brand” in Google and then Google will get 100% of the credit for that customer acquisition.
But this isn’t the truth right? Sometimes we
want need to know what channels introduced this customer to your brand, not only what drove the “last click” (previous step). This is because the customer journey is much more complicated than that, usually your website visitors come to your website several times and via different channels before converting.
Let’s take a step back before entering too deeply into the fun (but nerdy) digital marketing stuff. What I’m going to write about now is Multi-Channel Attribution Across Digital Channels referred to as “MCA-ADC”. This is also what I wrote about above.
But please be aware of that many C Level executives (CEO’s, CMO’s etc) may refer to Multi-Channel Attribution, Online to Store (MCA-O2S). This is when measuring how online advertising is affecting the physical (offline) store sales. If you’d like to read more about this I suggest you to read this article from Think with Google: “Proof That Online Search Ads Can Boost Offline Store Sales”. They could of course also refer to Multi-Channel Attribution, Across Multiple Screens (MCA-AMS) meaning for example how to track TV Commercials with Web Analytics. Here the usual dilemma is when people see your TV Commercial -> Search for you on Google -> Google Adwords click / conversion and again, Google receives the conversion credit instead of the TV Commercial.
TRACKING THE Full Digital Customer Journey
Let’s focus now on how you can track your customers across digital channels: Social Media, Direct, Organic Search and Referral. A random but normal path could be: Andreas sees your Social Media Ad -> Clicks to your Website -> Leaves the site. A few days later he searches for you on Google and clicks to your website through the Google Adwords ad and subscribes to your newsletter (Micro Conversion). A few weeks later he clicks on a cool promotion on your newsletter email and makes a purchase (Macro conversion).
Now how do we assign all these channels their appropriate value? Let’s take a look at some different Attribution models found in Google Analytics below:
Last-click or Last interaction attribution model. The last touch point will receive 100% of the credit for a conversion. Problem scenario’s with this are many. We are missing out on loads of important data when only looking at the last-click. We’ve done advertising on many different channels but our URL is very easy to remember, so why should it all be grouped in the “Direct” channel. This model is frankly very bad for obvious reasons.
Last Non-Direct Click. This is the default model in Google Analytics. The Last Non-Direct Click attribution model, all direct traffic is ignored, and 100% of the credit for the sale goes to the last channel that the customer clicked through from before converting. Problem scenario: I Google for “cheapest iphone 6” and find an interesting organic offer but I don’t buy it yet, I leave the site. After a few days when my pay-check is in, I Google for the brand name because I remember the store name, I click on their Google Adwords AD and now Google Adwords will get 100% of the credit. This is problematic especially for SEOs because SEO really had an impact here, but Google Analytics doesn’t count it as a converting channel in this case.
The Last Adwords Click model gives 100% credit to the most recent Adwords ad that the customer clicked on before converting. This model is used for Google Adwords and is different with the Google Analytics attribution. In GA a goal completion is counted only once per sessions. In Last Adwords Click a conversion can be counted many times per ad click. An example scenario: Andreas clicks on a Adwords Ad on the 26th of May comes back the next day 27th and makes the purchase. Here Adwords will get 100% of the conversion credit the 26th may while Analytics will report the conversion for 27th.
The First Interaction or First click model gives 100% of the conversion credit to the first touch point. Problem scenario: I click on an Advertisement for summer 10% discount but can’t find anything I like so I leave the website. I return a month later when I see a “Fall 5% discount ad” and decide to buy clothes for the fall. The problem here is that the first interaction, in this case, the “summer 10% discount” campaign will receive 100% of the conversion credit and not the Fall campaign which should be the correct one. However this can be a good model for unknown brands to understand what channels are driving the first interactions to their website & brand.
The Linear attribution model. Now we are starting to analyse the full customer journey. But still far from great. The linear model will assign all the touch points equally the same credit. This is unfortunate, because some touch points are more or less important, for example micro & macro conversions. Problem scenario: I click on a great ad on Facebook and enter a website, however I don’t find anything interesting so I leave. Later I see a great campaign for christmas presents and decide to purchase. Now both Facebook and the second channel with get equally the same credit: 50% and 50%.
The Time Decay attribution model. This is the model that makes the most sense for me. All touch points will receive credit, but the latest interaction where the conversion is made receives most of the credit. This seems logical, right? This model can also be customised with time delays and percentages, very good!
The Position Based attribution model. This gives 40% of the credit to the first and last touch point. Could make sense if we run an TV campaign for an unknown brand where the first (direct) touch point is very important, smaller touch points in the middle get to share 20% for getting the visitor back to the website, and the last interaction where conversion happens receives 40%. I would also use this model for many e-commerce websites.
What model should i use?
Personally I think it’s impossible to say which model you should use. It depends on what you need to measure. But the attributions are very important to really understand your digital customer journey. Forget “Path Finder” or similar tools, these will show you too many different paths that are not giving you any insights. It’s impossible to choose one path to analyse. A better idea is to try to understand how Social, Paid, Organic, Referral and Direct work to assist each other. But usually I recommend to not assign a value to “direct”. Direct is not a marketing channel, but a result of your marketing activities.
Scenario 1) You have a unrecognised brand and want to measure brand impact
My recommendation here is to use the “First Interaction (click)” attribution. With this attribute we can give more credit to the first driving channel that actually introduced your brand to the customer in the first place.
Scenario 2) You have an e-commerce website and want to measure performance.
I would suggest to start with Time Decay to see how the campaigns are working and how the channels assist each other. You probably will also get important insights from the “Position Based” attribution so you can give most credit to the first touch point (first contact with the customer) and finally the last touch point where the conversion happened. As stated earlier, this model gives 40% for the first touch point, 40% for the last touch point conversion and 20% divided for the channels in between.
WAIT BEFORE YOU JUDGE YOUR CHANNELS
Too many times are companies judging some of their digital channels too fast. Before really understanding what is going on. Questions like is this channel really working because we don’t see any ROI? Where should we increase or decrease our budgets? How should we really divide our marketing budget?
Google Analytics Multi-Channel Funnels Overview
Assisted Conversions tell you how many conversions went through this channel on its way to the final interaction. This simply means that assisted conversion can happen anywhere in the flow but not as the last interaction. An example: A visitor searches for “Black shoes” and enters a website via organic search, he leaves the website without making a purchase and then returns the next day via an Google Adwords Ad. Now Google Adwords will get the conversion credit and organic search will get the assisted conversion credit.
Let’s start by taking a look at Google Analytics -> Conversions -> Multi-Channel Funnels Overview
Blue: Paid Search (Google Adwords in this case), Yellow: Direct, Green: Organic Search (SEO). What we can see here is that all these three are overlapping quite a lot. This means that we most definitely have attribution problems here. For example my questions that arise now are:
1) Are we measuring Performance (in this case Google Adwords ROI & CPA (Cost Per Acquisition)) correctly?
2) Is the SEO Reporting showing underestimated performance figures?
3) What correlation does the Direct traffic have to organic & paid channels?
4) Have we set a value for micro conversions?
5) What insights can we draw from the correlation between Organic & Paid?
To fully understand this we will now move over to attribution models using the:
Google Analytics Attribution Model Comparison Tool
Let’s go to Google Analytics -> Conversions -> Attribution -> Model Comparison Tool
In the above screenshot from the GA Model Comparison Tool, I have chosen to compare “Last Interaction (Click)” default model against Linear and Time Decay. If I was in a meeting with a CMO I would advise to simply follow the arrows in the red marked square.
Insights drawn from the Linear Attribution Model
First if we look and compare the “Last Interaction” data with the “Linear” data we can notice a few interesting things: Paid Search has actually more conversions and a lower Cost Per Acquisition than expected. The correct CPA is $12.45 and not $13.25 and Adwords also has a few assists as expected. Of course because of this we can see that the conversions in “Direct” has decreased because we now know where they really came from. Organic Search (SEO) is performing better than expected as well. If we look at the red marked area Organic Search has: 13.78% (Linear) and 12% (Time decay) more conversions than expected. I can already here draw the conclusion that organic is a very important channel for this client.
In this specific case the “Display” is tough to analyse because it has so little data, so I’m going to ignore that one. But if there was a few 00 after the 22 conversions we could draw the conclusion that display is really working well to introduce new customers, where it’s acting as a great branding channel.
If we’d like to take a further look into branding performance for this unrelatedly unknown client we can switch to “First Interaction” attribution model like below:
As seen in the image above. If we choose “First Interaction” where we give 100% of the credit to the FIRST touch point with the customers – Display is really doing this well. But now as mentioned, this is a small channel for the moment so we won’t draw any actions in this particular situation.
Wait, what are we looking at?
The most interesting part to look at in the “Attribution Model Tool” above is the red marked area in the right column. This shows you the % change in Conversions compared between your attribution models. For example Display as First Interaction has 59.09% positive change in conversions compared to “Last Click”. If the goal is branding and understand what drives the first interaction, then simply follow the percentages (up & down) in the last right column. Simply allocate your budgets accordingly.
No model is perfect and never will be. But the important part is to have a holistic approach to these challenges (and opportunities). Remember, the further away we get from an actual conversion funnel, the more difficult will it become to measure it. I won’t be going any deeper into the insights and actions process in this article.
SOCIAL MEDIA’S opportunity & challenge
People share stuff nowadays – take this into count. If you get 10 people to share a product of yours on Facebook, with an average of 388 friends this could reach up to 3380 people. If you are investing a lot of time and effort into social media, remember to analyse how social media is assisting your conversions. This will make you understand it’s value better. Maybe you had underestimated or overestimated it’s real value to your business?
If things weren’t complicated enough: A big challenge that social media is up against is also attribution within it. We’ve learned to think beyond engagement and followers, but this still has a big impact when building your own brand community. I am guessing a further digging into this could be done using cohort analysis to really understand the performance impact of social media. An interesting experiment would be: Run a 24h campaign to acquire Facebook Fans and engagement and using a Cohort Analysis follow how these fans engage with your brand over time. Please comment if you have any experience doing this :-).
Strategic Budget Allocation
By not only looking at the standard Google Analytics Acquisition reports but fully analysing how the different channels assist and work together to drive performance is a crucial part of allocating your marketing budget.
The Google Analytics Attribution Model Comparison Tool and the Multi-Channel Funnels gives you a very easy overview of how these channels collaborate towards your strategic objectives.
Multi-device tracking is important to remember. A usual scenario is when people browse around with their iPads or mobile devices and then come back to make the final purchase with their computers.
To read more about Multi-Device tracking with Google Analytics real this full guide from Optimizesmart.com.
To read more about Facebook Multi-Device tracking read this article.
By looking further than “Last-click” you will be able to make smarter data-driven decisions in the future. By understanding and using the different attribution models you will be able to make smarter budget allocations for your digital marketing channels and understand how these collaborate to assist each other. The next time you are meeting with a CMO or CEO, remember to bring up this discussion.
Please comment & share this article if you found it interesting or helpful. Feedback appreciated. Thanks!