3 Marketing Attribution
Marketing Attribution is the science of assigning credit to different actors in a chain of events. The science of attribution is an important topic in marketing as device, channel and platform proliferation have made it increasingly difficult to attribute the right value to different marketing activities.
The main marketing attribution challenges that businesses are trying to solve are:
- Intra channel: How are my generic search terms supporting my brand search terms? Or, what is the role of my YT video campaign in supporting my GDN remarketing efforts? How much credit should I assign to my GDN remarketing efforts if I suspect this activity is over valued on a last click basis?
- Multi channel. How is Search benefiting from Display?
- Cross device: What is the effect of my mobile ads on my desktop traffic (conversion rates)?
- Online to offline: Is my online advertising budget driving sales beyond my digital properties?
- Offline to online: Is TV boosting my digital marketing performance?
- Cross platform: How are clients moving between my web and app properties?
The scope in this specific article will be limited to the first two (digital attribution),
How to start with Digital Attribution
It is often difficult if not impossible to do your digital attribution perfectly, but pretty simple to improve your current attribution efforts. As a result of the difficulties involved, many businesses have adopted a simple last click model, where one hundred percent of a conversion is attributed to the last touch point. This is a gross over simplification, but it is an attribution model that has served many businesses well by resulting in significant growth.
Many businesses I have dealt with have made the mistake of trying to go from zero to hero in one go. The best route however to better attribution is to become a little bit better every single week (or month). First of all, you should start by looking at what an investment in improving your attribution might bring you. With other words, is it worth it? Is there a lot of scope for improvement in changing your attribution logic? How much of your time and resources is it worth?
If your customer journey is short, both in length of time and in the number of touch points, often the effect of changing an attribution model is less impactful. A silly example. If ninety percent of your conversions is preceded by only one click, for ninety percent of your customer journeys, the difference between first click or last click attribution is exactly nothing.
In Google Analytics, you can check the Conversions > Multi Channel Funnels section, where you will find five different reports who all try to answer this question in one way or another. Keep in mind that these Google Analytics reports are all cookie based, so they will very likely underreport the complexity of the true customer journey. Google Analytics can’t track between the different devices that may have been part of the customer journey.
To invest or not to invest in Cross Channel Attribution
First, let us start by looking at how complicated the average customer journey is in how many touch points your customers have with with your website (and/or marketing channels) before converting. For longer more complicated journeys the attribution logic can be an important part of evaluating your different marketing efforts. So let’s use the Google Analytics Demo account to look at this for the Google Merchandise shop.
Go to Conversions > Multi Channel Funnels > Path Length report.
You wil see the following graph ->
Please note that in the drop down menu, we have un-ticked all the Goal conversions (like email signup for example) so we are looking exclusively at customer journeys that have resulted in actual purchases in the web shop.
You can increase the length of the customer journey. For some industries this makes absolute sense. If you do a lot of branding through the Google Display Network, you might want to look at impression assisted conversions as well (if you have enabled those).
What this graph tells you is that 55.21% of the Google Merchandise store transactions that happened in the specific time period you have selected (1st August – 1st of September for example) have had one touch point (usually a website visit; could be an GDN impression or Rich Media touchpoint as well) in the last 30 days prior to the transaction date. However, those 55.21% of the transactions represent a mere 33.62% of the revenue in the month of August.
This means two thirds of the Google Merchandise revenue is the result of customers journeys that consist of more than one touch point. Would a last click model do justice to all touch points in those customer customers journeys?
Secondly, let’s look at how you would value different marketing activity under two radially different models. To see the scope for improvement, I like comparing first click with last click models, and seeing how large the difference is in how you would valuable your marketing channels and/or campaigns. I chose the “last non direct model”, because conversions assigned to Direct are not traceable to any marketing activity, and want to see where I can influence results by changing my marketing mix. I add the linea model as an “in between alternative”, because there is room to add a third model.
So, I used the GA Demo account again. I select only transactions as the conversion types and increase the look back window to 90 days to make sure I see as much of the (same device) customer journey as possible. I change my primary dimension from “MCF Channel Grouping” to “Source / Medium”, to make these insights a bit more specific to the types of marketing activities that one runs. This results in the overview below.
You should focus on the last second last column. That percentage is the percentile difference in the number of conversions credited to the difference marketing activities or channels if you assign away from the last click and credit all the first click. That is the level you have to play with on a channel level.
Usually, you find display or upper funnel activity being undervalued on a last click basis. On the contrary, brand search, remarketing activity and affiliates are usually over valued on a last click basis. If you have a lot of channels where the difference between the models is huge an investment in better marketing attribution could result in significant efficiency gains.
It is good to be aware of how your marketing attribution logic affects your optimisation efforts.
Google Analytics Marketing Attribution shortcomings
Keep in mind that Google Analytics is cookie based and unable to follow users across different devices. In reality, customer journeys are often longer than Google Analytics is able to record.
Google Analytics is pretty amazing at reporting on the effectiveness of your marketing channels. Google Analytics is a webanalytics platform, so often a visit to your digital property is required. The free version of Google Analytics is able to record impression level data from GDN and YouTube (without a click resulting in a visit to your digital property), but not from other third parties (such as FaceBook, TubeMogul or AppNexus).
If you invest heavily in branding and prospecting, you should consider upgrading our Google Analytics to the paid version, and channeling your Search, Video and Display efforts through a platform such as Google Marketing Platform (GMP).
This will give you a lot of additional insights into the role the different marketing channels play in your acquisition efforts. Also, GMP has fantastic feedback loop that allows for an automation in bidding based on advanced data driven attribution models. This sort of technology comes at a price and is generally only worth it for very large advertisers.
If you are not in that club, no worries. The easiest gains often come from in channel optimisations.