Last year advertisers spent close to 194 Billion USD on digital advertising and this number is expected to reach 355 Billion USD by 20201. Digital advertising has become one of the most cutting edge markets today. This is something that has not happened overnight, the industry has come a long way from simple text ads in search engines to innovative immersive ads done via studying the user behaviour and bidding (< 100ms)  in real time auctions.
In all of this sophistication and maddening rush to acquire more customers for your business there are still a few questions which still remain unanswered.
Which platform(s) should i invest to get the maximum returns on investment.
What is the impact of a particular campaign on my overall sales and brand.
All of these questions lead to what today’s marketers like to call Attribution Analytics / Multi Touch Attribution / Multi Channel Attribution. 

 

To give you some context, in August 2013, Google commissioned Forrester Researchto explore the role of attribution measurement and how enterprises have adopted and leveraged this advanced measurement methodology. Forrester found that companies see the value of having an attribution strategy, with 67% of marketers indicating that attribution is highly valuable and helps them make smarter marketing and media decisions — but only 31% of marketers are really using it.

 

Lets talk about Multi Touch/Channel Attribution – In simpler terms this means when a user is exposed to your multiple touch points / channels / mediums of your application, how much each of these channels / mediums / touch points is contributing to the overall conversions and sales of your product.
The above diagram tries to depict a simplified version of a user journey, for example visitor c before buying a product, went on to the site organically, then received an email from the advertiser about the product and finally saw an ad on Facebook and bought the product. This is called as the Conversion Path of a User.
 
If you have google analytics installed and have set up goals in GA. Then you can see your top conversion paths by going to  Conversions -> Multi Channel Funnels -> Top Conversion Paths.

 

 

Understanding your conversion paths is the starting point to understand Attribution Modelling. As you can see in the above screenshot how complicated can the path leading to conversion can get. All of the attribution models which you have heard till now or will hear in future take this data ( and more ) as an input and then try to calculate the significance of  every touch point/channel which led to the actual conversion.

 

Now Let us see how some of the basic attribution models work. For simplicity sake i have exported the data from above, made it a little simpler to understand and have given every conversion a value of $1.

Last Interaction

The last interaction model is one of the most common, widely used models present today, In this model the last touch point is given 100% credit for the conversion. If we take the above example the last attribution channel would show the below results.

You can easily compute this data manually by counting the conversions where “Direct” interaction is coming in last. You can very easily derive that account to last touch attribution model, direct traffic is whats driving most of your conversions.

First Interaction

First interaction model is completely opposite to the last interaction model ( this seems a little obvious ). The first interaction model gives 100% credit to the first touch point which lead to conversion. Consider the main data, we will get the below results according to first touch conversion models. This model should be used if you are a new brand and are trying to make you mark. Since this gives 100% weightage to the touch point which got you traffic you can easily see whats working and whats not working for you.
You can now see in the above data that paid search came in 4 times as the first touch point and drives equal number of conversions as direct traffic.

Linear 

This model gives equal credit to all the touch points that lead to conversion, so if you have 4 touch points which lead to a conversion, everyone will get 25% share. This model is generally good when you have  long running campaigns throughout the conversion cycle.
As you can see that in linear every touch point gets equal weightage of the conversion value, like you can see that Paid search came twice in the first row of data and again in the fourth row.

Hence the total contribution for paid search becomes $3 + $0.5 = $3.5

Time Decay

This model attributes more value leading up to the conversion. This model is ideal if the sales cycle is short. For example, if you run a two day promotion then channels that were interacted with a couple of weeks before the conversion should have a lower value than channels that were interacted with during the promotion period.
Since, this involves time related data we will not show its calculation here, but if you want to calculate your conversions based on the time decay model, you can very easily do that like we did for the linear model. Just that instead of equal weightage we now give weighted weightage based on how close the touch point was to the actual sale.

Position Based

This model is also very similar to time decay and linear with a twist that you give maximum weightage to certain positions in the touch points, for example in the below model you can see that the first and last touch points contribute a total of 70% weightage of the conversion. The rest 30% is divided equally among the other touch points.

Customised

Customised models, as the name suggest let you define your own weights for the conversion.

Google analytics lets you easily define your custom models through an easy to use UI. As you can see in the screenshot below you can select a baseline model and then edit the percentages based on what you feel is the best reflection of your business. You can download the complete “Google Analytics Attribution Playbook” from here.

 

Which one should i choose ?

Here are some of the important things to consider when selecting an attribution model. Every business is different and its goals are equally diverse. Consider the following points when you define your attribution model.

  • What is more valuable to you: a product signup or a sales qualified lead or a product sales? If its a product flash sale, then you should go with linear, or time based decay models.
  • How much of your traffic comes in organically ? If most of your traffic comes in organically and searched for your product but does not buy, then you would not want to go with the first touch models.
  • What is the average LTV of your customer?
  • What is the average time it takes to close a customer ?
  • Do your customers pay in monthly subscription payments or one-time charges? If its a recurring business then probably you need to use a more long term strategy of holding the customer back on your platform.
The reporting and analysis part for ad campaigns has been ignored for a long time now, with new advertising networks coming in every day, they make it easy to run the campaigns on their platform, but when it comes to doing some analysis of the performance of the campaigns, its too difficult to understand the data. What most of them rely on is exporting the data in big excel sheets and then assume that the advertiser will spend hours understanding them. What we need is a tighter integration of the advertising data sources with the analytics platforms. Simply understanding the behaviour of the users who clicked on the ad to check out the product is not enough. Today’s marketers need more sophistication to understand their ad campaigns in a better way. We need independent measurement partners for web like there are for mobile.

 

At Adalyz, we are on a mission to unify all the advertising data sources at one place so that the digital ad marketing can understand the complete picture of his advertising spend across different channels. We are working on merging the analytics and advertising data tougher
I hope you enjoyed reading this article, please let us know if you liked this article or if you have any suggestion, comments in the box below.

 

References:

https://www.statista.com/statistics/237974/online-advertising-spending-worldwide/
https://think.storage.googleapis.com/docs/forrester-cross-channel-attribution_research-studies.pdf