Attribution Modeling is a framework used for analyzing how different marketing channels should be credited for a lead or conversion. It helps marketers to get a better understanding of how and when multiple touch points assist a conversion.
Basically, Multi Touch Attribution Modeling enables you to figure out where to put resources in the future. The framework will inform you that a certain channel resulted in a visitor buying something. So, if you can identify that a particular channel is the one responsible, you can allocate more budget for the upcoming cycle.
It’s about transparency. In our sweet little minds, we would love to assume that a visitor finds your website or views an ad and then immediately converts. But, as the purple alien in Avengers: Infinity war said:
With the increasing fragmentation of platforms as well as devices, attribution is becoming more and more complicated and challenging. In reality, a visitor is likely to visit your site multiple times before actually converting. That visitor could come through a social post, after an interval through an e-mail, and then directly to finally convert!
With the presence of multiple touch points in a user journey and each of them playing their parts, it can be difficult to measure the value of each marketing channel. Marketing teams put a lot of effort across various channels and hence, would love to know what is working and what is not.
Various model comparison tools are present in the market which can help you with this by applying several pre-defined or custom attribution model on your data. Some of the most common attribution models are First Touch, Last Touch, Linear, Time Decay and Position-Based. Let’s discuss their pros and cons now:
1. Last Touch:
This model assigns 100% credit to the last the channel from where the user came from.
Let’s take an example, a visitor first arrives on your website through an email, then through a display ad on some website and finally comes directly to convert.
In this scenario, the Last touch model will give all of the credit to Direct Channel.
Pros and Cons:
Last Touch model is the most basic and simple of them all. This is the default model for almost every analytics tool in the market.
Since the model considers only the last channel, it completely ignores the importance of what might have happened before coming to this particular channel.
Although, this can be a good option for businesses with a shorter purchase journey or the ones which focus more on sales generation.
2. First Touch:
This model is as simple as the Last touch model with a major difference that it gives 100% of the credit to the Channel which leads to the first interaction of the visitor.
For the last example, where, a visitor first arrives on your website through an email, then through a display ad on some website and finally comes directly to convert, all the credit will be assigned to channel Email as it leads to the first visitor interaction.
Pros and Cons:
Similar to the previous model, this model also focuses on just one channel and neglects what happened after the first interaction. Further, this can be vulnerable to things like cookie expiration period and lookback window.
However, this can be considered by businesses with the aim of demand generation or awareness.
3. Linear Attribution:
Linear Attribution is the first step towards multi-channel attribution. It assigns equal credit to all the channels in a user journey.
Remember the example we used, where, a visitor first arrives on your website through an email, then through a display ad on some website and finally comes directly to convert, the credit will be divided equally between all three channels.
Pros and Cons:
Linear Model is quite straightforward and gives a simple message that each channel has importance.
But, in (most of) cases, it might not be a good idea to consider a less important Email channel to be as equal as a highly impactful webinar or Affiliate channel
Position-based model focuses on the order of channel appearing in a user journey. By default, it assigns 40 – 40% credit each to first and the last channel that appeared in the conversion path while giving the remaining 20% to all the other channels that appeared in between the first and last channel.
I am sure, by now, you have learnt that example by heart, where, a visitor first arrives on your website through an email, then through a display ad on some website and finally comes directly to convert, in that scenario, the 40% credit will be assigned to email and direct channel while the remaining 20% credit is given to display.
Pros and Cons
This model barely recognizes the importance of channels that might have had more impact than the first(or last) channel. Also, the fact that it assigns equal credit to the first point of contact as well as the last touch point before conversion.
However, it can be a great tool for answerings the questions, ”What generated the demand?” or “What lead to the conversion?”. Also, this can be helpful for businesses which don’t emphasize nurturing a lead for a long time.
5. Time Decay
Time-Decay model assigns credit based on the difference between the time of channel interaction and actual conversion. The closer the channel is to the conversion the more credit is assigned to it. The model is based on a half-life whose default value is 7 days, meaning that a touch point occurring 7 days prior to the conversion will receive 1/2 the credit of a touch point that occurs on the day of conversion. Similarly, a touch point occurring 14 days prior will receive 1/4 the credit of a day-of-conversion touch point.
Alright, here we go one last time, using the same example, where, a visitor first arrives on your website through an email (Day 1) , then through a display ad on some website(Day 7) and finally comes directly (Day 14) to convert, in that scenario, the 57% credit will be assigned to direct channel, half of direct i.e 28% will be assigned to Display while the remaining 14% (half of direct) credit is given to email.
You are wrong if you are thinking that one model can answer all your questions. The models discussed above are pretty generic and make some harsh assumptions about your business and user behavior.
Custom Modeling: Penultimate Step towards Attribution Utopia
A custom model can you help you build a framework that is tailor fit for your business. It has the potential to get you the most granular look at what is leading the visitor towards conversion. The trick to take full advantage of the custom attribution model is to ask questions which will help you in getting a clear picture and context that will further direct you to complete the model.
Few of the potential questions:
1. How many touch points are there prior to conversion as well as how much time does it take to convert on an average?
As the number of channels increases, the model becomes more nuanced. You might want to give extra credits to more important channels like Social or to someone who is coming after attending a webinar instead of less important channels like Email.
Also, for business with longer conversion paths, you would like to give more credits to channels that are closer to the conversion as compared to the ones in the initial part of the journey.
2. How do you value certain user behavior?
As a business, you would want to identify/reward channels which lead to a certain important behavior. For example, you could give more credit to a channel which leads to Product Views or some other important engagement.
3. What is the purpose for attribution modeling?
It is important to know for what purpose are you creating the model. A business focusing majorly on demand generation would credit channels in the initial phase more as compared to the ones in the center or end. Similarly, a business focusing more on sales would reward the channels in the proximity of conversion.
It’s completely on you what path you are choosing for attribution. For quick answers, you may compare two – three models together and get something out of them. For the long run, you should design a custom model and maintain it over time according to business requirements or changes.
Just like everything in life, attribution perfection is like an asymptote, you can never reach 100% but you can keep working towards it and become better and better at it!