This post is first in part of the Blog Series we are doing on ‘Budget Allocation –> Factoring in Attribution Modelling, Customer Lifetime Value and Cross Devices Browsing?’ In this series on using Customer Journey Insights for Budget Allocation, we will have 4 blogs with 1 each on aspects of Attribution, CLV, Cross Device Browsing, and one final post connecting these 3 factors to apply in Budget Allocation.
Attribution Modeling – What and How?
Let’s begin with watching this wonderful goal!
Wowww what an amazing goal by Messi? For those who think that Messi alone scored this goal can skip. However if you think it was a Team effort of FC Barcelona, then read on!
Attribution Modeling is like a Football game. It is all about knowing who scored the goal shot, while who all in the team passed around the ball, and created the perfect opportunity for Striker to Score a Goal. As in football, some goals are easy shots (brand searches), some driven by Strikers alone while some involve great passing and perfect timing. (read Customer Journey).
How Much do you Need Attribution Models?
One of the first questions before starting Attribution Modeling is to know how important Attribution Modeling is for your Business. A simple thumb rule practice in Google Analytics for that is to:
Open Path Length Report in Google Analytics. -> Conversions > Multi Channel Funnels > Path Length.
Alright if a substantial amount of your transactions have a Path Length Interactions more than 1, then you need Attribution Modelling. More the transactions with more than 1 touch point, the greater is the need. If the transaction with 1 touch point has a share less than 50%, then you better start now itself.
As a next step, couple the data above, with Overview of Multi Channels Funnels. It will give you Visualization of how many transactions involve more than a touchpoint.
Conversion > Multi Channel > Overview.
An important point to note here is that we are discussing Attribution Modeling here for Online to Online Sales. Attribution can further be extended into 3 more broad categories:
|MARKETING & SALES CHANNELS||BUSINESS SALES POINT|
|Online Marketing to Online Sales.||Ecommerce|
|Offline Marketing to Online Sales.||Ecommerce|
|Online Marketing to Store/Online Sales.||Ecommerce and/or Stores|
|Online Marketing to Store/Online Sales.||Ecommerce and/or Stores|
The first two are relevant for E-commerce businesses, while all 4 are applicable for Businesses that can have both Stores and E-Commerce.
Without any further ado, let’s look at some of the Attribution Models that Google Analytics offers, and how they align with equivalent Marketing Strategy.
A.) Last Click Attribution = Team only of Strikers!
This model implies that the one who scored the Goal is the hero and rest of team is a bunch of Jokers! The sooner you abandon this (or First Click Attribution), the better it is! Basically it attributes that the SOLE reason for the conversion was the last marketing touch point of the customer, meaning all your other marketing be it Social, Search, Display, Referrals were crap!
THE RISK: What it can bring in an organization is a dangerous culture of the marketers being under unnecessary pressure as many of their efforts might not be good at final converting points.
The Team with ‘All Forward Players’ might make some goals, but will surely get a goal on them every single time the ball goes to opponents. All your Competitors would love that you keep on using Last Click Attribution.
B.) First Click Attribution = Team only of Defenders!
The exact opposite of Last Click is First Click Attribution. With due credit to Avinash Kaushik’s joke, the best way to describe First click attribution is ‘akin to giving my first girlfriend 100% of the credit for me marrying my wife’.
If Last click is risky, then this is lethal. At least in Last click you get some knowledge as to which Campaigns are good at making customers convert – e.g.: Affiliates, Targeted Discounts. With First Click Model you are living on faith!
N.B.: Don’t skip them completely. In the best case scenario, however with a very low probability, this and Last Click Attribution might help in finding a low hanging fruit. Doesn’t happens often.
C.) Linear Attribution = Everyone in Team is Equal!
A Step better than the above two, yet this is as saying that every member in the team participated equally. The Risk with this is that you will never improve players or change team. No Hiring, No Firing, No Promotion, No Demotion (Read No Growth!).
The Players/Marketers need to understand each other’s strengths/weaknesses and each other’s role in the Strategy. And this takes us to next models.
Having discussed the above 3 almost directionless models, and having my biases I will just skip the Last Non Direct Click (everything in GA outside Attribution model is this number) and Last Adwords Click (Till date I haven’t come across anyone using GA to talk about its usefulness! Phew Google is crawling this!).
Let’s move on to the next 2 Very Useful Models (Position Based/Time Decay). And if you can understand and evolve, there is a 3rd one, the ‘Customized Attribution Model ‘coming your way. Customized models, as the name says are not preset, and can be directly aligned with ‘Your Marketing Strategy’.
D.) Position Based = Team with ‘Ball Control’ Strategy
- This is one Model where you can start aligning Business Strategy. Are you a team that needs a good first exposure and good Converting channels (typical for e-commerce)? Then keep high numbers for First and Last Click (in example here, it’s 40% each), while the rest of the midfielders share 20% remaining credit.
- Customize the numbers here. Shuffle around the default credit for First and last Touch points. Explore !
- Is your focus more on creating Brand Awareness? Give more Value to First Click.
- Are you competing on Price, is your focus more on Conversion -> Give more credit to Last Click.
- Write Down your Marketing Strategy -> Check out Path Length in GA for your account
-> Accordingly Change the Default Position Model to your strategy and Your Customer Interactions.
E.) Time Decay = Team with Attack Strategy
This is one model which without any customization that can give you great Insights. Two simple concepts:
- The Closer the touch point is to the conversion; more is the credit it gets.
- The Farther the touch point is to the conversion; less is the credit it gets.
Just like Position based one can customize it by changing value for the Half Life cycle of Decay. Don’t worry, it’s not as complex as it might immediately sound! As Google puts Half Life cycle, simply means ‘An assist click occurring this many days prior to conversion will receive 1/2 the credit of a click on the day of conversion.’
In short if you think your business involves some research before customers buy, use a larger number like 15. If you are into fast moving retail, use a small number like 7.
Play around, change the numbers many times, and explore how it dramatically effects your Marketing Channels ROI numbers.
There is a lot you can come to know about your own business just by changing the Primary Dimension to Source/Medium and trying the half-life numbers from 1 to 30.
F.) ‘Customized Model’ = Big Daddy of ALL
Team aligned with Coach (CMO), Manager (Marketing Manager), Staff (Marketers) and Training (Marketing Plan)
Customized Model can take Attribution Modelling to Totally another Level. Once you start using this, you will realize how limited usefulness all the above models have.
Indeed, one needs to first explore the above models, starting from almost directionless Last/First/Last Non Direct Click to next level of Position Based and Time Decay. They help one to understand the usefulness of the Attribution Modelling, and to get acquainted with the usage of the tool.
Coming back to Customized Models, this is where I love using Attribution Modeling in GA. Treat it like your own child, play with it and take good care of it!
So, how to use it? It’s as simple or as complicated as your Marketing Strategy.
First the Process:
Begin with a Base Model –> GA gives options of Linear/First Interaction/Last Interaction/Time Decay and Position Based.
Once you figured out that, set your Lookback Window. This Lookback window is the same number that you can set on the Top Left for any Model. The concept is that if you think for your business an interaction before 60 days carries no values, then you set it to 60. Then only impressions and clicks upto 60 days will be considered. GA allows you to use any number from 1 (wonder who’s gonna put 1) to 90.
Suggested idea to find best number for Lookback Window – Just go to your Time Lag report (Conversions > Multi Channel Funnels > Time Lag) in GA, and change the look back window here (try 30, 60, 90). The moment you find 75% of the transactions are done, that’s the number you are looking for.
Now we get to the two smart parts,
Adjust credit based on user engagement. This has 2 choices Time on Site and Page Depth. Let’s take Page depth as an example. It would translate into that a campaign that sends a user that browses 10 pages in a session versus one that browses 5, will get more credit. Time on Site can require more carefulness, as most likely your clients will not be knowing how to find time on bounce visits.
Now the second smart part Apply custom credit rules. This is where you have 100+ options to choose from to set your filtered weights. If you want to give more credit to Email, or you want to credit visits more than 5 minutes, or value clicks more than Impression. You name it you have it J What’s more you can even decide the Size of the Credit too! –> More on how to get this number correct will be taken in another blogpost on Customized Models later. Choose the best filters aligned with your onboard Marketing Strategy.
Customized Models are the Heart of Attribution Models and if you don’t evolve to align it to your Marketing strategy, then you are missing out on the real potential of Attribution. Since using Customized Models is a vast topic in itself, therefore we will have one dedicated post on how to build your own Customized Models
Wrap Up of what we Learned Today !
Attribution Modeling in GA offers 3 categories:
Almost to Totally Useless- Last/First/Linear/Last Non Direct/Last Adwords Click. DON’T use them unless someone has a gunpoint on your head
Useful – Position Based/Time Decay Attribution Models. Just playing around with default numbers, and by experimenting you can get to many interesting Insights fairly easily.
Ultimate BLISS – Customized Attribution Model: Align this with your Marketing Strategy-> Keep on evolving it -> Use it in your Daily/Weekly/Monthly KPIs.
Further Note - Apply Attribution Models to ROI, and see how dramatically numbers look different and how your viewpoint completely changes about some channels. We have put in a made up example here. Try something like this rather than just focusing on Conversions or Value only:
The Devil is in Details: Extend Modeling from Channels and dive into Campaigns. This is where Real Insights are lying. A channel might have done well, and can give you a hawk eye view. But with Attribution you want to find out within a Channel which campaigns have the Highest Impact. Merge this with Customized Model and you know exactly if the Campaigns Planning has given the impact you wanted it to.
Super duper, now we are done with Attribution Modelling. Coming up next in this Blog Series on Planning Your Budget Allocation are ‘Customer Lifetime Value’, ‘Cross Devices Browsing’, and final post that combines these aspects - ‘Budget Allocation using Attribution, CLV and Cross Devices Browsing’.
- N.B.: This post has been based on using Attribution Models in Google Analytics, and thus using GA examples. However, the approach would remain similar in other systems like Site Catalyst.