Choosing the right marketing attribution model (6 Point Guide)

Half the money I spend on advertising is wasted; the trouble is I don't know which half. - John Wanamaker

How many times has your Company CEO has told you that " Let's rework on our attribution modeling? & How many times have you worked on it? " Tons of Times! Right? This guide helps you understand the different attribution models and when & how to use them as per your business model (and how to evolve as you grow your eCommerce business).

What is "Attribution modeling" in a marketing sense?

Consider this scenario: (Digest it properly as we are going to use this to explain the models)

1. A user clicks on your Instagram ad. Explore your product inventory but does not add any product to his/her cart.

2. After almost a week, the same user sees your ad on facebook and again clicks on the ad. He then adds a product to his cart. However, he didn’t buy the product.

3. Finally, the user searches for your product on Google Search console and clicks on a text ad and buys the product (Woohoo!) 

Now, how would you credit the conversion to Instagram, Facebook and/or Google? That's where attribution to marketing channels/activities comes in to picture. Got the idea? Now let’s dive into top 6 attribution models.

1: First click model

In this model, the first user interaction takes 100% credit for the conversion (purchase on a website, sign up or whatever).So in the above example, Instagram will take 100% credit.

Pros:

 If you want to focus more on the site visits and sessions, this model may be a fit for you. This is suited if you want to focus on the number of inflows coming to your website. Lowering the cost per click and improving the CTR would be the only optimizations here.

Cons: 

As I said, this model focuses on the “number“ of people coming to your e-commerce store and not on the quality of the audience. So this model does not take into account the conversions / add to carts / sign up forms this -audience got you. 

Pro tip: Not much levers to make conversion optimization here. Not recommended for stores with over 2000 sessions/day. Only good for early traction days where you are just getting your first 1000 customers/users.

2: Last click model

In this model, the last user interaction takes 100% credit for the conversion. So in the above example, Google Search ad will take 100% credit.

Pros: This model solely focuses on conversions and hence is widely used in the industry. Best suited for e-commerce store with a large number of sessions (more than 2000/day). Use this if you have a large number of cart abandoners or product viewers and want to optimize for only conversion events.

Cons: This model will tell you the half-baked story. You are no more giving credit to the channels which are driving you session/ new users/brand awareness. Once your eCommerce store has gained traction, you might not worry about brand awareness but you should always (always!) give due attention to sessions and new user inflows. I would not recommend using only this model. Combine the first and last click model to analyze both- new sessions and conversion funnel.

Pro tip: Use Google analytics attribution tool to play with first and last click attribution for every order you received on your eCommerce store.

3: Linear model

In this model, every interaction takes equal credit for the conversion. So in the above example, Instagram, Facebook & Google search ad will take 33.33% credit each.

Pros: This model assigns equal credit to each brand touch point. This helps you analyze the user journey rather than just a single activity.

Cons: As this model gives equal credit to a click on an email newsletter and to those actually converting ones, this model fails to give due credit to the channels giving outcomes. This makes the optimization for outcome-based ones a tedious task.

Pro Tip: When you have a good user base and are getting traffic through a lot of media, this model might become a little bit tricky. Hold On, You have Time decay model to the rescue.

4: Time Decay model

In this model, recent interaction takes greater credit than the older one for the conversion. So in the above example, the credit split will be Instagram (~14.25%), Facebook(~28.5%) & Google search ad (~57%).

Pros: This takes the good part of linear attribution in terms of multi-touch points and gives due credit to the recent touch point which drove the conversion. Marketers can use this model to optimize for brand touch point that drives revenue as well as the ones which increase the likelihood of a conversion in the near future.

Cons: Consider an example — first user touch point is a free product sample sign up. In this scenario, the model will give a very little credit to the first interaction just because the user interaction came in early. This gives lower credit to the user interaction which might give a conversion in next stage.

Pro Tip: This model is the good to go strategy if you have a good amount of data and lots of conversion numbers. Use this when you have a strong understanding about what works and what doesn’t work for your brand.

5: Position Based

In this model, first and last interaction takes 80% credit & 20% is divided equally among remaining brand touch points for the conversion.So in the above example, the credit split will be Instagram (40%), Facebook(20%) & Google search ad (40%).

Pros: This helps you to give due importance to the last(40%) and first(40%) touch-points while also giving the in-between touch-points at least some share. This means you can assign significant credit to the channel that introduced your brand to the customer as well as the campaign that eventually drove them to convert.

Cons: This could easily result in two very low-value touches being given too much credit. Think about a scenario: does it make sense for a first touch Instagram story (organic)is going to get the same amount of credit as the paid search ad that resulted in a purchase?

Pro tip: Don’t even think about using this in this cross channel(Search/Social/Referral/Email/Direct) & cross device (mobile web/desktop web/ mobile app/ tablet app) marketing world.

6: Data Driven

In this model, a weighted average of all the interactions with the brand is taken into account & then the most converting path gets the highest credit.

Pros: It allows to weigh each type of user interaction and not just the interaction position in the user journey. It may be the best fit for a marketplace model such as Amazon where you may want to understand user behavior rather than just clicks and vague engagement data.

Cons: Minimum requisite to use this model? 15,000 clicks and 600 conversions in the last 30 days. For a small to medium eCommerce store, this may not be a case. I would not recommend going into the complex model when you are just starting your store.

Read More about data-driven attribution modeling Here.

Bottomline:

You know which attribution model works better for your only when you experiment enough. 

Additionally, it may change over the course of time as your eCommerce business gains traction.

It all boils down to how you think about each of the user journey marketing channel and give due credit to each one of them. Understanding user engagement level is also another important aspect of making a conclusion just on the basis of data. Remember

The right kind of data + Gut Feeling = Marketing magic 

PS: Do comment below and let us know how you are using attribution models to scale your eCommerce business?

I would love to help you out in defining right attribution for your eCommerce business.