Attribution, simply put, is a way to assign credit for sales or leads back to the initial marketing activities that drove it.
An attribution model is a way of assigning more or less credit to a marketing campaign, content, or channel depending on where in a customer’s journey they encountered it.
To demonstrate how attribution models work, let’s look at a hypothetical customer journey.
A Sample Customer Journey
Somebody discovers your brand through an ad on Google on a Monday. They browse around briefly, then leave.
On Wednesday, they scroll through their Facebook feed and see a retargeting ad and remembered the product they were looking at and visit again.
On Friday night, they’re hanging out at home on their couch with the tablet and recall your brand name from the previous day and search on Google, this time clicking on an organic listing. While at your site, they see a product they like, but it’s out of stock. They see an “email me when this product becomes available” call to action and sign up for the email reminder.
Next week, when the product comes back in stock they get the reminder email and visit your site again from that campaign. But they want to make sure they’re getting the best deal, so they go to a competitor’s website to check the price on a similar item.
Lastly, since they’re familiar with your brand now, instead of reopening the email they got earlier, they just go directly to the site, search for the item, and buy it for $100.
Applying Attribution Models
Now that we’ve laid out that customer journey, let’s look at how credit for that sale is assigned under a few attribution models.
Last Click Attribution
Last click attribution, sometimes called last touch, is the most common model and the most problematic for two reasons:
- By the time the last touch happens, a potential customer is already educated and familiar with your brand. Which means, they come directly to your site or search for your brand name in Google.
- Almost every analytics platform uses last touch attribution by default. annual shareholder reports tend to also use last touch attribution when talking about digital results. Why? Because they’ve been that way forever and old habits die hard.
Revisiting our customer journey using last click attribution, direct gets all the credit for the sale, completely obscuring the marketing efforts that came before it.
Incredible! Our brand strength is off the charts! Why are we even paying for traffic? Or to send out these expensive email blasts? Or to work with our SEO agency? We should stop doing all of that and just rely on word-of-mouth.
For more on the last click attribution problem, check out this post.
Edit: Commenter Sam Gabell made a good point in the thread below. Technically, Google Analytics uses Last Non-Direct Touch attribution by default, which is an important distinction. Under Last Non-Direct, Email would get 100% of the credit for the customer journey in our example. However, if the fifth and final touch were somebody coming back after doing a branded Organic search, email wouldn’t get any credit in that scenario.
First Click Attribution
Under a first click (or first touch) attribution model the opposite is true; only the first stop in the user journey would get the entirety of the credit for that $100 sale. None of the other touch points leading up to that sale would be credited at all, even though they played an important role in the transaction.
So going back to our customer journey, if we paid $2 to generate that initial ad click, and the first touch attribution model tells us we generated $100 through that ad, we’d rightly think that our ROI for paid search is 50:1!
Wow! We should just shovel our entire marketing budget to Google Ads at that rate!
So why don’t we just give equal credit to all the touch points in a user journey and simplify everything? That’s what a linear model does.
We call it the “Participated Award” of attribution models. You did it, little Johnny! There are no winners and losers, but you tried really hard and that counts for something.
So everybody gets their $20 bucks, everybody’s happy, and we’re not necessarily any wiser about which channels are most effective for the business.
Then there’s u-shaped, also called position-based attribution. Now we’re starting to get somewhere intelligent for retail.
We give the first touch lots of credit, for introducing someone to our brand. Last touch gets lots of credit for closing the deal. And then we sprinkle some loose change on the comparison shopping touches in between.
But is this really what we need? Maybe we sell a really big ticket item, more than $100 and less of a potential impulse buy. In this case, comparison shopping and research touches in the middle of the funnel become more vital.
Then there’s a ramp model, which gives increasing credit to touches as they get closer to the sale. This is also called a time decay model in some attribution tools because it also factors in how long ago the first and middle touches occurred and downgrades the level of credit they get accordingly.
A ramp model is more effective for long sales cycles that require keeping people engaged with your brand over lengthy stretches of time, as well as B2B products and services that require long term commitments and contracts.
Getting Started with a Model
Just wrapping your head around what models are out there is a good first step, but the real gold is backing out to a customized attribution model based on your business and the most common customer journeys you encounter.
Google Analytics’ Model Comparison Tool allows you to apply any of the models we mentioned above on your historical data in a non-destructive way, so you can start to understand the value of considering your performance in a new light.
A word of warning though: once you get hooked on attribution, there’s no going back!插逼视频