Updated on May 13, 2022 | 11 minute read
Why is attribution modelling important?
What models can I choose from?
The last-click model
The first-click model
The position-based model
The time-decay model
The linear model
The data-driven model
The right model for you - I want to...
...keep things simple
...focus on top-of-the-funnel engagement
...attract people at the moment of purchase
...convert traffic into paying customers
...run short promotional campaigns
...use the most accurate model
Attribution models are used to better understand your customer journey, and the search terms and campaigns involved in conversions. According to Google, attribution modelling is:
“The rule, or set of rules, that determines how credit for sales and conversions is assigned to touchpoints in conversion paths”.
In layman’s terms, attribution modelling informs Google Analytics and Google Ads which channel or keyword is assigned the credit for a conversion.
The clicks recorded can be any form of interaction with your digital presence. This includes engagement with your social pages, Search, Display or Shopping ads, organic searches, or links to your website.
Regardless of what brought the user to make a purchase, these touchpoints can vary in their influence to nudge the shopper towards checking out.
The models can help you understand your customers’ journeys, informing you of high performing search terms, and where you should increase your CPCs to appear for the most relevant, ready-to-purchase shoppers.
Recent research shows that consumers engage at least eight times before purchasing a product. As such, we can’t assume the final (or first) interaction is entirely responsible for the conversion, whether all of your channels worked together equally or if some were more effective than others.
Attribution modelling uses real data to provide a clear understanding of how different channels help to convert given prospects.
A last-click model attributes the credit for the conversion to the last click made by the user in their entire journey leading to their purchase. It’s also the default model in Google Analytics, meaning this is the model you will currently be using unless you’ve already made changes.
The last-click model is simple to use.
However, the same simplicity can be a disadvantage. With little information, the last-click model fails to provide a full picture, attributing all credit to the final click which resulted in a purchase.
It’s often the case that brands are discovered by an unbranded search term or even a number of searches before they visit your website and checkout. Using this model would suggest all your conversions are the result of branded searches or direct web traffic.
In this instance, it would appear that your ads or social channels don’t perform as well, as engagements building brand awareness have been omitted in the ultimate attribution.
A first-click model attributes the credit for the conversion to the first click made by the user in their entire journey leading to their purchase. Here’s a scenario:
With a first-click model, that first click would be attributed the credit for the conversion, suggesting the first keyword demonstrated high purchase intent, which isn’t accurate in this example.
The first-click model demonstrates how your customers came across your product or website.
Like the last-click model, it doesn’t give a full picture of the customer journey, highlighting the initial click as the most important which might not always be the case.
A hybrid of the previous two, this model gives 40% of the credit to both the first and last ad clicks, spreading the remainder over the other interactions in the customer path to purchase.
This model illustrates the full journey from initial click right through to the purchase.
Yet, dividing just 20% of the credit over any interaction between the first and final click means some may be undervalued. Similarly, the initial and final stages may be assigned more credit.
Use this model if you most value the touchpoints introducing customers to your brand and the final touchpoints which result in sales.
The time-decay model assigns more credit to ad interactions occurring closer to the conversion.
The credit is distributed using a seven day half-life, meaning that interaction with an ad eight days before the conversion is assigned half as much credit as an interaction occurring one day before the customer makes their purchase.
This model puts an emphasis on interactions closest to the conversion, demonstrating search terms and keywords used when your customer is ready to purchase.
However, the spread of conversion credit may not represent an accurate weighting across each click, and may not demonstrate the bigger picture.
This model distributes the credit for all conversions equally across all interactions in the path-to-purchase. Linear modelling can provide an idea of which keywords and channels work and which don’t.
If your ad serves for a keyword and isn't clicked, it simply won’t receive any credit. This model is useful if the aim of your campaigns is to maintain awareness with your customers, gradually moving them through the sales funnel.
With the linear model, retailers can see each step in the path.
However, by assigning an equal value to each stage, the data may be skewed and some steps may be attributed with too low or too high a value.
Using a data-driven attribution model enables retailers to draw on historical ad clicks and conversion data. The credit is distributed by calculating the actual contribution of each interaction across the path.
By comparing the paths of customers who convert to the paths of customers who don’t, the model identifies patterns among those ad interactions which ultimately result in a conversion. The converting paths may have specific stages with a higher probability of leading the customer to make a purchase.
These more valuable ad interactions are then attributed with higher credit, demonstrating to retailers the purchase intent of customers using the keywords associated with that ad or touchpoint.
A data-driven approach is naturally the most accurate attribution model available to retailers.
Unfortunately, this model is only suitable for retailers with a high volume of site traffic and conversions.
To use this model, Google states you must:
If you don’t have enough data, you won’t see the option to use data-driven attribution., and it will revert back to the last-click model.
Similarly, if your data drops below 2,000 ad interactions or below 200 conversions for the conversion action in 30 days, you will no longer be able to continue using the model.
The last-click model is great for retailers looking for simplicity. The model remains useful to communicate to marketers the keywords and channels nudging shoppers to checkout, so we can better understand how popular or successful certain ads are performing.
The first-click model may be the most suitable approach. As this is the stage where your target audience engages with your brand for the first time, the first-click model helps in understanding how your audience found you, and what strategies, keywords and campaigns work for you.
If your business is primarily transactional, and your sales cycle doesn’t involve a consideration phase, then the last-click model is a suitable approach for you.
Since the linear model provides an idea of the keywords and channels receiving clicks in the path to purchase, you can garner insights into the search terms used at different levels of purchase intent.
Understanding which search terms denote low/high levels of purchase intent helps to optimise your bid values on Google Shopping, to reduce ad spend and deploy more budget to high intent keywords.
The time-decay model is best for short 1-2 day promotion campaigns, as you can assign more credit to interactions during the short promotional period. This means that the clicks with larger values are more closely related to the campaign you’re running.
The data-driven model uses historical data to calculate how each step is credited. The information collected by the data-driven model can be used to inform your CPCs if you’re already benefiting from an automated bidding strategy on Google Shopping.
Ultimately, your attribution model should inform you of high-intent keywords to target and help to optimise your CPC bids on Google Shopping.
We hope our guide has been helpful in understanding attribution models for Google ads and in finding the right model for you and your business.
Are we a perfect match? Book a call to chat about your Google Shopping strategy and Bidnamic’s technology with one of our Google Shopping experts today.
Or learn more...
Karman is a Data Analyst at Bidnamic. She works alongside clients to optimise their Google Ads accounts. She also plays a major role in the transition phase, ensuring a smooth experience and continued success.