The ultimate goal of attribution modelling is to understand which channels are helping you achieve your business goals. Attribution provides you with the stepping stones to success – by understanding the paths your users take, you can leverage performance to increase profitability and reduce wasted spend.
Users take a number of paths en route to converting, often it is not as simple as “search keyword > clicked ad > read content > converted”. Although this can be the case, there are often multiple touch points involved in the conversion path. The conversion path report in Google Analytics is a great place to analyse these touch points. Simply head to GA > Conversions > Multi-Channel Funnels > Top Conversion Paths, and take a look at these, for example:
When we look at a different industry with a much longer decision-making process, the results are very different:
Both examples show the paths taken by users to get to a conversion. We can see that paid has played a role within a number of these paths, whether it’s the only path or one of many. When we track success in Google AdWords we need to know where the ad fits into the conversion path, allowing us to optimise and leverage our budget effectively.
When we understand how campaigns are attributing to the success of business goals, we can get an accurate picture of how the account is performing. Using the wrong attribution model runs the risk of wasted spend and missed opportunities.
So, what attribution models do you have to choose from?
In Google AdWords, there are six attribution models:
Let’s start with the default model: Last Click
The default model used when setting up a conversion in AdWords is Last Click. This model credits 100% of the conversion to the last click in the conversion path, like so:
This model favours efficiency as you will be able to see the last touch point where the user converted. Is it the right model to use? Probably not. This model significantly overvalues your branded terms and favours customer recycling, it gives a very limited scope of performance and only takes into account the last path. Take this conversion path for example:
With the Last Click model implemented, 100% of the credit will be assigned to Direct. When you look in AdWords, you’ll see the keyword has a click but no conversion, when in reality the user searched for a keyword, clicked the ad, then came back directly to the site and converted. This scenario would result in 0% of the credit being attributed to the keyword in AdWords, suggesting it isn’t a profitable keyword because it has driven up CPAs.
Based on the example above, we would look at performance by channel and assume that Direct has been driving conversions – which would be incorrect.
Last Click verdict: Not a preferred choice. It’s the default option, easy to use, and can make PPC managers look awesome. However, Last Click is often inaccurate and does not allow you to get a true value out of your campaigns.
Next up is Last Click’s best friend: First Click
This model follows the same premise as Last Click; however, all of the credit is attributed to the First Click, like such:
This model favours the first touch point and will assign 100% of the credit to the first click in the conversion. So if a new user searches for a term > clicks through on paid > returns directly > converts via a social ad, the credit will be attributed to the initial paid ad.
The problems start to surface when we look at the performance in AdWords, as based on this scenario we would see a conversion assigned to a keyword in the account and assume it is performing well, right? The reality is that the ad did not directly lead to a conversion. It did play a very important role by being the first touch point, but it was not the sole contributor in the conversion path.
In light of this, we would look at our conversion data and assume that paid traffic is performing well, but this would neglect the direct and social touch points, which will not play well for accurate reporting.
First click verdict: Not a preferred choice. This model is similar to Last Click, except the credit goes to the first click instead. This means your data will be inaccurate, and it will be difficult to find the true value of your paid campaigns.
Let’s look at the level playing field: Linear attribution
This model is where it starts to get interesting, as linear modelling attributes credit to every touch point in the conversion path. Like such:
This model favours every touch point in the conversion path, let’s take #2 in the example above: Paid Search > Direct
This will attribute 50% of the conversion to the paid ad – when you refer to your AdWords account you will see 0.5 in the conversion column. This model adopts a ‘team’ approach – if a channel assisted in the conversion process, it will be credited.
The benefit of this model is that every touch point is considered, so it creates a level playing field where you will be able to see where each and every keyword in your AdWords account attributed to a conversion.
However, this model undervalues key touch points and overvalues minor touch points. With this model implemented, where do you focus your paid search efforts? It makes it incredibly difficult to decide which keywords are performing well, which aren’t and where you should best attribute your budget. As with First and Last Click, basing your performance off a Linear attribution model will lead you to make assumptions for the paid campaigns that may not be right, restricting the output from your campaigns and potentially wasting spend.
Linear attribution verdict: If are after a level playing field, this model will do the trick. This will allow you to see where your ads play a role in every single conversion path, alongside a % of what is has attributed. This model does undervalue key touch points, and can make it tough to attribute value to your best performing campaigns, so approach with caution.
Now let’s look at the gradual incline: Time Decay attribution
This model is based on exponential decay, assigning the majority of the credit to the last touch point and a lessened value to the earlier touch points. Like such:
This model is more complex in that it uses an algorithm to delegate the credit to the correct touch points, looking at the channel closest to the conversion and working backwards. This model assumes that as the user gets closer and closer to converting, the importance of the channel will increase in line with the intent.
The benefit of this model is that we can now start optimising based on interactions. By placing the strongest weighting on the final touch point, we can assume that without this the user would not have found you at that sweet moment when they decided to convert, right? Well, not really. What about the first touch point? If that wasn’t present, then the user would never have discovered you and entered the conversion path. What’s to say that if your ad wasn’t there they wouldn’t have just clicked through organically?
We have to admit that this model does take things a step further than the previous models we have run through. However, it lacks the ability to recognise the interaction which initially hooked in that customer – which arguably has just as much importance as the last.
Time Decay verdict: If you wish to optimise for touch points closer to the conversion, this model is the one, as the touch points amalgamate the value increases all the way up to the conversion which will have the most credit. Is this model perfect? Not quite. Despite providing excellent attribution for conversion optimisation, it lacks credibility and heavily undervalues the first touch point.
The best of both worlds: Position-Based attribution
This model combines the best of Linear and Time Decay, Position-Based attribution attributes 40% to the first and last touch point and the additional 20% across every touch point in between:
Like Linear, this model takes every single touch point into account, whilst still allowing you to optimise for the first and last touch points. Using this model will allow you to optimise your campaigns and keywords based on the most pivotal points in the conversion process. Out of the box, this model really starts to get a grasp on accurately attributing value; to take this to the next level, you could customise the percentages to match your goals.
There are drawbacks, however: this model assumes that the first and last touch point deserve exactly the same weighting. To explain better, think of this scenario: a user clicks an organic listing first to arrive on your site then six interactions later converts via a paid ad – is it right to assume that the organic listing and paid ad should both receive 40%? Probably not. The ad was there at the pivotal moment and drove the user to convert, yet we still will assign the same heavy weighting to the initial organic click.
Position-Based verdict: This model is great and if we had to recommend one to use for AdWords out of the box, it would be this. If you do, make sure to consider your weightings to meet your goals. Compared to the attribution models we have run through so far, this one is by far the best – however, it can require a bit of tweaking from time to time.
The perfect scenario: Data-Driven attribution (DDA)
This model is the holy grail of attribution models for AdWords, as it gives credit for conversions based on how people search for your business and decide to become your customers. This model differs to all of the above as it uses your conversion data to calculate the actual contribution of each keyword across the conversion path, making the model completely different for each advertiser.
Ever wanted your own attribution model? Now is the time. But wait, it’s not that simple – this model has pretty hefty data requirements:
“As a general guideline, for this model to be available, an account must have at least 15,000 clicks and a conversion action must have at least 600 conversions within 30 days”
On top of this, conversion requirements have to be hit consistently every month, which are 10,000 clicks and 400 conversions – so still a considerable amount. The saving grace with this model is that it is still in beta, hence the robust data requirements. We can assume that the more and more accounts that are switched over to Data-Driven, the lower the threshold and the more likely it is to be that smaller accounts can switch over to this model.
According to Google, “DDA typically delivers more conversions at a similar cost-per-conversion than last click attribution.” This is all well and good: however in most cases it is out of reach – especially for accounts that are not recording vast levels of conversions. This model lends itself to ecommerce and big budgets, which is rather annoying (for now).
Data-Driven verdict: If you fall under the threshold, this is the one to go for in Google AdWords. The reality is, a huge percentage of AdWords accounts will not be anywhere near the monthly data requirements so you’ll have to wait until all AdWords customers can try this new model.
Do it yourself: custom attribution models
If required, you can customise your attribution models in Google Analytics. This can be done at view level in Google Analytics – just head to admin and click “Attribution Models”:
Once in here, click “+ new attribution model”:
Once here, you will be able to choose a baseline model, which will be any of the above (not including Data-Driven). You can then play around with features such as the % weighting, half-life of decay, lookback window and add in custom credit rules.
A popular way of customising attribution is to change the weightings on the position based model. Say you wanted more credit to go to the last touch point that the first, then you could do the following:
There is a drawback with customising your model: it’s not available in the AdWords interface at the moment. This model will have to be applied in Google Analytics, you’re able to add segments and custom reports to filter the paid traffic but you won’t be able to see them in the AdWords conversions dropdown:
Custom Attribution verdict: Take control of your attribution, but it won’t show in your AdWords interface. This is great for users who have a real focus on where they want to attribute value, and backing this up with custom reports in analytics can help you get a true view of performance and how to optimise your campaigns.
What happens when you change the attribution model in Google Adwords?
Will conversions drop? Will they increase? Will they change? Will my AdWords account implode?
Relax. All you have to do is head into AdWords, change the model for the correct conversion and save the settings. Once you have done this, the model will be updated and you will be tracking straight away.
Last Click: Assigns all credit to the last click the customer makes. If an ad was at the start/midway in the path it will receive no credit.
First Click: Assigns all credit to the first click the customer makes. If an ad was at the end/midway in the path, it will receive no credit.
Linear: Assigns credit equally across all touch points. If there were four interactions in a conversion path that included an ad, it would receive 0.25 of a conversion.
Time Decay: Incrementally assigns credit the closer the user gets to the conversion. The first touch point gets the least credit, and last gets the most.
Position-Based: Assigns 40% of the credit to first, 40% to last and the additional 20% across all touch points in between.
Data-Driven: Requires a minimum of 15,000 clicks and 600 conversions in 30 days. This model uses an algorithm to look at your conversion data and assign credit.
Custom: Tailor the above models (not Data-Driven) to meet your goals. However, these will not be available in the AdWords interface.
So, which attribution model should you choose in Google Adwords?
Do you have over 15,000 clicks and 600 conversions per month in your account? If so, we recommend using the Data-Driven model.
For all other accounts, we’d recommend using the Position-Based model as a starting point. As an agency, we only use Last or First Click on rare occasions, so go with Position-Based and you’ll be off to a flying start with your attribution.https://www.hallaminternet.com/contact/
Want a hand with your attribution modelling? Our team will be happy to help.