It might not be the most glamourous side of digital marketing, but attribution modelling is an essential part of understanding the roles of various channels in their customer journeys and the revenue that your business makes online. This guide looks at why digital attribution modelling is needed, the various types of model and what the differences are, and also how to make sure that your business is using the right setup for your specific requirements and objectives.
Attribution modelling refers to methods that marketers use to determine and give credit to the various channels that contribute to a user conversion or sale. People often have multiple touchpoints with a brand before making a purchase or conversion decision.
Research shows that around 92% of consumers have no conversion intent the first time they visit a brand website. So, if the first time a potential customer visits your website is via a PPC ad after a generic search for the type of product or service that you provide, but then they later return via another channel to find more information e.g. through organic search using a more specific keyword/search term before signing up to your email list, does the original PPC ad get any credit for this?
Say that this same potential customer doesn’t make a purchase for some time, but now receives your weekly email newsletters and eventually becomes a real customer through this channel. They would never have converted into a customer without both of their earlier engagements, but unless you have an effective way to attribute value to PPC and organic search for this conversion, the email gets all the credit for the revenue. This actually ends up giving you an inaccurate and inflated ROI for your email activity.
The default way of attributing sales or conversions from online marketing channels is usually the last channel that the user engaged with before they converted. However, if that single channel is given full credit for the conversion and any resulting revenue, it doesn’t give you a true reflection of the importance and role of all the other channels that the individual had an interaction with in the build-up. As marketers, this makes it much more difficult to strategise effectively across the marketing mix, and can also mean it’s harder to justify marketing spend on certain channels if the part they play in the customer journey isn’t supported by reliable data.
In a worst case scenario, if the digital marketing attribution modelling that you’re using isn’t right for your business, you will make poorly informed decisions that could be detrimental to your bottom line. For example, if you end up ignoring or de-prioritising a specific channel that actually plays an important role for some conversions then you could lose that revenue entirely as the journey no longer joins up in the right way for some customers.
Before we start to delve further into different types of attribution models, it can be useful to familiarise yourself with your current setup in Google Analytics. The Attribution section is a little hidden away under the ‘Conversions’ section in GA but, at the time of writing, it now also has its own section (beta) so can easily be found on the left sidebar in the main GA dashboard. Both routes are show below.
If you already have attribution models set up, you will be taken to your list of projects when you click on this. If not, you can choose to start a new attribution project.
There is no silver bullet when it comes to attribution modelling. There can be multiple factors that play into finding the right way to track and report on attribution, which can vary between difference services and products – so it can get really complicated, really quickly, to try and attribute absolutely everything. There will usually be a balance to find where you can use one or two models for the majority of your analysis and reporting, but supplementary models may also be needed too for you to get a clearer view of the bigger picture to shape future marketing strategy and budget requirements.
Here, we look at the most common attribution models that you can set up in Google Analytics to help you determine which will fit best for your business. You can use several different models to evaluate various funnels, if your business requires it.
Sometimes also called ‘last interaction attribution’ or ‘last-touch attribution’, this is the default way of attributing channel credit for a conversion, where 100% of the value is attributed to the final interaction that the lead had before they converted.
If you’re using the default attribution model in whatever analytics platform you use, it will usually be this that is being reported on. Whilst this model ignores all of the potential interactions that the customer had with your brand and is therefore not as useful where the purchase decision is a considered one, with a long buying cycle, this model can be fairly accurate for some types of business.
If you have a short buying cycle, or sell something that is often considered an impulse buy and there aren’t often many different touchpoints in the buyer journey, this model may be a good fit for the majority of your reporting to highlight your strongest channels.
At the other end of the journey is first click attribution, also called ‘first interaction’, which gives 100% of the credit for a conversion to the first touchpoint and interaction that the customer had. So if the original visit came from a Facebook ad, even if no conversion was made on that visit, the paid social channel gets the credit when it does result in a sale.
In a similar way to last-click attribution, this model is useful for businesses that have a buying cycle on the short side, if a fairly high percentage of conversions happen on the first visit.
Alternatively, if one of your business goals is bringing new customers to the site rather than returning visitors, reporting on the results of this model can be really useful for analysing which channels are working best for this specific objective. Ideal for new brands in the launch phase.
This model assigns 100% of the value for the conversion to the final channel interaction that wasn’t ‘direct’. A direct visit is one from a user that goes straight to your website from a bookmark or by typing your website into the address bar of their browser – they will only be able to do this if they are already familiar with your brand, so attributing to the direct channel isn’t always meaningful if you’re trying to piece together the whole buyer journey that brought them to that point.
Using a last non-direct click attribution model, you can attribute credit to the previous interaction that made them bookmark you or remember you, which can be useful for businesses that have a shortish buying cycle or a higher than average amount of conversions from direct traffic and you want to better understand what triggers that behaviour.
It won’t be so useful for businesses that have a longer buying cycle or a more considered purchase journey that involves several touchpoints with your brand before conversions usually occur.
A linear attribution model gives credit to each channel that was involved in the interactions that the customer had with your website on their way to the conversion. The credit it gives the channels is not weighted in any way but is split equally. For example, if a customer first visits the site from Pinterest, then comes back via a blog post they see on Facebook, then eventually makes a purchase for £75 following them clicking on a remarketing display ad, all of these three channels will be given credit for £25 of the revenue.
This model ensures that all channels are credited for the end result, which can give useful insight into customer journeys and show bigger picture value across channels, but the lack of weighting does mean that it won’t be clear which channels played a bigger role in the eventual revenue than others.
Time decay attribution, sometimes also called ‘time delay attribution’, is similar to linear attribution in that it gives credit to each channel where interactions occurred in the lead up to a conversion. However, rather than splitting the credit equally between channels, time delay attribution weights it by giving a higher percentage of the conversion value to the more recent interactions. The longer the time delay between interactions, the smaller the percentage of credit given by this model.
This can be useful for businesses that have a product or service with a long sales cycle, because it essentially mirrors a relationship with the lead. The most recent interactions are assumed to have the most influence on the eventual outcome of a sale or conversion.
However, just because a particular interaction happened near the start of the journey, doesn’t mean in reality that it was of lower value or influence to the eventual outcome, so it won’t always give a true reflection of channel value.
Position based attribution, which is sometimes also called U-shaped attribution, weights the attribution between channels/interactions in a slightly different way again.
It assumes that the first interaction and the final interaction before conversion are the most important in the journey, so gives each of these 40% of the credit. The remaining 20% is shared equally with any other interactions that happened between these.
In most cases, it’s a fair assumption; the first time a customer visits your site is important because it starts the relationship. The final interaction is also important because it tips the customer over the line so it’s vital that you understand how and why this happened in order to shape your future strategy.
This model is only available for GA 360 accounts that have sufficient data (at least 15k clicks and 600 conversions within 30 days to start), but it essentially uses Google machine learning to compare converting and non-converting visitors to find the most important interactions for your business. This model will apply its own percentages for credit, based on the data available.
Cynics (who, us?) might believe that the results coming from this model will always be weighted towards the more costly paid channels because that is how Google makes the majority of their revenue. However, if you have access to Google’s MCF data-driven attribution model, you can draw your own judgements based on your own audience and knowledge of your customers.
It might be that one or more of the above models fits pretty well for different parts of your business and your most important funnels, and there isn’t a real need at this stage to further break down attribution. If you’re unsure whether this is the case or not, you can seek expert advice from a digital marketing attribution modelling agency to see if there is anything important that you’re missing.
However, if you do want to delve further into attribution for a specific funnel, you can create custom attribution models in Google Analytics.
A custom attribution model allows you to choose the weigh that you give each channel touchpoint. You can base this on what you believe to be the most important touchpoints in the buyer journey of your specific customers.
In order to set up a custom attribution model accurately, you’ll need to have lots of data about the full journey of your customers, so this approach is best suited to businesses with a long and complex buying cycle, but where you have lots of information about how and why people use specific paths to sale. You can set up 10 custom attribution models in each GA reporting view and start with a default model, which you can then tweak in the ways you choose.
If you have a hypothesis about a conversion path, this can be tested with a custom attribution model. This would be related to a specific type of behaviour or interaction. For example:
If a user completes a purchase within 48 hours of clicking on a retargeting display ad, the display ad should get 60% of the credit for the sale, the final interaction (assuming it’s different) should get 20% and the remaining 10% should be attributed to the first interaction.
For many businesses, custom attribution models will be a step further than they need to go to understand and report more accurately on the ROI of their various marketing channels.
For many marketers, reporting on attribution effectively to the business stakeholders can be tricky. If those stakeholders have a great level of knowledge then it makes things easier, but when time is at a premium, you’ll want to compile and communicate the most important data relating to the business as quickly and easily as possible. Too much data can be a bad thing in this type of situation if it just muddies the water.
If you use just one attribution model across all conversion funnels then this shouldn’t be too difficult, but if you use several different models for different types of customer then things can start getting a bit complicated.
Thankfully there are solutions that will produce visual reports. For example, using Google Data Studio, especially in combination with the Supermetrics connector, can bring the information that you need into one place, which can make a big difference to effective reporting.
If you would like any more information on setting up attribution modelling that answers the right questions for your business and helps you better evolve and scale up your strategies and budget, get in touch with the team at Hitsearch for a chat.