Are you attributing your online sales or enquiries to the right marketing channels?
If you’re using Google Analytics, you’re probably inclined to say ‘yes.’ I’d be inclined to tell you there’s a substantial chance you’re not. That is because, by default, Google Analytics only tells us half of a story:
- According to Google’s data (as shared at the Google@Manchester event on 30th October 2012), around half of online sales are multi-interaction
- By default, Google Analytics attributes sales to the last non-direct click.
Attribution Modelling Tool in Google Analytics
Google recognises the importance of attribution and made assisted conversions data available to all Google Analytics users. In addition, they gave premium Google Analytics account holders exclusive access to an ‘attribution modelling tool.’ This tool made sifting through the copious assisted conversions data a walk in the park by enabling users to define an attribution model that works best for them and calculating channel values based on it (as well as comparing various attribution models side by side).
Fast forward a year and that tool is now available in the standard version of Google Analytics. Yay!
Google made the announcement at its annual Analytics Summit last week.
We’ve had early access for some of the campaigns we’re working on and have had chance to dig through the tool and the various models.
How Does the Attribution Modelling Tool Work?
When you have access to the tool, you’ll find it under conversions >> multi channel funnels >> conversion attribution tool in Google Analytics. The first time you check it out, you’ll see this:
This shows the revenue attributed to each of the marketing channels (referral, paid, organic and direct) under a last click attribution model – i.e. the value of the transactions is attributed in full to the channel that generated the last visit that led to the sale.
You can then change that to any other model or show the figures with another model selected in order that you can compare the figures side by side. Here’s what that looks like:
What this lets you see is how the various channel values stack up on different interaction models in a quick glance.
If you’re going to use this tool, I’d recommend you set up custom channels. The images above show the default channels as they already exist in Analytics. But it’s worth setting up custom channels based on your specific needs. For example, at Tecmark, we generally separate brand organic and non brand organic so we can see the value of each separately (and create attribution models that take each into account separately). Similarly, you can separate your paid search traffic and referral traffic into different channels as well.
Deciding on an Attribution Model
Every business is different and will be using different marketing channels, so an attribution model that works well for one business might not suit another. My belief is much in line with Google’s: there’s no such thing as a perfect model. But that doesn’t mean to say you can’t use attribution modelling to substantially improve your understanding of the performance of various marketing channels.
Google’s Attribution Models
Google has included several attribution models by default and has enabled users to create their own custom models or modify Google’s default ones. Here’s an overview of those various models.
With this model, 100% of the value of a sale is attributed to the final interaction with your site. This is a common model already employed by many businesses. This model could give you an insight into which channel, in multi interaction purchase journeys, has most value in terms of actually converting your user to a sale.
On the flip side, there’s no weighting given to the channel that was actually responsible for a user finding your website in the first place and it’s likely that none of the subsequent interactions (including that final ‘converting’ interaction) would have taken place were it not for the first one.
Last Non-Direct Click
This is the Google Analytics default model and the one anyone who takes Google Analytics data as is will be employing. This is much like last click except that if the last click was direct (someone typing your web address into their browser or visiting you from a bookmark saved in their browser) then it will attribute the transaction value to the interaction before this.
With first click attribution modelling, 100% of the value of a conversion or sale is attributed to the user’s first interaction with your website. This means all of the credit it given to the channel through which a user first discovered you, even if that user didn’t convert on their first visit and converted on a later visit. The channel through which a customer finds you is hugely valuable and it’s likely that sales from subsequent visits would never have happened without this first interaction.
Although there’s no denying the importance of the first interaction for those whose focus is marketing to acquire new customers, the interactions in the journey are critical too. For those whose marketing focusses on achieving more sales from existing customers, first interaction is likely to offer less value than other models.
With linear attribution modelling, the value of the conversion is split equally between all interactions along the journey. So for a 2 interaction journey, both touch points will receive 50% of the value or for a 5 interaction journey, each touch point will be credited with 20% of the conversion value. This is a decent model for those whose marketing efforts place equal emphasis on new customer attribution as on achieving sales from an existing customer set.
With time decay attribution modelling, the credit is distributed based on how close to the sale the interaction was. So, for example, interactions on the day of the sale will have more credit than interactions a week before the day of the sale.
Position based modelling is something of a cross between first interaction and last interaction modelling. It lets you place emphasis on first and last and less value on mid journey interactions. This might mean you split the credit 50/50 between the first and last or that you give first and last 40% each and split the rest between the middle interactions.
Custom Attribution Models
You’ve also got the ability within the Google Analytics Attribution Modelling tool to create custom models. One that I’ve created for one of our clients for example I’ve dubbed ‘Emphasis First Click,’ where I’ve set it so that the first click is always given twice the value of the subsequent clicks. But the custom model is where you can set something up based specifically on your business goals and what you need to know about the performance of your marketing channels.