The emergence of free and low cost analytics tools that help us truly understand our customers (and potential customers) are now widely available. However, the effective adoption of these tools is rare, and answering questions about ROI in digital marketing is still often seen as a dark art rather than a practical process
Digital Marketing ROI
In this blog I want to discuss the key issues of Attribution Modelling by showing how it allows us to move practically toward understanding the customer journey and show how this can help us move toward calculating true ROI. I also want to highlight why this isn’t as easy as it may first seem, highlight some of the key pitfalls and look at some simple options.
Defining a Conversion
Before we get going we need to define what we mean by a “Conversion”. A conversion is when one of your goals is completed. This could be an online sale, a form being filled in, a report being viewed or a file being downloaded. It is essentially somebody doing something you want them to do on your website. Setting up goals is essential to get best use of your analytics, and without it calculating ROI or using attribution modelling is impossible.
Attribution Modelling
Attribution modelling is a hot topic right now, and at first glance it looks like a quick solution to working out ROI. We purchase an Attribution system, we then track and look at each of the interactions our customers have with us online, we weight these appropriately and then see how each touchpoint contributes to our bottom line. However, the devil is in the detail with attribution modelling, because this is still, as the name suggests, modelling.
True Attribution
Another key point to remember is that most of the solutions available will generally look at online channels only. Factoring in what impact your print or TV ads have had takes things a step further, and is not an easy thing to do. It is possible, but its complex. We’ll come back to this topic at a later date.
Google Multi-Chanel Funnels
A very good step toward attribution is using Google Multi-Channel funnels. The video below is a great introduction to what multi-channel funnels actual is. I’ll also explain the key premise below.
Last Click Attribution
The most important ability of multi-channel funnels is to go beyond the last click. Analytics packages, when applied in a basic way, will generally tell you where your completed goals came from. That is, what was the traffic source that delivered the user that converted. That’s not to say they hadn’t visited your site 10 times before from various other sources. Multi-channel funnels goes beyond this and lays out the different digital channels that a user used before converting. You can start to see things like although your email campaigns might not generate much direct sales, they do contribute to a large percentage of your sales overall.
Last Click Example
The image below shows two paths to conversion. The first includes an email click through and then sometime later a search. The second just has a search that leads to conversion. Analytics that just looks at last click would see these as being the same, as the last click was the search in both cases. You could then make the mistake of thinking that the email added no value. However, without the email, the first conversion would not have happened (well, maybe! We’ll discuss that later).
The Difference Between Attribution and Multi-Channel Funnels
Attribution modelling attempts to take this one step further and understand the value of each step and then allows us to test out this value attribution. This means having some sort of weighting mechanism that attributes value and we’ll discuss this more below.
Attribution and Modelling – that’s the point
Essentially we are taking each of the touch points a customer has online, such as email, online ads, social media,etc and we then look at what order they experienced or interacted with these touch points in the lead up to the conversion (the point at which they did the thing we anted them to do, which need not be limited to a purchase). In order to correctly understand how important each touchpoint is we using a weighting mechanism, looking at things like how much did they interact, how long for and how recent was? By looking at each of these factors we are deciding how important this touchpoint is and then assigning it an appropriate percentage of the revenue it helped generate.
For example, a video viewed for 2 minutes yesterday will generally have more influence on our behaviour than an ad that we saw but didn’t interact with a week ago.
So, we get some software and we set everything up to track and give us some data. We also need to set up some rules as outlined above. If we get these rules wrong, or we bias them incorrectly in some way, any results we get will be bogus. For this reason many Attribution solution providers also offer consultancy to help work through this stage. So far so good, but probably a little more complex than we first imagined. This is when things really start to get interesting.
Fail Faster
At this point we could start blindly following what our attribution system tells us and look at the revenue it tells us our online touch points are generating. Based on this we may take a touchpoint that looks expensive but ineffective and cut it from our plans. What we will very quickly see is that this changes the values for all of the other touch points in many cases. Some that seemed very effective now don’t work at all! The reason is that each touchpoint impacts the others, and in order to do some effective with attribution modelling, the work is only actually just starting.
At this stage we need to start testing. What works and what doesn’t? How do different touchpoint interact? Whats the best mix of digital channels? How are online and offline interacting? This all takes planning and resourcing, and without this you’ve just wasted a lot of money on data you can’t really use or interpret.
Why Bother?
So looking at the complexity of attribution modelling, why bother? Fundamentally because we can suddenly achieve the holy grail of marketing, something that has never been possible before. We can actually calculate ROI, know what parts of are marketing mix are effective and improve both the bottom line and the user experience. Its hard work, but if you get there, you’ll be in a position that 99% of organisations aren’t.
Any experience of trying to carry out Attribution Modelling or calculate your ROI? Let us know. Also happy to hear from providers of analytics and attribution software on why your system is better than all the rest.