If you're spending money on marketing at all, at some point you're going to come across the same challenge that every single company faces as they become more sophisticated with how they look at marketing spend - how can you accurately attribute credit for new customers to your marketing efforts?
Did a particular ad bring in a user, or was there some other truly organic reason for the user’s visit? Did an ad influence the use of the product? Was it a combination of multiple ads that finally drove the customer? How much money should I be spending on Social vs Search ads?
The only way to answer these questions with any confidence is to first feel good about your attribution data. If your attribution is bad you are making decisions on wrong (or at best incomplete) data, and run the chance of not only making bad decisions and putting dollars behind less effective marketing, but also paying for customers multiple times and wasting money. And, as the saying goes, you don’t want to be renting customers, you want to be buying them.
At Pickaxe we’ve found that there’s no magic bullet solution, and in fact the best way to do attribution is to make sure you have multiple methods. And the most important thing is that you fully understand the problem, because getting attribution right is a mix of the rights tools and processes being employed across the entire organization.
Before you start, are your attribution tools, conversion events, and pixels implemented correctly?
The first hurdle of attribution is accurate data collection – if you don’t have the tools accurately in place to send conversion events (visits, registrations, cart additions, purchases, free trial starts, etc) to the ad networks, then you’ll end up spending way too much on campaigns that don’t get optimized for the right signals.
Remember, all the ad networks are essentially just robots trying to get you more of the right kind of click – and to be successful, you need to ensure you’re sending the right signals back to those networks.
To read more about setting up Google Tag Manager, an attribution SDK like Kochava or AppsFlyer, or a data management platform (DMP) like Doubleclick Campaign Manager (DCM), check out our post on marketing data audits.
The first layer of attribution starts with Media Mix Modeling.
The best place is start is to get a broad understanding of how the various marketing channels are contributing to your core business goals, and to understand the impact of each channel overall. The way to do this is through Media Mix Modeling.
To create your Media Mix Model, you start by pulling in all your data from across your marketing channels (and hopefully you’ve done all the hard work of ensuring the conversion events, pixels, and other data analytics were set up correctly to make the reporting more accurate and useful).
Here’s what you data can gather, and from where:
Ad Network Reports
Facebook, Google Adwords, Doubleclick, TikTok, affiliates, and other ad networks will all send you the details of how much you spent, how many people saw the ads, how many people clicked on the ads, and the number of your new customers they think they deserve credit for acquiring (and in many cases, more than one network will claim the same user, which we’ll talk about more later!).
In the majority of cases, they won’t give you user level data, so they won’t tell you which specific customers they think they got, just how many and on which date; however, if you’ve set up conversion events and your own attribution, you can do your own work to figure out the connection between each specific customer and the marketing that drove them.
Broadcast & Above the Line
If you are doing more than just performance marketing and mixing your marketing dollars across other traditional mediums, you definitely want to incorporate those into your Media Mix Models.
Generally, your above the line advertising, including broadcast media ads and out-of-home ads, will just only give you a rough statistical estimate of how many people saw your ads, but they are still important inputs into your model.
Build Your Model
When you combine this data with your organic social media, you can get a pretty simple set of data.
With a data table like this containing a few months of data, you will have enough data to use some basic linear regression to fill out the last column of data and in turn get a general sense of the relative impact of each marketing channel.
And, of course, if you need help, there are tools out there that can gather and automate this whole process, including our very own data Insights Platform.
But why does my Media Mix Model look so different than what my attribution tools are showing? Bring in Multi-Touch Attribution to make your picture more accurate.
The problem with Media Mix Modeling is that while it’s relatively simple, it never matches what you see in the attribution tools.
For example, Facebook might claim 20k customers but you’ll only see 400 based on last-click attribution in your own data.
In our experience this is why your media mix modeling needs to look at the multi-touch attribution data as a sample, this way you can get some estimates for how many customers see ads on more than one network!
Doing this will help you better understand the critical overlap and relationships between marketing channels. For example, is your Facebook spend increasing the number of customers that your affiliate partners are claiming? If you only look at last touch, you’d have no real way of knowing.
And don't forget to account for those viral moments, content launches, or other tentpole events!
Even once you have a model that combines Media Mix and Multi-touch, you might find yourself saying, “But we launched Black Widow that weekend, of course we got tons of customers. That’s going to skew our results going forward.”
That’s when we recommend a third layer of attribution – Content Attribution.
We use our platform to help identify things like the first product or piece of content a user landed on, and how it can be attributed to the user’s conversion event.
For example, for an SVOD we look at the First Video that a user watched after signing up (or the first that they got 20 minutes into), and the number of Free Trial users who watched something and then converted.
Or for an eCommerce customer we look at the first products added to cart and the number of return transactions.
The important thing is to remember that nothing happens in a vacuum, and use smart ways to account for all those “known unknowns”.
Now that you've got all of that, normalize against the baseline!
As your last step, you have to evaluate your new marketing efforts against what would have been expected to happen normally.
For example, in our Insights Platform we show the range of expected outcomes, and can you show what the lift was against what would have been reasonable to assume would have happened without it.
However you do it, as your marketing mix becomes more complex, you need to find the smartest ways to eliminate the noise and understand what is happening with every dollar you spend.
But That's Not All
Getting smarter about Media Mix and Multi-Touch will make you better.
All the effort you’ve put into improving your attribution and modeling your marketing spend will have a wide impact on your ability to understand other relationships between the user experience and your marketing spend.
- How to use binge loyalty and sampling behavour to identity which content is increasing retention and which searches are leading to churn
- Which channels are bringing in low value customers?
- Which customers are sticking around and which ones never come back?
- Building better lookalikes!