One of the most powerful aspects of digital marketing is the ability to measure results immediately. Compared to traditional media like TV, Radio, or Out-of-home advertising, digital marketing allows you to measure exactly when and where ads were seen, how they were interacted with, and whether they led to a purchase. Well, most of the time, anyways - in this article series we're going to talk about the basics of attribution and measurement, different ways it can be done, and how you can apply attribution to your campaigns.
While the ability to measure views and interaction are one of the most powerful benefits of digital marketing, they're a bit of a double-edged sword: Having the ability to see every interaction can lead to over-emphasizing the wrong interactions, or missing the big picture. How do you understand the value of your marketing or an individual ad when you have multiple campaigns in the field, and a customer may be exposed to dozens of ads in different places? If customer journeys may involve exposure to video ads, in social networks, and through search, how do you determine who gets the credit for a purchase from a customer exposed to ads in each channel?
Attribution is the key to solving this problem.
What is Attribution?
Simply put, attribution in digital marketing is a system for determining which ads get credit for a sale or other action, and how much credit those ads receive. Attribution systems can range from being very simple in the case of last-click, to very complex, using Markov models or incrementality. But even at their most complex, they're still just about understanding which ads are doing work for you and how to ensure you're assigning that credit properly so you can make better decisions when it comes to channel mix and optimization.
With the definition out of the way, it's time to look at some common systems of attribution, talk about how they work, and talk about the pros and cons of each.
Last-Click Attribution
This is the easiest, simplest form of attribution. When someone clicks on an ad, visits the site, and makes a purchase, the final ad they clicked on gets 100% of the credit.
Pros
This is the easiest form of attribution to understand, and it's very intuitive - the ad interaction (the click) can't exist without the ad, and they wouldn't be on the site without clicking. The purchase immediately follows the click, so you can rest assured that those two actions are linked. Many, if not most, of the direct response campaigns running today are basically set up using last click attribution, and it's the default measurement for any marketer looking at click-through rate as a measure of success. Many attribution systems default to last click attribution.
Cons
Last click attribution is easy, but it's also lazy. If someone is served four ads before they finally click on one, you are saying that the first three didn't matter, and potentially ignoring their cumulative effect when it comes to grabbing someone's attention. You're also putting a ton of emphasis on the click action - and clicking on ads is not a common occurrence. Whole studies have been done to show that it's a very specific kind of person who clicks on ads, suggesting that if you're only measuring clicks for success, you're basically saying that you should only be marketing to the subset of your customer base who are also people who click on ads. That's not great, given that a "good" click-through rate for a campaign is 0.1 percent!
Additionally, it's not uncommon for someone to visit a retailer website, then come back later. Maybe someone clicks on an ad, visits your site, then you bring them back later with retargeting. In those cases, you're only giving credit to the retargeting ad, and not the ad which drove the initial click. This can rapidly skew your metrics.
First-Click Attribution
On the other side of things, you can assign credit to the first ad to generate a click.
Pros
This solves your "what about my retargeting?" problem by giving full credit to the first ad someone clicks on.
Cons
This has all of the same problems as Last-Click attribution, while also shortchanging your retargeting. It's all still relying on people who click on ads as well.
First/Last Touch Attribution
In these systems, you assign credit based on ads served or seen, without requiring an interaction. If someone is served an ad on a website, then goes to your site later and makes a purchase - even without a click - that ad gets credit. This includes what's often referred to as "view-through" attribution, where viewing an ad is enough to get credit for a purchase.
As with click attribution, you can assign the credit to a different ad shown, usually the first ad or the last ad before the conversion.
Pros
The biggest value here is that you aren't dependent on clicks to measure success, and you can now assign value to ads which are merely seen. This also means you end up measuring a lot more events, which gives you more ability to optimize and learn about how your ads may be influencing customers.
Cons
The biggest problem with this system is that "view-through" part. A lot of marketers (rightfully) are skeptical about giving credit for a purchase for just showing someone an ad. After all, that person may have been likely to visit the site anyways, or make a purchase. Unlike with click-through attribution, it's harder to make the case that the conversion wouldn't have happened without the ads. If someone saw an ad two months ago and then made a purchase on my site today, do I really want to give them credit for that and say the ad was the reason it happened?
And whether you do first or last touch, you're incurring the same problems as first/last click, assigning too much credit to a particular ad in a series.
Multi-Touch Attribution
Now we're starting to get into the more complex models - and the most complex model we'll talk about today. In a multi-touch Attribution model, you assign partial credit to all of the ads somebody saw before making a purchase or converting.
There are several ways to do this, of course: In a linear attribution model, you assign equal partial credit to every ad in the chain leading up to the conversion. On the other hand, you can choose to assign credit based on positioning. The most common way to do this is to assign more value to either the early or latest ads shown to someone before their conversion. You can also assign more credits to certain types of action, like making clicks worth more than impressions. In a time decay model you assign a greater portion of credit to the most recent ads, based on some kind of half-life for earlier ads.
Pros
This is a huge step up from looking at a single event in someone's purchase journey, and is ideal when you're working with ads in multiple channels and with multiple vendors. You can also tweak the model to best suit where you feel credit should be given, helping drive optimizations down the road. You can also adjust those touchpoints later based on data.
Cons
Multi-touch attribution is very powerful, but also difficult to use effectively. Setting it up can be a challenge, particularly if your ad server isn't equipped for it, and even if you know how to set it up, you need to know how you're going to weight interactions - and your decisions there could be wrong. Those initial weights are very much arbitrary and while you can tweak them using data later, if you're too far off in your weights initial set-up you may not have good data to use later for changing them.
Next Time: Attribution 201
Attribution and measurement are critical aspects of running successful digital campaigns. But while attribution can be complicated, it's not magic - an experienced partner can help you navigate the solutions in the space and set up a custom solution in your ad server. And it's not just about knowing the digital ecosystem - different brands and products with different purchase journeys will naturally require different ways of measuring attribution. If your product is an easy one-click online sale, then last-click attribution may be the way to go, or multi-touch with a heavy emphasis on last click. If you're selling a product which would never be purchased online, such as a car, then you'll need multi-touch attribution in place or something more sophisticated. Stay tuned next time when we dive into the next level of attribution systems.
Interested in talking about or implementing a better attribution strategy? Drop us a note in our Contact form and let us know!
Comentários