Posted at Sep 17, 2019 7:12:00 AM by THAT Agency | Share
Attribution models give you information about how customers engage with your website. Some of that information is much more useful, some of it less. It's easy to make marketing attribution mistakes that lead you to overlook value. It's crucial to know which attribution model to use for your business, your marketing strategy, and your website.
Look for the signs that your attribution model isn't working for you as well as it should. If you're confused at the results of your attribution model, that confusion exists for a reason. The information that a useful attribution model gives you won't always confirm your assumptions, but it will always make something clearer. If there's confusion involved, it might be due to these signs you're using the wrong model:
1. You don't know the model well
Attribution models are popular because they can be so useful and they can reveal a great deal of information. They can be relied on to refine and streamline a strategy.
That popularity and trendiness can lead an organization into thinking that it's more important to have attribution modeling in place than to understand why you're using that model in the first place.
Why are you using that particular attribution model instead of another?
What questions do you hope that attribution model answers about your marketing strategy?
Do you understand how that attribution model can be misinterpreted or misunderstood?
If you're not able to answer these questions, then you do not have an attribution strategy in place. Like any tool, it's only useful when used in the correct way. Throwing money at it just to have one before you know why you want that particular model is likely to lead to misuse, misunderstanding, and a wasted tool.
If you buy any other type of tool, you spend your money wisely so that you know you have the best tool for the job. Why would you suddenly throw that proven logic out the window when dealing with attribution models?
2. You use Last Click Attribution exclusively
This is one of the most common marketing attribution mistakes. It might be easy to think you have all these numbers coming in from customers directly entering your URL your searching for your brand name. You must have phenomenal name recognition, right?
Yet these customers have more often made their purchasing decision thanks to prior visits where they read your content and browsed your products. They may have found these through social media, review sites, influencers, and searches hitting SEO content. They've likely re-visited through re-marketing or following a social media page.
Is that sale really due to so many customers magically knowing your brand out of nowhere? No, they know to enter your URL or search for the brand name because they've been to the site so many times before. That's what did the heavy lifting on their purchase. Yet with the information you have here, you may start over-investing in branded searches that really aren't driving or expanding your customer base.
You can tell here that the choice of which attribution model to use is obscuring a lot of incredibly useful information. The attribution model itself isn't lying – it's giving you accurate information. It's just that the scope of information being sought out isn't very useful. It can lead to bad marketing decisions based on a failed understanding of which routes are working and which aren't.
Accurate information in this case would rely on a model that gave weight to first click and midpoint visits as well. That could be a linear, positional, or time decay model.
3. Your customized model confirms all your assumptions
Every good thing that you assumed is going right with your marketing is just being confirmed. Every assumption you had is being shown in the attribution data. Your marketing decisions have never been wrong. That's convenient.
Now it could be that you were just spot on in your marketing decisions and the attribution model is proving this out. That is a real possibility. It does happen. Something else that happens is that someone customizes a model to the point of abstraction.
Most sports fans will know that you can look at statistics until you find something to like. Even if a player's other statistics are dreadful, if there's one in there you can hang your hat on and argue with people about, then that is your new favorite statistic.
We treat a lot of other things like sports these days – politics, news, budgets, our social media. When you look at your marketing with a customized attribution model and give more and more weight to the elements that are working well, then you have your new favorite statistics to argue. The thing is, doing so doesn't erase all the other statistics you just gave less value to. They're still meaningful and they can still cost you money if you don't pay attention to them.
Models with customized factors and weights can be incredibly useful when you're looking for them to answer a question. If you're just looking for them to prove a point, then chances are good that the model's not useful.
4. You're not verifying lead contacts
Let's say you email or even direct mail a number of leads. Have you ever compared this list to your new sales?
Many who receive mail or email won't bother to enter a specific code or URL. They may just search for your brand. They may click on a bookmark. There's some distrust for email marketing nowadays, so they may feel safer doing a brand search because they feel wary about clicking on an email link.
That means that none of these attributions are actually showing the usefulness of your mailing or email campaign because the attribution model itself isn't giving credit to the leads. This is where you run a match-back to identify new sales who match the leads you generated.
5. You're not segmenting your marketing with holdouts
A holdout test requires you to segment your marketing into groups. One group receives your marketing strategy as normal. Another group has an element that you'd like to test withheld. That's your holdout group.
After some time, you look at the difference in sales rates between the two groups. Does the group that received that marketing element have a higher sales rate than the group that didn't receive that element? How much higher? That can be understood as the difference that one element makes.
It's important when doing this that you're working with large enough sample sizes. Beyond sales figures alone, attribution models can quickly help you understand not just the difference in sales rates that a marketing element encourages, but also the difference in how each group chooses to engage.