When brands first became serious about marketing, they had no way to tie it to real results. The Don Drapers of the world convinced companies that they had to advertise to build their brands. Yet, when they were asked to prove that their advertising tactics were effective, companies had to rely on Don’s smooth charm to convince them that the ads were providing results. This ultimately meant that, when it came time to cut the budget, marketing was always one of the first areas to be slashed. As much as we like the Dons of the world, it is truly hard to believe something without any data to back it up.
Marketing attribution changed all that by allowing marketers to directly tie their efforts to referrals, website traffic, search results, leads and sales conversions. It gave marketers the insights needed to understand what worked and what did not. Best of all, it gave them the ability to prove their worth when it was time for budget cuts.
However, the problem is that the attribution models currently used by many practitioners often do not tell the whole story. To truly understand customers’ journeys, you need to understand all of their actions, including how they behaved across devices. Instead, many companies focus only on the last thing the customer touched before making a purchase. This isn’t wrong, it’s just missing the bigger picture. If a customer sees a Facebook ad and then clicks through to buy your product, it does not necessarily mean the Facebook ad was the only thing that influenced that purchase. As a result, it is important to build an attribution model that tracks all of your marketing initiatives to glean insights into how they impact the business overall. So how do you create an attribution model that takes the entire customer journey into account?
Cultural Challenges Regarding Your Attribution Model
In the current corporate culture, there are a few reasons why implementing new attribution models may cause friction. First, there are any number of methodologies to create your attribution model. Last Touch, First Click, U-Shape, Starter, Player, and Closer are a few of the more common rules-based methods marketers use. Second, many organizations have used the same model for years without giving thought to change. Lastly, the resources to implement a new model may not be available. These reasons can sometimes lead to in-fighting when someone new tries to come in and shake things up with a different way of attributing marketing effectiveness.
In fact, when your team members come to rely on one particular attribution model as their bible, it can lead to a holy war of sorts. Team members will justify all sorts of actions to defend their particular “holy model.” It is extremely difficult to bring all of your team members to believe in one true attribution model. Unity may only come after a long, drawn-out battle and many conversions.
Even if you finally establish one methodology that all team members can agree on and use consistently, you still have other teams within the firm who may be using other methodologies. What it leads to is inconsistent modeling, and ultimately, no single source of data on which to base decisions regarding marketing budgets.
Avoiding Confirmation Bias in Attribution Insights
Every person on every team wants to use the attribution model that positions them in the most favorable light. This is part of what causes the conflict discussed above. But even more than that, when choosing a model to use, many people fall prey to confirmation bias. Naturally, we want to use the model that shows us what we already believe. Even if you are not on the social team, but you believe Twitter is great for building brand recognition, chances are you will choose the attribution model that proves your existing bias to be correct.
This tendency to choose an attribution model based on confirmation bias is precisely why Gary Angel argues that data science is not a science at all. More often than not, people are naturally subjective when choosing an attribution model. Not only does this lead to significant corporate in-fighting, but also takes a tool that is meant to be objective and turns it into something inherently biased.
Is Algorithmic the Holy Grail?
So, great. Attribution models could solve all our problems, if only we could get past ourselves to let them. If only there were another way — luckily, there is.
Attribution models have now turned to algorithmic modeling to examine millions of customer touchpoints. This allows marketers to view actions across platforms as well as process the overwhelming amount of data that is included in Big Data. We can tackle these huge amounts of data while simultaneously getting rid of the bias.
Algorithmic models look at both what was successful and what was unsuccessful. However, since all journeys are different and can include things like word of mouth that are difficult to measure even in the best attribution models, some data is still left to interpretation. For example, if we are examining display, we may see that one display ad was in 50 percent of the successful journeys, while another display ad was in 30 percent of the journeys, and still another was only in 10 percent of the journeys. Suddenly, if you are just looking at display, it seems clear which seems to be the most effective.
Algorithmic models can go a step further, though, and help understand which display ads work best with which audiences. In doing so, this attribution model begins to align with customer analytics. Ultimately, this type of modeling can help to understand which things are working and with whom. As a result, we can better understand how best to utilize our marketing budget and what impact every team member’s hard work has.
There is no 100% Correct Answer, But Give Peace a Chance
To ensure that choosing an attribution model does not devolve into a holy war of sorts, it is important to remember that no single model is perfect. The goal is to look at various elements to understand how effective each one is and why. Utilizing attribution models should be an objective process rather than one that is ruled by ego and a personal belief system.
Is algorithmic attribution the one true answer? No. Algorithmic attribution is a step in the right direction, but it is not for every organization. As outlined above, there can be many factors working against you from trying to implement it. The key is to make sure that everyone has an open mind and remembers that you are working toward a common goal. Run tests and compare different models to actually spot the differences in them so that you can learn from it. No one attribution model should become your bible.