Crossing The Lines of Attribution

Online marketing attribution and its impact on customers

Online marketing attribution is one of the most important aspects of digital marketing. It’s the science behind assigning values to individual touchpoints throughout the customer journey. It means that marketers are getting a bird’s eye view of how advertising affects and impacts customers and revenue across each touchpoint. Cross channel attribution is attribution that tracks all channels across the entire marketing mix, regardless of their place in the funnel. All of this is to help businesses optimize their ad spend, in part, by using data to best understand the identity of their customers and what their marketing is actually doing to persuade them.

This is all accomplished through an attribution model. And an attribution model is only going to be as good as the data on which it’s based. Many marketers jump to detailed model discussions before they ensure that they’re data is sound. Some go to the other extreme, insisting on factoring in every possible touchpoint, online and off. This can result in an expensive and complex implementation that often fails to deliver actionable insights. A successful attribution model will cover all bases, tracking and weighting each touchpoint from the first ad click on one device, to the final conversion on another. This all leads to marketers becoming able to accurately evaluate their multi-channel advertising performance and be able to make more informed advertising campaign decisions. There are two main schools of thought when it comes to attribution models: Rule Based models and Statistical models.

Rule-based attribution model

As the name suggests, uses rules to credit contributors to a conversion. To be more specific, these rule based models use predetermined formulas to allocate conversion credit across marketing touchpoints. Unfortunately, they will not take any historical data into account. To better explain through analogy, it is very similar to any fad diet. These diets give a list of very specific rules assigned to specific foods while forbidding others, all through these vague basic understandings of nutrition. They don’t take into account your past nutritional history, health, genetics, etc. In a similar sense, rule based models are based on simplistic views of marketing data. These models are usually either time decay, positional, or first/last click.

The time decay attribution model in omni-channel marketing

This model is a multi-touch attribution model that gives some credit to all the channels that led to your customer converting, with that amount of credit being less, or decaying, the further back in time the channel was interacted with. The assumption here is that the first advertising channel your customer interacts with only plants the seed, and that over time their interest in actually making a purchase will grow with repeated exposure through various other marketing channels. Because of this, the way that this model assigns credit to different channels may be interpreted as a rising level of interest, as well as commitment, from the customer.

Time decay attribution is helpful to measure longer sales cycles, as time between channel interactions will serve to highlight the differences in conversion credit they receive. Timed campaigns also do well using this model as the time measurement aspect is what this model is based around. Another benefit is consistency, since the formula is fairly standardized, so fluctuations in activity are easily measurable. Using a set formula that applies the number of days prior to conversion as a key variable, each marketing channel receives credit based on the assumption that the more time that has passed since the first interaction, the closer the customer got to the actual sale. But how does it all work exactly?

Let’s say that your customer is watching a traveling show on their CTV. They see a location that piques their interest, which leads them to a quick organic search about the area. They click a few links, possibly read an article about it and move on, likely to forget about it if not reminded. A few days later, while scrolling between pictures of pets and fancy dinners on social media, they see an advertisement for cheap flights/hotels in the area they were curious about. Remembering how interesting the area looked and contemplating the last time they might have taken a vacation, but the customer still might be unsure. A little further down the line, they might be listening to a music streaming service at work and get an ad break from the tourism board of the area gloating about how everyone should visit once. At this point the customer is about to pull the trigger. They’re aware that they can afford it and by now know the website’s name by heart. They type it in and purchase the tickets. That is how time decay models work.

Positional Attribution and how its different from rule-based and time decay models

Positional attribution is a bit different. The positional model is also a multi-touch attribution model, but it gives the first and last touchpoints a specific percentage of credit. The remaining credit is then distributed evenly across all other touchpoints. The positional model usually exists within two different values: 30 percent or 40 percent. This means that either 30 or 40 percent of the credit is given to both the first and last touchpoints, with the remaining 40 or 20 percent of the credit distributed evenly among the remaining touchpoints. This approach views the first and last touchpoints as the most important aspects to driving sales and attributes more than a single interaction for the journey’s success while maintaining its simplicity.

To explain the way the positional attribution model works, take a consumer searching for new shoes. They most likely will start with a very broad search for what they’re looking for like “white sneakers”. As they research and find more options, the terms will keep getting more specific until they settle on a single brand, type in their name and make a purchase. Using the positional model, 40% of the conversions would be given to “white sneakers”, 20% would be given to everything in-between, and the remaining 40% would be given to the final decision of the specific brand. Position-based modeling gives you different and potentially more valuable insights into your campaign, as you’d be able to see and take action based on top-of-funnel keyword data. Unfortunately, this can lead to missed opportunities as there could have been altered conversions within the first and last clicks.

First & Last Click Attribution

First/last click attribution is the most popular and widely used rule based attribution model just for its simplicity alone. It attributes 100% of the credit for the sale on the first, or last, touchpoint. In the simplest terms, it’s just dividing up positional attribution, while disregarding all previous or following touchpoints. Though both these choices are quick and easy to implement, it leaves the door open to the possibility of double counting conversions. This is due to the fact that platforms like Google and Facebook don’t share data with one another, but both take credit for first/last click interactions. Working with an independent digital advertising partner (like Stirista) that integrates with both publishers and will deduplicate conversions is the easiest way to remedy this situation.

Statistical attribution models

Statistical models, again just as the name states, rely heavily on statistics and the data provided. These models are built with algorithms that use click data to define an empirically based value for each channel’s contribution towards a conversion. Statistical models can compare each individual touchpoint, allowing for a predictive approach. They can also quantify historical data and apply machine learning for more accurate results. The main type of statistical attribution model that is used is known as multi-touch attribution (MTA) analysis.

Statistical MTA is the most sophisticated, as well as accurate, attribution model. It uses historical data which is then put into algorithms that are designed that learn what each touchpoint means to a consumer, and then dynamically assign credit to each channel. To make sure the data you receive from MTA analysis is accurate, it relies on three main points: user identity/characteristics, media touchpoints, and sales/conversion data.

User identity and characteristics within MTA analysis are, at the basic level, identifying individuals with specific interests, preferences or prior relationships with the brand, as well as measuring whether certain messages/campaigns were effective. This information is generally obtained from either the marketer’s’ own sales or CRM records or external 3rd party data partners, which together provide the marketer a more informed sense of a consumer’s demographics, past purchase history with the brand, web content and browsing behavior, or other types of information that might be important for a marketer to make an informed decision about the value of the advertising opportunity. This data can be categorized, or segmented, broadly as audience data and device data.

Understanding multi-touch attribution (MTA) analysis

Media touchpoints within MTA analysis are specific data points a consumer experiences from the start of the campaign through the end. Media touchpoint data will usually show the format or type of ad (display, video, CTV), the overall cost of the ad, and whether or not the user engaged with the ad (clicked, shared, completed, hovered, etc). MTA attribution models benefit by including as much information as possible about user engagement with the media, as engagement is considered to be an important proxy for whether the user was receptive to the message. Understanding the media journey will help getting an accurate readout on different media channel’s effectiveness on how important it was to the overall journey.

Sales and conversion data are a part of MTA attribution used to measure how much ads have influenced consumer behavior. This data specifically supplies the metrics of the campaign results, otherwise known as key performance indicators (KPIs). All of this data goes together to create the optimization metric that enables marketers to take action on MTA insights by shifting budgets across or within different channels. It also allows marketers to adjust creative measures and offers within the market to increase overall performance and retention.

To better understand how these all fit into the MTA analysis process, let’s use an example of a woman named Veronica. She is a 35 year old manager of a home furnishings retail store. She has never really purchased much other than bedding from the store she works at, but has signed up for the store’s loyalty program with her phone number and personal email address. Based on her past purchases, the retailer would not assume that Veronica would have any interest in premium cookware, yet they notice that her account has viewed a few listings for pots and pans in the past few days. They also notice that Veronica has viewed the same listings from her work computer. Using this information, the retailer was able to deduce that Veronica would be a perfect candidate for a targeted advertising campaign using advanced marketing attribution methods.Ultimately, Veronica ends up seeing the ads and decides to go ahead and purchase. It’s fairly easy to see where each of these points come into play throughout this short example.

At Stirista, Identity is our Identity

Being able to run omnichannel campaigns, we’re able to help group all of this information together in one place so that it’s quick and easy to use across nearly every platform. Being an identity marketing company, there’s no need to wait or sort through mountains of legacy data when our identity graph does all of that for you. It gives you a complete view of the personal details that individualize your prospects, all the way from consumer interests, to financial detail, to professional information, and beyond. Our Visitor ID graph gives you granular visitor analytics, is able to track website traffic from online and offline campaigns, and allows for retargeting across every channel for better retention.

At Stirista, Identity is our Identity, and we have a simple mission: help marketers generate revenue with our identity-level data. As marketers ourselves, we know that one-size-fits-all solutions and decayed data don’t work, so we built our modular and real-time OMNA Identity Graph from the ground up to arm you with marketing data that actually works.