A Brief Overview Of A Complex Process
There are a few other components of digital marketing that are less understood than Attribution Modeling. In this article, I will provide a brief description of attribution modeling, its purpose in digital advertising, and the most typically supported models used today. My goal throughout this article is to demystify the process of defining attribution, while also providing guidance to start using its power in your digital marketing!
Part One: “What Is Attribution Modeling?”
Attribution Modeling is the process of identifying conversions as they happen across different mediums. In today’s digital marketing landscape, conversions do not happen linearly. The process of someone finding your website, reviewing your offerings, and then purchasing (of completing a sales funnel) happens across a multitude of channels. Each of these channels provides their own analytics data as their attribution “source of truth.” However, in the interest of showing value, these different channels often create and define metrics that are SKU-ed to their benefit. We see this conversion cycle reporting problem all the time in e-commerce websites, especially with Facebook and Google.
Facebook and Google traditionally require websites to install tracking scripts and pixels on the commerce sectors of websites. They use these scripts to track their user’s entries and exits from your website. Because they both are tracking their users and events separately, they aren’t able to correctly credit each other with conversion action as it happens on the site, creating an overlap. The most effective way to compete with this overlap in data is proper Attribution Modeling.
Attribution Modeling is best described as the timeline that happens each time your website experiences a conversion. That user has a start point (top of funnel entry) and endpoint (conversion or bounce) to every single interaction they have taken with your business. This timeline follows users across each different channel as they engage with your website from their computer, phone, or tablet.
The goal of attribution modeling is to clearly define the conversion start and endpoints, while also crediting internal steps (or touchpoints) with the appropriate conversion value. This process is definable in several different steps (or models) that are typically broken down in this grouping: last interaction, last non-direct click, last ads click, first interaction, linear, time decay, position-based, and custom attribution modeling. Each of these models allows you to apply your conversion cycle credit to different components of your sales funnel, using these different expressions. I will include explanations for each model here:
1. Last Interaction Attribution Modeling
Last interaction Modeling is simple to define. Whichever channel (organic traffic, paid traffic, social traffic, etc.) and medium (Google Organic, Google Ads, Facebook, etc.) generated the actual conversion event gets 100% of the conversion credit. Almost entirely out-dated as a model (a holdover from days before tracking users between websites), last-click gives the least conversion value to only one click. In over 90% of use cases, this model generates lower conversion efficiency than any other model. Google’s (the originator of attribution modeling) Agency Partners and internal Ads Specialists will all recommend switching this attribution model to one of the other options. Avoid it!
2. Last Non-Direct Click Attribution Modeling
Not utilized directly in conversion tracking funnels, Last Non-Direct Click Attribution allows you to remove direct (or branded) channel engagement from your conversion modeling. This helps provide more insightful reporting of the effectiveness of your website by separating out direct or brand-related traffic and relies more heavily on your advertising channels. This lets you get a clearer picture if your secondary channels are helping or hindering your marketing efforts. It also allows you to clearly check your most monetized channels' impacts on your conversions overall, removing the filter of your brand’s highly convertible traffic.
3. Last Ads Click Attribution Modeling
Exactly like it sounds, within an advertiser's ecosystem, it allows you to filter conversions by paid conversion point. This means you can source the account, campaign, ad group, and creative/keyword that was last utilized prior to the conversion metric. Used most heavily as a model in Google’s Product Suite, this model is more for internal advertising research than specific conversion funnel targeting.
4. First Interaction Attribution Modeling.
The opposite of Last Click, this Attribution Model credits all conversion value to the first interaction of an associated conversion event. This model is very rarely used, however, it is great for extremely long sales funnels were tracking the internal attribution steps typically won’t provide more value to the marketing effectiveness. The first interaction also allows you to very clearly value the initial outreach of the client more-so than their followup steps or experiences. Except in special cases like defined above, this model isn’t typically used other than for research purposes and trying to more effectively gauge your cost per acquisition values.
5. Linear Attribution Modeling
This attribution model is typically used for applications with large amounts of conversions. When you get into 10K+ conversions per month, you want to gauge the overall effectiveness of your marketing, rather than particular funnel steps. This allows you to segment your conversion revenue into funnel steps that more closely align with your ad spend. This model is not intended for typical advertisers, and by this point, most users will be better off with a custom attribution model at this scale level.
6. Time Decay Attribution Modeling
This attribution model is designed for retail verticals. Typically, retail online marketing is powered through brand identity, and this attribution type rewards more value to conversions getting completed over time. This allows advertisers to recoup more of the advertising cost based on the customer lifetime value, instead of the immediate initial conversion metric.
7. Position Based Attribution Modeling
Most effective for lead-based businesses, position-based attribution modeling applies most of the conversion value to the first/last interactions, then applies a little conversion value to the internal steps. Position Based Modeling is most generally used in lead-based digital marketing because it allows you to attribute conversion value between the top and bottom points of your conversion funnel. This allows you to view your lead generation approach including all the funnel steps utilized throughout the conversion event. This data then allows you to weigh more accurately the value of your entire marketing approach, rather than a specific channel or medium. If your business goal is to generate leads online, use this attribution model exclusively!
8. Custom Attribution Modeling
Custom attribution modeling allows you to build your conversion value in direct alignment with your business goals. This includes the specific % you want for each interaction step, different values per channel/medium, and complex multi-stage value associations. Not for the faint of heart, Custom modeling lets you really control your conversion funnel flow, and define the most valuable individual characteristics of your conversion funnels.
Part Two: “Now What?”
Now that I've explained the usage process for basic attribution modeling, I want to break down how the different models will impact your sales funnel conversion performance.
1. Professional Services
One of the most competitive industries in terms of base advertising costs are professional services. This vertical should use the Position-Based Attribution Model for their Conversion Tracking. This model attributes the conversion value to the first and last interactions primarily and allows you to target two distinct conversion events within your sales funnel. Effectively, you are reporting more targetable conversion points for each lead you generate, and this model aligns with machine learning models more effectively than other lead-based methods.
2. Ecommerce Companies
Depending on the scale of the enterprise, eCommerce companies will typically use Position-Based, Linear, or Custom Attribution Models.
A. The Position-based model works best for eCommerce stores with hybrid conversions that are split between online and offline sales. Typically, phone calls generate a larger amount of revenue than online purchases. Using Position Based Modeling helps connect digital interactions (like ad clicks) with real-world interactions (phone calls) by separating those conversion actions into 2 conversion points.
B. Linear Modeling allows you to view conversion metrics equally throughout your entire digital property. Rather than focusing on the costs associated with advertising, this model helps identify your digital brand power. It also helps companies in high growth mode more aggressively target conversions with less regard to spend.
C. Custom models are catered specifically to the business-specific sales funnel. This functionality is much more complex than the average marketer will want to undertake, and you will want to look at several use cases before attempting a setup on your own.
3. Retail Stores
Retail stores will typically track conversions with the Time Decay Modeling. Retail stores typically drive revenue by acquired customer lifetime values. This model makes repeated long term conversion events more valuable since those customers are less expensive to retain for longer-term revenue. It also helps gain more context for the time lag between promotional events, specials, or digital channel creative. content.
Part Three: The bright future of attribution modeling as the source of truth.
Attribution modeling provides a ton of digital marketing power, however most companies use models that are self advantaged. Google and Facebook use their conversion tracking engines to track their conversions on your website, not the reverse. We need to start using our website in the process of attributions, rather than relying on the platforms belonging to large companies that hold user data hostage, filtering results that generate profits rather than positive user experiences.