Digital marketing analysis often revolves around ROI and ROAS. It’s necessary to understand what campaigns or ads are generating what return, but the question is, what happens when multiple channels interact to create conversions?
Here’s an example: Jeremy is scrolling through Facebook and spots an attractive ad that catches his eye. He clicks on the ad, checks it out, and checks the price with shipping by adding it to his cart. However, he then fails to checkout. Later on, he sees a Google ad for that same product, and this time he thinks, actually, I’ll go and complete that checkout.
In this situation, both Facebook and Google would count the conversion from your website, doubling-counting attribution to both channels.
You can guess how complex this might get with lots of touchpoints, and as the “advertising rule of seven” suggests, people are more likely to convert after seeing a product many times.
Marketing attribution models solve these issues by crediting different levels of success to each touchpoint. The most common and simplest model is the linear attribution model. By applying different models to your conversion data, you can find out which conversion path produces higher conversions.
Intro to marketing attribution
Marketing attribution is the process of analyzing the marketing touchpoints a consumer interacts with on their journey to purchasing or converting.
Attribution helps determine which channels impacted a customer’s decision to convert and their progress through the sales funnel.
Sometimes, a customer may encounter just one touchpoint, e.g., they’ll land on a site via organic search and purchase immediately, but in many situations, there are multiple touchpoints involved in a sales journey.
How to apply attribution
Once your marketing mix is diverse and you’re getting a fair amount of clean data from your UTMs, cookies, pixels, etc, you can begin ingesting customer journey data into tools such as Mixpanel, Smartlook, Google Analytics, Full Story, and Amplitude.
The majority of these tools make UTM tracking simple and sort data into categories so you can visualize and analyze customer journeys.
The goal is the same - creating a single truth source that defines and describes customer journeys so marketers can optimize channels alongside segmentation data that exposes different journeys for different segments.
In Google Analytics, you’ll have a selection of different attribution models (below).
By comparing models, you can see how your conversions are reweighted as the model assigns different credit to various parts of the customer journey. This enables you to locate the customer journey(s) that drive the most conversions, unlocking insights and opportunities for optimization.
Obtaining tracking data
You can get this event-driven data by using UTM parameters attached to URLs. User tracking techniques are becoming trickier for marketers due to the rise of ad blockers, GDPR, and privacy initiatives from Apple and Google, but marketers can still get mileage out of user tracking. However, marketing mix modeling (MMM) is the only future-proof way to measure marketing ROI.
UTM parameters from URLs are parsed into a Google Analytics cookie. UTM parameters enable you to ingest touchpoint data into Google Analytics, Mixpanel, or other funnel analysis platforms.
There are five generic types of UTM parameters:
UTM source identifies the entity sending traffic to the site. An obvious one is a social media channel, but the source could be pretty much anything where a URL is stored.
The medium identifies the medium or content type via which the traffic was generated, including:
- Banner ads
- CPC ads
- Emailed newsletters
- Facebook ads
This part of the Utm enables you to track the campaign by its name, which might be a nickname like SpringSale or a promo code.
UTM_term enables tracking of paid keywords.
Finally, Utm_content helps distinguish different pieces of content from the same ad. So, there may be multiple CTA links in a banner or ad, one being a buy button and one being a banner. This further aids analysis, albeit it’s not usually necessary.
Once you’re gathering touchpoint data via UTMs, you can begin measuring marketing attribution. This varies with different sales funnels, and UTMs are only effective when most touchpoints are digital and involve URLs.
In the B2B world, sales funnels are a little more oriented around calls, meetings, etc. Thus, attribution is a somewhat different beast. It may also be necessary to attribute conversions to calls and other touchpoints that can’t be measured using UTMs. Moreover, organic or direct search doesn’t require/involve UTMs, but still acts as a touchpoint in Google Analytics or other funnel analysis platforms.
However, for digital marketing, getting clean data from UTMs is the first step toward multi-touch attribution.
Attribution vs Marketing Mix Modeling (MMM)
Attribution and marketing mix modeling are similar, but attribution is a bottom-up event-driven approach to marketing modeling while MMM is top-down, focusing on analyzing inputs (e.g., channel spend and campaigns) to outputs (e.g., sales).
MMM broadly matches the spikes and dips in outputs with inputs, while also modeling for seasonality and other external factors. With a good-functioning MMM model, it’s possible to run simulations on data to predict campaign impact. On the other hand, MMM lacks the granular behavior analysis of marketing attribution - though it can obviously be combined with an attribution model.
Below is a comparison of attribution and MMM, with A/B testing offering a middle-ground for analyzing specific short-term decisions.
MMM is the more sophisticated over-arching model, whereas attribution is powered by shorter-term events and user behaviors.
One of the key differences between MMM and marketing attribution is that marketing attribution doesn’t cater to TV, radio, direct mail, and other marketing efforts. Moreover, while B2B platforms (e.g., CRMs) like Hubspot contain their own attribution features, attribution is best-suited to a smaller circle of digital marketing channels.
Finally, MMM’s focus on macro-level events enables marketers to assess marketing impact in the context of product changes and external factors like seasonality.
Single touch marketing attribution
The most common form of marketing attribution - and the default in both Google Analytics and Mixpanel - is last-touch attribution. First touch attribution is the opposite of last touch attribution.
Last touch attribution gives full credit to the last touch the lead converted from. So, if someone interacts with multiple touch points but converts from an Instagram ad, for example, last touch attribution provides that final touch point with 100% of the final sale.
First touch attribution gives full credit to the first touch in the journey, without considering the middle or last touches. First touch attribution is useful for understanding where customers entered the journey.
Last touch attribution makes sense on some levels, because it’s very simple to execute tracking-wise and places focus on converting channels. However, in multi-channel marketing, it’s likely not robust - as each channel contributes to moving individuals through the funnel. With last-touch attribution, you might be tempted to optimize and invest in your last touch at the expense of other channels, only to find that your revenue goes down. This is because the other touchpoints play a major ancillary role in the final sale.
Last touch and first touch attribution are also relevant to websites. For example, say you have a blog and a landing page - last touch attribution will attribute all conversions to the landing page. But what if many of those converting customers visit your blog first? What if the vast majority of your highest value customers spend more time on your blog? Last touch attribution may push attention away from your blog, which could prove costly.
Last touch and first touch attribution is the simplest form of attribution. However, linear attribution and other types of multi-touch attribution take things a step further and require more set up regarding tracking.
It’s important to get a clean flow of good tracking data before considering any type of multi-touch marketing attribution. This is fundamental to really understand your customer’s journeys through the sales funnel.
Where conversions occur after contact with multiple marketing touchpoints, it’s necessary to create an attribution model that respects the contribution of each channel. Failing to do so makes it difficult to calculate ROI or ROAS when dealing with longer, more complex customer journeys.
Full visibility of one’s channels and their contribution to conversion enables markets to tune marketing mix models (MMMs) and optimize the channels that matter the most. For example, you might find that some channels are excellent at bringing customers through to a store but are poor at converting them, whereas other channels are excellent at that final conversion.
Linear marketing attribution
Linear attribution is the most basic form of marketing attribution model. Take the above example that involves two touchpoints; the Facebook ad and Google ad.
Without attribution, both are given 100% of the credit - which is incorrect. Linear attribution splits the credit evenly between each touchpoint. Facebook receives 50%, and Google receives 50%.
Take a more advanced example where a customer engages with five touchpoints; Instagram, the website itself, a newsletter, YouTube ad, and Facebook ad. Here, each is assigned 20% of the credit.
Revenue is then also distributed across the linear funnel. So, for example, if the product sells for $100 then each touchpoint in the above receives $20 of the credit.
Pros of linear attribution
Invokes multiple touchpoints
Unlike last touch attribution, linear attribution does take notice of all touchpoints involved in a customer’s journey. Touchpoints include:
- Organic search
- Paid search
- Social media
- Paid social media
- Smartphone apps
- Software apps
- Email newsletters
- Email promotions
- Loyalty programs
- Promo codes
- Direct sales calls and meetings (B2B)
- eCommerce (e.g. Amazon, Etsy, and eBay)
Consider then that each of these can be further divided into sub-touchpoints. So, for example, you might be running multiple social media accounts with different campaigns, or may have separate websites for different countries or different sub-brands.
Linear marketing attribution attempts to divide conversions between touchpoints, which is useful when dealing with relatively simple multi-touch customer journeys.
Linear attribution enables marketers to assess all relevant touchpoints equally, as the model credits every touchpoint without over-emphasizing the first or last touchpoint. This is more accurate than single-touch models when dealing with journeys that involve three or more touchpoints.
Easy to set up
Linear attribution is relatively easy to set up with UTM tracking in GA or other tools. Then, once you’ve obtained a smooth transit of clean tracking data, you can start to see which touchpoints do the work under different models.
Enables you to find weak channels
By analyzing conversion paths with linear attribution, you’re creating a level playing field that emphasizes weak channels. This makes it easy to see which channel is playing a weak overall role in the conversion path.
Cons of Linear Attribution
Linear attribution doesn’t really emphasize the nuance of the customer journey. If you want to delve into what marketing channels help build awareness at the start of the conversion path or encourage purchasing at the end of the conversion path, you’ll need to use different models.
Touchpoints treated the same
Sometimes, interaction with a touchpoint is incidental or even coincidental. For example, someone might “like” a product on social media, come back to the website to read a few blogs, then purchase a product. In this situation, crediting that initial social media like equally to the website might ignore the blog's major contribution to the sale.
Other attribution models
A handful of other attribution models emphasize other stages of the sales funnel.
These are selectable in analytics platforms like Google Analytics.
1: Time decay
Time decay attribution provides more credit to touchpoints towards the end of the cycle. The logic for this is that touchpoints closer to the end should be catalyzing sales vs those at the beginning. Those touchpoints at the start are also valuable but contribute less to the overall attribution. For example, a cold call may receive a small portion of the credit, then a meeting, then a demo, and the final deal, which receives the most credit.
The U-shaped model emphasizes touch points at the start and finish of the cycle, while also distributing credit to those in the middle. The U-shaped model emphasizes touchpoints that capture early interest, and those that help close the sale.
The w-shaped model focuses on the first, last, and middle points. This emphasizes the lead creation and awareness stage, the middle of the journey, and the final touchpoint, which helps close the sale. The points in between still receive some credit, but typically only share 10% each, with the three main touchpoints receiving 30% each.
4: Full path
Full path attribution is similar to the W-shaped model, but uses four touchpoints receiving 22.5% each while dividing the remaining 10% between the remainder. Again, this is better for longer customer journeys.
The above models may seem arbitrary, and in truth, they are. However, these standard marketing attribution models enable marketers to analyze their figures when allocating different weights to each touchpoint. In reality, many sales journeys are different and benefit from a custom approach where the business/marketer builds its own formula to describe customer journeys.
If you’re using Google Analytics, you’ll be able to access data-driven attribution if you have at least 300 conversions and 3,000 ad interactions in supported networks within 30 days. This adaptive model takes website, shop visits, and conversions from Search (including Shopping), YouTube, and Display ads. Instead of arbitrarily crediting touchpoints, the data-driven model uses your data to compare the paths of those who convert with those who don’t.
Data-driven is an algorithmic way to credit channels, but you can also do this without GA by modeling your own conversion path. Most other customer analytics platforms provide their own features for analyzing journeys using AI and ML.
Summary: Marketing attribution models
Marketing attribution is a bottom-up behavior-driven strategy for discovering customer journeys from initial lead capture to conversion.
By analyzing touchpoints with different attribution models, it's possible to workout which customer journeys produced the most conversions.
It's worth noting that customer data is becoming harder to track, which is increasing the appeal of marketing mix modeling (MMM).