Multi-Channel Attribution: Understanding Your True ROI

By | July 1, 2016 | Marketing Insights

Proper implementation and analysis of Multi-Channel or Cross-Channel Attribution models are possibly the most important factors to properly calculate and understand digital marketing Return on Investment (ROI).

Here’s what an attribution model does…

Essentially, these will attribute value to the myriad marketing touch points that lead the customer to complete a purchase or online goal. The February 2016 CMO Survey forecasts a 66% increase in spend on marketing analytics in three years — upon which 35.3% of decisions will be based.

What does this mean for marketers?

We need the ability to prove the value of our efforts and strategies. With data more readily available with each passing day, clients and decision makers will expect that marketers can provide valuable insight into the mix of marketing activities that lead to the greatest ROI.

A full analysis of Multi-Channel Attribution would include both online and offline channels, as well as attribution across multiple online devices (desktop, mobile phone and tablet), that consumers may utilize in their buying journey. While this is complex, it is not impossible. And there are many myths about Cross-Channel Attribution that lead to companies avoiding putting in the effort.

While there are many myths about Multi-Channel Attribution, the process is not without its challenges. Today we are only going to look at one piece of this puzzle, attribution modeling across digital channels and a single device. This will provide us with a starting point in understanding how attribution works and how we can work this into our overall analytics strategy.

We must keep in mind that looking at attribution solely from a digital perspective leaves certain gaps in the data, particularly when a customer interacts with us across multiple devices or when a customer begins their buying journey online but completes the conversion (purchase) in the store. You can read more about those challenges here at Multi-Channel Attribution: Definitions, Models, and a Reality Check by Avinash Kaushik.

Fortunately for us, Google helps us make significant strides in this area with its “Store Visits” metric in Adwords, which ties ad clicks to actual store visits based on the user’s Location History on his or her smartphone.

One last note before we look at some examples: The very first step in determining the proper attribution model is gaining an in-depth understanding of the business, including the products/services, the industry, the various buyer personas, the current marketing activities, and anything else that we can get our hands on.

The Many Variations of Attribution Models

Lens of Buyer's JourneyLet’s look at the many variation of attribution models through the lens of one example buyer’s journey.

This buyer began her journey doing a Google search for our services and visited our website, then read around for a bit without converting. Next, we were able to get in front of her via social media, resulting in a return journey to our website. As she continued down the buying funnel, she performed another Google search and clicked on one of our Adwords campaigns for her 3rd visit to our site. This time we targeted her via an email campaign resulting in a 4th visit to our site.

Finally, when she was ready to complete the purchase or conversion, she remembered our name and visited our website directly, completing a conversion with a goal value of $250.

Now let’s take a look at the way in which this $250 can be distributed to the various marketing channels.

 

Last-Click Attribution (Last Interaction)

A form or attribution modeling in which the last click to the website or the last channel via which a consumer interacted with us prior to the goal completion is awarded 100% of the credit for the conversion. This is one of the most simplistic forms of attribution.

Advantages: It is easy to implement, and other than the fact that we are at least beginning to think about multi-channel attribution, this form is something of the past and should not be used as a long-term solution to our analysis.

Disadvantages: Last-Click Attribution ignores the fact that the customer interacted with us at various points in their buying process. While the customer may have navigated to our website directly and made a purchase, their first interaction may have been via an organic search where they were first made aware of our brand. Then we may have received additional interactions via multiple channels. Last-Click Attribution ignores these previous interactions the consumer had with our website and awards all credit to the fact that they navigated directly to our site when they made the purchase.

Last-Click Attribution

 

Last-Non-Direct-Click Attribution

Any direct traffic to your website is ignored, and all the credit for the conversion is awarded to the last channel that the consumer clicked through prior to converting. For instance, our example demonstrated four touch points we had with the customer prior to them navigating directly to our site and converting. With Last Non-Direct Click Attribution, 100% of the credit for the conversion is awarded to the Email Campaign since the Direct visit, which actually resulted in the conversion, is ignored.

Advantages: Again, this is easy to implement and is a starting point, but last non-direct click attribution is not a long-term solution due to various channels and campaigns it ignores that assisted in the conversion.

Disadvantages: Similar to last-click attribution, this model ignores all the previous channels via which the consumer interacted with our site. It is also inappropriate to ignore direct visits to the site because that dismisses the value of creating brand recognition. We are unable to make informed decisions with this model because we attribute zero value to all the digital campaigns that ultimately lead to the conversion.

Last Non-Direct Click Attribution

 

Last-Adwords-Click Attribution

Another attribution model that awards 100% of the value to only one channel. In this case, all of the credit for a conversion is distributed to the last Adwords click.

Advantages: I can’t seem to think of any value to using this model, if we are going to take the time to change Google Analytic’s default settings, then we should use a more valuable model.

Disadvantages: Like the other attribution models, this one awards 100% of the credit for a conversion to one channel. We undervalue all of the work done by our other campaigns and overvalue our Adwords campaign, which can result in over-allocating marketing funds to Paid Search and under-allocating funds to other channels.

Last Adwords Click Attribution

 

First-Click Attribution (First Interaction)

This time, 100% of the credit for a conversion is attributed to the first click to the website. No matter how many interactions it takes for the consumer to eventually complete the conversion – all value is given to the marketing channel that first brought the user to our site even if it takes an additional 100 interactions to convert.

Advantages: No pros here either. Providing all value to the first channel entirely dismisses all other touch points and the actual interaction that resulted in the conversion.

Disadvantages: We ignore the channels/interactions that actually result in a sale or goal completion. This can lead to a misinterpretation of the data, as we aren’t assessing the value of the of the digital marketing campaigns that ultimately create the conversion. We may incorrectly allocate funds to the channels that poorly convert because we aren’t assessing the value of the channels with the highest conversions.

First Click Attribution (First Interaction)

 

Linear Attribution

The Linear Attribution Model attributes equal credit to every channel in the process. For our example, there are five touch points prior to the conversion, which means that every channel receives 20% of the credit or is attributed $50 of the $250 goal value.

Advantages: We are starting to distribute value to each of the touch points in the buyer’s journey. This means that we understand that our marketing doesn’t operate in silos, rather each channel and customer interaction is working in unison and assists in the ultimate goal of driving conversions.

Disadvantages: While Linear Attribution is better than any of the models that awards 100% of the value to one channel, not every channel should be considered equal. The problem with assuming that every channel provides equal value towards the conversion is that it provides no insights into which channels perform best at the various stages of the buyer’s journey.

For instance, our social media campaign may effectively drive people to our website who are interested in products, but those same people may be just beginning the buyer’s journey and not ready to make a purchase until a later time.

Why exactly does this matter?

With Linear Attribution, our Social Media campaign would receive the same credit for a conversion even if they are rarely the channel that actually results in the conversion — this could lead to a misinterpretation of the effectiveness of our social media campaigns. If the consumers visiting our site via social media are just becoming aware of our products or services, but are not ready to buy, then we should adjust our messaging to appeal more to buyers in the research phase versus attempting to create a conversion. Linear Attribution does not provide this insight and, in this case, would lead us to overvaluing the effectiveness of social media in creating conversions.

Linear Attribution

 

Time-Decay Attribution

The marketing channels closest to the conversion get more of the credit than those interactions that occurred further away from the conversion. This model forces us to think more about our sales cycle and the length of time we expect the typical sales cycle to be. Within Google Analytics, we must designate the half-life of decay (the default is seven days), and this exponential decay continues within the lookback window you have set within Google Analytics (the default is 30 days).

As an example, if we set the half-life to 10 days, then an interaction occurring 10 days prior to the conversion will receive ½ the credit of the interaction that occurs the day of the conversion. An interaction occurring 20 days prior will receive ¼ the credit of the interaction that occurs the day of the conversion.

Advantages: Time-Decay Attribution appears more logical than any of the models we have talked about thus far — given the fact that it assigns more value to those touch points that occurred closest to the actual conversion. This seems to make sense because the earliest touch points did not result in conversions and, thus, are probably less valuable. Whether or not this is actually true can be argued. However, this is potentially the best non-custom model available within Google Analytics and is an incredible starting point for our Attribution journey.

Disadvantages: The model assumes that previous visits are less valuable and that the further away they occurred from the conversion is a factor in their value. While this assumption seems to make sense, it isn’t necessarily true.

Think of this situation…

A consumer is researching for products like yours and finds your site via organic search — during that first visit, he or she finds the ideal product and decide to buy, but this person needs to wait 20 days until he or she saves the money. Over this 20 day period, this consumer visits the site via a social media campaign just to look at the product again and clicks on a retargeting banner to make sure the same product is still there. And finally, 20 days later, this consumer goes directly to your site and convert. In this case, it was actually the very first touch point (organic search) that was the most valuable; however, it will receive the least value under Time-Decay simply because it was furthest removed from the actual conversion.

Time Decay Attribution

 

Position-Based Attribution

This model is more highly customizable and takes aspects from both First-Click and Last-Click Attribution. With Position-Based Attribution, we can split the credit between the first touch point, the last touch point, and the middle touchpoints. The default setting in Google Analytics assigns 40% of the credit to both the first interaction and the last interaction with the remaining 20% being distributed to all of the middle interactions. This is customizable, so Google Analytics allows us to change this distribution to what we feel is appropriate.

Advantages: We are beginning to think about how each of our channels interacts with one another within the buyer’s journey, and we are assigning credit to both the converting channel and the assisting channels. The customization allowed by this model is also favorable as we can use any data we have gathered thus far and input this information into the Position-Based Model.

Disadvantages: While this out-of-the-box model is better than most of them, it still generalizes the positions. We have already identified that First-Click and Last-Click Attribution are not the ideal models; however, we are essentially combining characteristics of both in this model. And as we know – two wrongs don’t make a right.

For instance, in one buyer’s conversion path, the first interaction may have been via organic search and in another buyer’s conversion path the first interaction was social media. With a Position-Based model, both would receive 40% of the credit for that particular conversion. However, with the first buyer, the organic search may have had a much larger impact on the customer’s decision to convert than the second buyer’s social media visit did – but both would receive the same credit. This limitation is tough to overcome with any out of the box model.

Position Based Attribution

 

Data-Driven Attribution Model

Finally! We’ve reached the attribution model that we want to be utilizing – Data-Driven Attribution. Google Analytics 360 white paper on Google Attribution 360 explains that data-driven attribution models determine the importance of each touch point as data accumulates “from the bottom-up.” Data-driven attribution utilizes your data and compares conversion paths looking for patterns and the appropriate distribution of credit to each of our marketing channels.

Unlike all of the out-of-box models, Data-Driven Attribution is a constantly evolving model that will utilize actual data for appropriate distribution — not simply conjecture.

For instance, a data-driven model would help us determine which conversion path is most effective at generating conversions. Perhaps social media is more effective at the beginning of the buyer’s journey rather than near the end; the data will tell us this information, and we can adjust our budget and messaging accordingly. This allows us to most effectively calculate our ROI from each channel and allocate our budget as needed.

 

Using Multi-Channel Attribution to Calculate ROI

Now that we have implemented a customized attribution model for our business, it’s time to look at the results and utilize the ROI Calculations that we discussed in a previous post. Let’s say that based on our attribution model and goal values, we have the following data in regards to our digital marketing campaigns over the past year.

Digital Marketing Channel Attribution Goal Values Cost of Campaigns Return on Investment (ROI)
Social Media $8,000 $6,000 33%
Email Campaign $5,500 $6,000 (8%)
Paid Search $60,000 $30,000 100%
Organic Search $70,000 $15,000 367%
Direct $65,000 * *
Display $12,000 $3,000 300%
Total $220,500 $60,000 267.5%

 

*We did not add any cost or ROI data to Direct Traffic because we could attribute every other channel as having an impact on the Direct Traffic through brand recognition. Ideally the data will be able to tell us which channels most effectively create brand recognition, thus generating direct traffic. However, for purposes of this example we will exclude the cost and ROI of Direct Traffic.

We used our ROI Calculation: ROI = (Increase in Total Goal Value of Both Micro and Macro Conversions – Cost of SEO Campaign) / Cost of SEO Campaign.

Now that we know the individual ROI for each channel, we can look at our allocation of funds and determine the budget allocation that maximizes our return. We will talk about interpreting this data in a future post!