How to Navigate Attribution Modeling in Google Analytics

By Tinuiti Team

Aug 10 2015

Attribution Modeling & Multi-Channel Tracking

An attribution model is the rule, or set of rules, that determines how credit for sales and conversions is assigned to touchpoints in conversion paths.

In ecommerce, attribution can be a complicated concept for retailers to understand. It’s imperative for retailers to know which marketing channels or campaigns are having the most impact on their customers. By dissecting the pathways that customers take to discover and make a purchase, retailers can better inform their budget and marketing strategy.

Most retailers can agree that a large portion (if not all) of their marketing budget will be devoted to their online business. In response to the vast landscape and multiple touchpoints within an online purchase such as email, social media, SEO and PPC – it’s important to understand how each of these channels relate and support each other throughout the shopper’s journey.

Defining the Multi-Channel

In Analytics, conversions and ecommerce transactions are credited to the last campaign, search, or ad that referred the user when he or she converted.

Two questions most retailers ask:

What role did prior website referrals, searches, and ads play in that conversion?
How much time passed between the user’s initial interest and his or her purchase?

The Multi-Channel Funnel reports answer these questions and others by showing how your marketing channels (i.e., sources of traffic to your website) work together to create sales and conversions.

Even though a purchase is made via a Google search, customers could have been introduced to a brand via a blog or while searching for specific products and services.

The Multi-Channel Funnel reports show how previous referrals and searches contributed to a sale.

Note: For a retailer there are a variety of “channels” (e.g. online-to-offline, computer-to-mobile, digital-channel-to-another-digital-channel). The scope with which we’re talking about in this blog article is strictly Across Multiple Digital Channels.

Attribution Modeling Example:

Below is an example to help explain how each of the attribution models apply to a conversion:

“Laura finds shoes.com by clicking one of it’s AdWords ads. She returns one week later by clicking over from Facebook. That same day, she comes back a third time via one of shoes.com’s email campaigns, and a few hours later, she returns again directly and buys a pair of sneakers.

Last Interaction attribution model: In this case, the Direct channel—would receive 100% of the credit for the sale.

Last Non-Direct Click attribution model: In this case, all direct traffic is ignored, and 100% of the credit for the sale goes to the last channel that the customer clicked through from before converting (in this example) the Email channel.

Last AdWords Click attribution model: In this case, the first and only click to the Paid Search channel —would receive 100% of the credit for the sale.

First Interaction attribution model: The first touchpoint—in this case, the Paid Search channel—would receive 100% of the credit for the sale.

Linear attribution model: Each touchpoint in the conversion path—in this case the Paid Search, Social Network, Email, and Direct channels—would share equal credit (25% each) for the sale.

Time Decay attribution model: In this case the touchpoints closest in time to the sale or conversion get most of the credit. In this particular sale, the Direct and Email channels would receive the most credit because the customer interacted with them within a few hours of conversion. The Social Network channel would receive less credit than either the Direct orEmail channels. Since the Paid Search interaction occurred one week earlier, this channel would receive significantly less credit.

Position Based attribution model: 40% credit is assigned to each the first and last interaction, and the remaining 20% credit is distributed evenly to the middle interactions. In this example, the Paid Search and Direct channels would each receive 40% credit, while the Social Network and Email channels would each receive 10% credit.

Attribution Modeling Comparison:

The Model Comparison Tool allows retailers to compare the impact of different attribution models on their marketing channels. Within the tool, the calculated Conversion Value (and the number of conversions) for each of the marketing channels will vary according to the attribution model used.

For example, a channel that predominantly initiates conversion paths will have a higher Conversion Value according to the First Interaction attribution model than it would according to the Last Interaction attribution model.

The Model Comparison Tool allows retailers to compare up to 3 different attribution models at once to measure the impact of their marketing channels.

According to Google, when evaluating the effectiveness of your channels, use attribution models that reflect your advertising goals and business models. Regardless of the model(s) you use, test your assumptions by experimenting. Increase or decrease investment in a channel as guided by the model output, then observe your results in the data.

There are a variety of tools available to determine this information but for this example we will focus on Google Analytics.

“Google Analytics is the best place to start. There is a lot of valuable data in there – but first retailers should define what they are trying to accomplish with this level of analysis, otherwise they risk getting lost in a mess of details,” David Weichel, Paid Search Director at CPC Strategy said.

Attribution Conversion Paths:

The Multi-Channel Funnel reports are generated from conversion paths, the sequences of interactions (i.e., clicks/referrals from channels) that led up to each conversion and transaction.

By default, only interactions within the last 30 days are included in conversion paths, but you can adjust this time period from 1-90 days using the Lookback Window selector at the top of each report. Conversion path data include interactions with virtually all digital channels. These channels include, but are not limited to:

paid and organic search (on all search engines along with the specific keywords searched)
referral sites
affiliates
social networks
email newsletters
custom campaigns that you’ve created, including offline campaigns that send traffic to vanity URLs

david-weichel

3 Key Factors of Attribution

Be Consistant
- Which attribution model you choose is trivial – pick one and stick to it
Use Comparisons
- The real value in MCA is comparing an alternative model to your default model
- Percent changes can say a lot about a channels influence
Create A Hypothesis
- Data may seem overwhelming but be careful to avoid analysis paralysis
- Formulate a theory – backed by data
- Form and test your hypothesis

“In the end, make sure to consistently come back to compare your models against one another to see if your hypothesis was indeed correct. It may very well have not been true, but at least now you know,” Weichel said.

Note: There are a few different types of multi-channel attribution but for this example, we’re focusing strictly on Digital Channels. This doesn’t account for online-to-store or across-devices or anything else that in reality impacts how we’re attributing value to channels.

Step 1: Start With Channel Budget Allocation

Figure out which channels are leading to the most New Customer Acquisition and invest more in those channels.

In Google Analytics, go to Conversions > Attribution > Model Comparison
Make sure you’re only looking at Ecommerce > Transaction data
Compare Last Interaction vs Position Based vs First Interaction
Choose Source / Medium as your Primary dimension

View conversion data by Last Interaction (the default in Google Analytics) would provide the example below:

Note: (direct) / (none) is high on the list.

If you sort by First Interaction conversions, you may see a completely different story (below):

(direct) / (none) turns out to have a much smaller impact in new customer acquisition than we originally thought
Valuable data is in the percent change (metrics on the far right)

You can alter the percent change metrics to calculate Conversion Value instead of Transactions, and in this case you see an even bigger jump (example: +79.62%)

For this example, the data tells you to invest more into google / cpc if you want to continue growing the business.

Step 2: Dive Deeper into Each Channel

Now you know to invest more in AdWords, but the next question is where do you increase campaign budgets or raise keyword / product bids?

Adjust your Dimensions view
You now have Campaign views that you can compare

Drill into a Campaign by clicking the blue link (This can inform precisely which ads/keywords are driving new customer acquisition and sales.)

Step 3: Beyond Search Funnels Data in AdWords

Unfortunately AdWords data doesn’t differentiate between First click and all other clicks.

So where in the funnel does it actually exist?

Is it an Introducer? Or is it a Closer?

Depending on which one, retailers should be treating it differently. According to Weichel, retailers can use Conversion Segments to further refine the data here and if they only want to look at Transactions that resulted from multiple AdWords clicks then they can do so here.

Below is an example of setting up a conversion segment to do just that:

Here retailers can see within AdWords clicks, which ones are happening First, and which ones are actually taking the credit in Analytics (and AdWords for that matter since AdWords uses a last-paid-click model as well).

“Retailers can use this to inform their analysis within AdWords, because that Brand/Trademark campaign might not be doing so well if it didn’t have all these other campaigns to first introduce users to the site.”

attribution-modeling