Attribution Modeling & Multi-Channel Funnels With Google Analytics - Beyond Last Click

Attribution Modeling & Multi-Channel Funnels With Google Analytics

Cardinal Path articleThis post was authored by Scott Shannon, a Senior Analyst and Google Analytics Certified Seminar Leader and by Melaine Fuentes, a Sr. Consultant of Analytics and Insights, both at Cardinal Path, a premier Digital Intelligence and Optimization firm with offices throughout North America and some of the top minds in the Digital Intelligence community.

"Half the money I spend on advertising is wasted; the trouble is I don't know which half."
John Wanamaker

In the digital age, everything should be measurable, from consumer behavior to the performance of advertisements. This means transparency in advertising spending is approaching unprecedented levels, and the expectation of accountability and proven ROI is coming to the forefront of marketing departments - far different from the days of Mr. Wanamaker, in which it was acceptable for a successful capitalist to waste only half of the advertising budget.

Customer Insights & Data Informed Decision Making in Marketing

The traditional mass media advertising system is being challenged by digital advertising, which has made great strides in recent history. Our ability to measure digital channels is also unfolding, and is now seducing marketers and advertisers who believe in data informed decision making. At the center of this rapidly-changing space is a familiar business constant: the customer.

Another reason why advertisers are pursuing digital strategies is the promise of new customer insights, in real time. What color cars are people building online? Which products are piquing the interest of consumers this week? What is the geographic profile of individuals who add an item to their cart, but fail to convert? What else can we learn about these customers or potential customers? Which advertisement produced the best click through rate, and ecommerce purchase rate?

Our ability to better understand the digital experience of customers requires that we measure multiple touch points with our digital assets and piece together information from a multitude of data sources. The customer experience can include any number of the following: visits to a site, paid search ads and organic search engine results, visits directed from other sites and marketing tactics such as display impressions, email marketing and social networking. It is because of this complex consumer experience that there are many ways to calculate advertising ROI and attribute "conversion credit".

After all, a customer may be exposed to many advertisements, and may learn about your products and services on your site, and on other sites - so how do we determine which parts of the visitor experience actually influenced a given purchase?

Multi-Channel Funnels

Currently, web analytics tools like Google Analytics, CoreMetrics and Site Catalyst use a "last touch" attribution model to assign credit for conversions. What this means is the last point of interaction prior to conversion is credited for generating the online sale, or "conversion".

Since it is possible that a consumer's decision was influenced by multiple touch points prior to conversion, Google Analytics took the step of introducing Multi-Channel Funnels officially on August 24, 2011. This feature can be used to better understand the value of upstream advertising that may be catching the attention of consumers early and influence a subsequent conversion.

Whether you are a small, medium, or large enterprise (or private, public, or NGO), the chances are that you are doing more than one type of promotion – from email campaigns, to paid search (cpc), to social media and display advertising. Even offline channels like TV, radio, billboards, and print can be measured using free services like Google Analytics, by directing visitors to your website – a "drive to site" strategy.

So, which promotional effort is the most effective? How do your promotions interact with each other? Do people first search for you on Google based on a need, or do people hear about your products and services on Facebook or Twitter before visiting your site? Once people hear about your offerings do they buy immediately (e.g. Justin Bieber posters), or do they talk it over with a friend or partner before purchasing (e.g. vacation travel packages). How do they come back to your site a second and third time before making a purchase, or converting against a key goal?

Measurement across different channels (email, social, cpc, display, TV, print) has been notoriously difficult, until recently. Since Multi-Channel Funnels were introduced into the free version of Google Analytics, new power and new insights are in the hands of all marketers, both large and small. So, how easy is it to set up Multi-channel Funnels? Pretty easy, just complete the following five steps.

5 Steps to Get Started with Multi-Channel Funnels in Google Analytics

  1. Ensure your promotions are measurable (campaign links are tagged).
  2. Ensure Google Analytics is installed on your landing pages (tagging pages).
  3. Ensure you have clear calls to action and a clear sense of what you want your visitors to do on your website (what goals do you have when visitors come to your site for a given promotion?)
  4. Ensure your conversion goals are configured in Google Analytics (setting up goals in Google Analytics). You may also use Ecommerce conversions as well, but it is still a good idea to create a conversion funnel for the checkout process using Google Analytics Goals.
  5. Test the tagged links and landing pages to ensure your clicks show up in Google Analytics under Traffic Sources > Sources > Campaigns, and that your goal conversions are also registering under Conversions > Goals.

Once you have all of the above in place, launch your promotions and wait a few days for data to collect. Next, login to Google Analytics and navigate to Conversions > Multi-Channel Funnels.

Out of the box, Google Analytics will show you which promotions and channels are driving your desired business outcomes. Often this will reveal the "true value" of social media or display advertising, which are called "upper funnel" channels, because they often result in a customer's first exposure to a company, product, or service in the longer path to purchase.

Multi-Channel Funnels Venn DiagramThis initial exposure may be achieved by spreading general awareness in social media, or by posting advertising images and videos across the Internet using display advertising, and this may be the beginning of a longer path to purchase, or "Goal Conversion".

The interaction of channels can be visualized with a Venn diagram (screenshot in the right), where the overlap in the circles shows the degree to which visitors had both paid search and organic search, and where the size of each sphere indicates the volume of converting online visitors coming through each digital channel at some time prior to conversion.

Things start to get really interesting when we can see the historical touch points of our customers across our digital channels. The following image shows the pattern of visitor interactions with our digital assets (the "Channel Grouping Path"), and the number of conversions for each combination, as well as the resultant ecommerce dollar value (conversion value may be a combination of ecommerce and dollar values assigned to goals).

Multi-Channel Funnel Path Report

With a large volume of display advertising we see that ad impressions are common to many paths to purchase ("display" with an eye symbol is an ad impression, and "display" without an eye symbol is a click on a display ad that resulted in a site visit). After seeing an ad in the last 30 days, many visitors come back to the site using a bookmark or by typing the URL (direct). In addition we see returning traffic using search engines (paid and organic) and other websites that link to ours (referrals).

In order to see the total impact of each digital channel the "Assisted Conversions" report is the best, because it allows us to see what we thought was true prior to the Multi-Channel funnels reports existed and the degree to which we were wrong (screenshot below). The "Last Touch Interaction Conversion Value" column is the "original" last touch attribution model, and the columns labeled "Click Assisted Conversion Value" and "Impression Assisted Conversion Value" each reveal the history of the visitor experience, and the real influence of each channel.

In the example below we see that display advertising would have been seriously undervalued using a last click attribution model - at only 8% of the click assisted dollar value, and at less than 1% of the impression assisted dollar value (impression assisted being the maximum measurable value of display advertising in this case).

Assisted Conversions Report

Comparing the last interaction and the click assisted conversion values we can quickly see which channels were being undervalued, and also which forms of advertising tend to lead to immediate conversion. In this example, paid search has a higher last interaction value than assisted, which is important to know when determining your advertising channel mix over the course of a large marketing campaign.

Attribution Modeling In Google Analytics

To further understand how various marketing programs are working together Google Analytics provides an Attribution modeling tool that allows for building, customizing and comparing attribution models. Attribution modeling is in many ways the "Holy Grail" in fine-tuning marketing future marketing activities. This feature allows you to rapidly compare attribution models, and ultimately tailor your preferred attribution model to fit your business needs the best. There are four huge benefits to using this feature:

1. No Additional Tagging

Much like Multi-Channel Funnel reporting, attribution modeling does not require additional tagging or implementation, and is currently available with Google Analytics Premium Edition (Editor Note: This feature is now in public beta for all Google Analytics accounts, including the standard version).

Google Analytics Attribution

2. Easy To Compare

The ability to view up to three different attribution models side-by-side, and use the percent change to measure variance across channels, sources, campaigns and even specific keywords.

Attribution Models comparison

3. Customizable

The custom model builder allows credit weighting based on specified rules including site engagement and type of interaction.

Attribution Model Builder

Any of the "out of the box" attribution models can be used as a starting point for building a custom attribution model. As explained on the official Attribution help article, the default models are:

  • Last Interaction model: attributes 100% of the conversion value to the last channel with which the customer interacted before buying or converting
  • First Interaction model: attributes 100% of the conversion value to the first channel with which the customer interacted.
  • Linear model: gives equal credit to each channel interaction on the way to conversion.
  • Time Decay model: most heavily credits the touch points that occurred nearest to the time of conversion.
  • Position Based model: allows you to create a hybrid of the Last Interaction and First Interaction models. Instead of giving all the credit to either the first or last interaction, you can split the credit between them. One common scenario is to assign 40% credit each to the first interaction and last interaction, and assign 20% credit to the interactions in the middle.

The advanced options under "adjust credit for impressions", allows us to weight all display advertising impressions, and apply an even greater weight if the impression occurs X minutes in advance of an online visit that ultimately converts. This means that we decide which ad exposures will be valued more based on what we believe makes the most sense.

Advanced Attribution Models

4. "Rapid prototyping"

Custom attribution models can be created in minutes and viewed within the interface for comparison against all the other attribution models. "Rapid prototyping" is the most compelling feature of the custom model builder, because it allows us to test and compare various models and learn.

Custom Attribution Model

Attribution modeling is the Holy Grail for advertisers, because it allows us to weight different touch points visitors have with your Website according to what makes the most sense for a given business or marketing initiative.

Closing Thoughts

In the digital age everything is measurable, but it is important to "bake-in" measurement when starting to plan an advertising campaign. We simply have more options today than we had in the pre-digital age... and no longer should we accept that half our spending cannot be measured, because it can. Using systems like Google Analytics Multichannel Funnels, and attribution modeling allows us to reclaim control over our ROI, and our ability to effectively use these new measurement tools results in a competitive advantage, and greater efficiencies in marketing spending.

Learn more about Attribution models in this Tutorial Video

Subscribe To Our Newsletter For Monthly Updates



Your e-mail will be kept private

Greg | January 2013

B2B - They browse your site ten times but never fill in a form. Then they call and talk to Sales, and become a customer. In Analytics they never did anything you track as a conversion. So how would you optimize for these people?

Crucially, can marketers tell me what percent of their new customers do this. What if it's 90%?

Couldn't be 90%? Really?

Do you have any data to back that up?

So.... I'd like to read your thoughts on this.

Scott Shannon | March 2013

Hi Greg,

There ways of integrating your Google Analytics with your call center. Here is a list of tools and services that do just that:
http://www.google.com/analytics/apps/results?category=Phone%20Call%20Tra...

Also important to remeber that ANY customer interaction with your Website can be configured as a Goal in Google Analytics, so "click to contact us by phone" is a possible conversion goal you can use in Multi-Channel Funnels (MCF), and relating these offline conversions to Website sessions will allow you link together your Web Analytics data and the sales data occurring through the call center.

~Scott

Karl Kleinbach | January 2013

Greg,

Some quick thoughts that may be helpful.

What percent of leads come in via phone? Are they being tracked accurately in CRM, suggest using package to enable tracking of phone leads by source, can be very sophisticated or just as simple as using a different phone number on the web site or landing page. Article here: http://econsultancy.com/us/blog/8027-how-to-track-phone-call-leads-in-go... (also look into Twilio for this capability).

Should be able to tell what percent of leads call in with GA and web forms that capture campaign data when compared to total leads in CRM and those marked as phone leads. Though your situation is no doubt different, in my experience the percentage of new leads coming in via phone has been single digits once you exclude repeat business from existing accounts.

Consider also using a Marketing Automation platform to get insight into activity prior to conversion. Send out message to new customers with link to useful welcome pack resources etc. Once they visit the link you'll be able to marry up pre-conversion website activity with closed customers. You won't get everyone, but should be enough to develop insight into what is actually happening. MA platform will also help you sync lead data from website with data in CRM so not relying so much on data entry or sales team to enter lead data.

Put the sold customers who do click through into a identified segment so you can report on the group as a whole as to campaign etc. (either in CRM or dump to Excel).

Rambled a bit here, but I hope this is helpful.

Kevin Kerner | February 2013

Hi Greg and Karl - One thought... We've used www.logmycalls.com with some success. It's a great application that gives us really refined call statistics and it integrates with GA. Give it a try!

Great article BTW. Love the primer on attribution. Thanks!

David | February 2013

Wow that custom model builder looks seriously cool, but just wondering how much data or insight you would need to set it up accurately. I'd love to be able to dedicate enough time to get my head around that and how you could use it to it's full potential.

I would imagine some industry statistics might be helpful as should you be giving generic search traffic a lower value depending on where that phrase sat within the purchase funnel?
Google Analytics Setup (higher weighting)
Google Analytics Help (lower weighting)
Google Analytics Logo (nil weighting)

Scott Shannon | March 2013

David,

The attribution tool in Google Analytics Premium Edition (GAPE), is seriously cool (state of the art). The best feature of Google Analytics Attribution Modelling is in "flexibility", "ease of use", and "experimentability", meaning we can test various logic scenarios for weighting credit to channels very quickly and compare the results side-by-side. Skeptics would say that this flexibility will be used make agencies and departments look good, but really this puts power into the hands of the people making decisions based on this data - it is pretty easy to ask for a second opinion in terms of what the optimal attribution modelling logic is, and compare the results side-by-side.

Some of the best ways to determine weight are:

Position in the path - first touch is important, because without it there may not have been subsequent contact or a conversion! The look back window is increasing from 30 days to 90 days according to this post by Justin Cutroni at Google: http://cutroni.com/blog/2012/11/01/2012-google-analytics-summit-recap/ This increases the likelihood that "first touch" is the actual "first touch".

Branded keywords - We can change the weighting of CPC and/or organic traffic if the visitor already knew the brand of the organization or product line (this is really more like a direct visit, where a search occurred out of convenience rather than necessity...). So weigh these at less than 1X.

Time Decay Logic - If a display impression occurred a few minutes prior to a conversion (whether the ad was clicked on, or resulted in a direct visit to the site), then we might want to give display a greater credit for this, because there appears to be correlation...Perhaps 3X weight if the impression occurred within 15 minutes prior to conversion (don't forget it takes time to browse, decide, and check-out).

Direct - Credit zero if direct, because this value will be spread over the other channels.

Cost based logic - If you are spending 200K on social and 400k on CPC, then consider weighing and social touch as 2X and any CPC touch as 4X (in addition to any other recency or path position logic, you can layer the weighting logic !)

Cem Ozturk | November 2013

Really great article. I learned a lot by reading this. Google Analytics is one of the best tools I have used thus far.

Post new comment
The content of this field is kept private and will not be shown publicly.
Refresh Type the characters you see in this picture.
Type the characters you see in the picture; if you can't read them, submit the form and a new image will be generated. Not case sensitive.  Switch to audio verification.
Online Behavior © 2012