Understanding Google Analytics Multi Channel Funnels

Understanding Google Analytics Multi Channel Funnels

Last month I had the pleasure of participating in a Google+ thread started by Daniel Waisberg to discuss the usefulness of Multi Channel Funnels (MCF) in Google Analytics. While I agree with my colleagues who said that the path analysis report like the one shown in the post are indeed not actionable (by the way, that picture represents the path used by a single user before converting!), I do feel that MCFs are a great addition to Google Analytics and have found valuable insights for clients. In this post I will share a few of my favorite conversion segments and MCF analysis methods, as well as discuss what MCFs add (and don't add) to GA.

Understanding Multi-Channel Funnels

If you are new to Google Analytics, or to this feature in particular, here is a video created by Google that explains the idea and value behind it.

Defining Multi-Channel Funnels attribution reporting

It is important to clarify how MCFs differ in their attribution reporting from standard Google Analytics reports. (Hat tip to Adrian Vender for pointing out this help page which does a good job explaining things). The most important difference is in the way that MCFs deal with Direct traffic.

In standard Google Analytics reports, if a user visits a site via some sort of campaign (i.e. thereby setting the utmz campaign cookie with values such as source, medium keyword, etc) and then returns to the site by typing in the URL to the address bar or a bookmark, the subsequent visits are reported as additional visits from whatever source/medium/keyword was last recorded. *Multi Channel Funnels do not take the campaign cookie into account when reporting direct traffic.* The impact will oftentimes be seen as larger 'last interaction' conversion values for direct traffic in MCF than the corresponding 'Direct' traffic source in other GA reports.

Direct Traffic Attribution on Google Analytics

Personally, I find this departure from the standard way that Google Analytics (GA) handles last click attribution a bit challenging. My preference would be to see the Last Interaction Conversions remain consistent with other GA reports. Indeed, I tend to look at MCFs in order to understand how areas with large spend (in general PPC and SEO) are giving me a greater bang for my buck than what I'm seeing in standard GA reports. At the very least, the ability to toggle between the MCF model and the standard cookie based model would be helpful. For now, some deep dive analytics is simply what is called for as long as this fly is in the ointment.

Interestingly, for ecommerce transactions the normally not so useful Time to Purchase report (Standard Reporting => Conversions => Ecommerce => Time to Purchase) provides some new insights in light of Multi-Channel Funnels. Normally, I call the Time to Purchase report "not so useful" because it only shows count of visits or days to purchase from the moment the utmz campaign cookie was set, not taking multi-touch attribution into account. However, noting how Direct Traffic is reported in MCFs, we can see how many conversions would actually be reported in standard reports as Last Touch.

Time to Purchase report on Google Analytics

Analyzing Data from Multi Channel Funnels

The discrepancy between the way standard reports and MCFs deal with Direct Traffic notwithstanding, I have found the following techniques allow me to clearly identify areas where clients are getting a bigger return on investment on inbound marketing spend than we would have otherwise known about without MCFs. By far, I find the most useful report the Assisted Conversions report.

Branded and Non-Branded Keywords

In the same way that segmenting between branded and non-branded keywords is a must in regular GA reports, it is also a must for MCFs. The beautiful thing about MCFs is the ability to see (and quantify) what web analysts have known all along - that other touch points often proceed branded searches. One of the first things I do when I begin working with a new profile is to create a copy of the basic channel groupings....

Google Analytics Basic Channel Grouping

.... and then separate out branded keywords from Paid and Organic Search.

Non branded search segment

I then create a new grouping to include just branded keywords. The Custom Channel Grouping ends up looking somethings like this:

Multi-Channel Funnel Grouping

Getting Actionable Data

Once Channel Groupings are configured to match one's particular website and tracking needs (noting that I always feel branded keywords should be separated out), we can begin to apply conversion segments to the Assisted Conversions report. One of my favorite conversion segments is the following:

PPC Conversion assists report

By applying this segment, we're able to see in quantifiable numbers the impact of paid search spend had on conversions in other channels. Noting my comments above about how MCFs treat direct traffic, I pay more attention to the non-direct channels that receive last touch conversions. The following is looking at the basic channel groupings.

PPC assisted Sales

I love the fact that I can see the impact of Paid Search on other channels. The actionable takeaway in most cases has to do with bid management. For ecommerce retailers, I rely heavily on the Adwords ROI calculation to determine if a paid search campaign is maximizing profits (keeping the client's margins in mind as well, of course). With MCFs, when I can see for a fact that some Adwords clicks are indeed impacting sales down the road or not, I am able to make a much more informed decision about whether or not it makes sense to boost bids (as I'm sure that many readers out there will nod their heads in agreement that the paid search marketplace can be quite competitive). For advertisers doing retargeting, MCFs expose how users got to the site before the retargeting sealed the deal.

To see how often my paid search campaigns also include a branded search in the users path towards conversion, I apply the following filter in the Top Conversion Paths report on top of a custom channel grouping.

Branded and Paid Traffic segment

I don't spend time looking at the rows upon rows of path information itself as it is not actionable. That bears repeating, as I've seen too many clients bang their heads against a wall trying to extract meaning from the linear path data in the Top Conversion Paths report. Don't waste your time trying to find insights in row upon row of path data. I do, however, take a look at the sum of the transactions that had both a paid and branded interaction to see what extent my super-converting branded keywords were or were not influenced by paid search campaigns.

In a similar vein, I recommend using conversion segments to determine if particular keyword segments or display advertising campaigns etc are having an upper funnel effect or not. Conversion segments are very extensible and are useful in confirming whether or not poorly converting traffic sources are indeed a waste of time and resources.

Challenges with Multi-Channel Funnels

I've already mentioned a few challenges people face with Multi Channel Funnels (namely, wasting time trying to unpack path reports and the discrepancy between Last Interaction Conversions and the Last Touch attribution model in standard GA reports).

Another significant issue, in my humble opinion, is that the look back window is only 30 days. For sites with a longer sales cycle, 30 days just doesn't cut it. For those businesses, passing an anonymous visitor ID into a custom variable is a necessary prerequisite for an analyst to do his or her own multi touch attribution analysis via the API or through a CRM integration. Of course, Multi-Channel Funnels do not currently support attribution modeling, though it is something on the Google Analytics Premium roadmap.

Lastly, MCFs have a lot room for improvement when it comes to the options available in conversion segments. While the Adwords options are quite robust, I would love to create segments by events, custom variables, page paths, and landing pages. It would be great to see if a user took a certain action on a site during a visit (for example a video view, or email signup) and then converted a few visits later down the road.

Currently, Google Analytics is not designed for visitor segmentation (advanced segments are visit level). The addition of Multi Channel Funnels was a major upgrade in this area, though there is still a long way to go. With the almost dizzying pace that the GA team has been improving their product over the past few years, I would not be surprised to see another "game changing" product update in the next year.

Concluding thoughts

The addition of Multi Channel Funnels to Google Analytics just over half a year ago was a huge improvement to the product. While MCFs do have room for improvement and can be confusing and difficult to use at times, I believe they have tremendous utility. In a web ecosystem where advertisers and marketers large and small are spending real money to try and get ahead, Multi Channel Funnels can provide meaningful and actionable insights when used properly.

Subscribe To Our Newsletter For Monthly Updates



Your e-mail will be kept private

Jonathan yunger | March 2012

Anything that helps us move from metrics to actionable items is helpful. Thanks for helping make MCFs more usable.

Anna Lewis | March 2012

Hi Yehoshua, love this post, it's great to see posts that don't just say create a brand non brand grouping and leave you to work out what the numbers mean yourself!

One question though, if you apply the conversion segment excluding last interaction for a medium, would you be able to add the conversion value of this to the conversion value found in the normal reports to see the full value of this medium? The reasoning behind this would be that you then see the assists and normal lasts, but the discrepancy might be that some conversions count as assists in MCFs but are lasts in normal reporting, as the last might have been direct.

Hopefully you understand what I'm trying to explain, basically, I'm looking for the full conversion value that a source has given but not sure if and how I can get this.

Thanks,
Anna.

Yehoshua Coren | March 2012

Hi Anna,

I'm pretty sure I understand your question. :) I think that creating the conversion segment you mentioned and then trying to add the assist value to the last touch numbers as reported in the standard reports won't really be effective. The main issue, as you mentioned in your question, is the discrepancy between how MCFs deal with direct traffic compared to the standard reports.

To get the "full conversion value" for a particular source of traffic as you describe it, you can simply add the value of the assist conversions to the last interaction conversions in MCFs. If you were to then subtract the conversion value from the standard reports from that number, you would know how much "assist revenue" the traffic source is actually creating if MCFs indeed treated direct traffic the same way as the standard reports.

Something to note, the "full conversion value" that you're trying to calculate assumes that every click / visit to the the site that led to a conversion receives full attribution. I wouldn't use that sort of model when doing attribution analysis as I believe it inaccurately gives too much credit to each touch point.

Thoughts?

Yehoshua

DEP Ecommerce Consultants | March 2012

Hey Yehoshua,

Can you give more of break down on how you are filtering out the branded searches? Typically when I filter out branded searches I'll put the actual search term in the filter. Example

exclude keyword containing x

exclude keyword containing y

Your example shows

exclude keyword matching regex branded keyword

However when I do both I'm not getting the branded metrics filtered out.

Yehoshua Coren | March 2012

@DEP Ecommerce Consultants -

You are correct. To exclude branded keywords, use regex like you mentioned in your comment. In the example in my post, I was just using 'branded keyword' as a placeholder. The branded keyword itself should be specific to each site.

For others reading these comments, I'm glad that you mentioned using regex. Regex is definitely the way to go since there can be so many variant spellings (and misspellings) of a company's branded terms.

Gerard Rathenau | April 2012

Thanks for sharing your knowledge. One question:

Do you advise to use conversion segments 'days' and 'touchpoints' before conversion to optimize Retargeting campaigns?

Matt Byrne | May 2012

Brilliant insights however, I do have a question/request?

I have seen a number of posts that talk about making sure not to use just the standard channel breakdown as defined by Google. Because of this I have therefore split out all of my channels including making PPC and SEO be split by brand and non-brand terms.

The problem comes when trying to work out how to use segmentation in addition to your actionable insights methodology (given the SEO and PPC split mentioned above) to determine the correct (purely assisted) orders/transactions as the way in which the segments are structured is extremely unclear (with the and/or statements potentially being construed in different manners.

I believe that I have managed to determine the purely assisted conversion value of PPC (brand and non brand) as within the segmentation there is an Adwords keyword metric), however, trying to achieve this purely for non-brand SEO terms is much harder (especially as our brand has two words that I'd like to associate as meaning brand to us - either or both of them needing to feature)

Would you have any advice/recommendations on how you might set such a segment up to the assisted conversions report?

Many thanks

Matt

Yehoshua Coren | May 2012

Hi Matt -

I'm not sure if I fully understand your question. Could you please elaborate? In what way(s) would creating branded vs. non-branded keyword segments be different if the medium was paid search vs. organic search?

Thanks,

Yehoshua

Matt Byrne | May 2012

Yehoshua
What I am trying to ask for assistance on is some guidance on how you would set up segmentation to be able to determine what "brand" SEO terms bring to the party. i.e. what orders they only assist in. my brand terms would include "scot" and/or "stow" in them.

Secondly, I would want to know non brand terms i.e. any natural search that did not include "scot" and/or "stow" in them.

This is all related to the assisted conversion report and your actionable takeaways part of the post.

Many Thanks

Matt

Yehoshua Coren | May 2012

Hey Matt,

What I recommend doing is first excluding your branded organic keywords from the branded organic keywords in a custom channel grouping. In your case,

Non-branded organic:

Include > Medium > Containing > Organic
Exclude > Keyword > Matching RegExp > scot|stow

Branded Organic:

Include > Medium > Containing > Organic
Include > Keyword > Matching RegExp > scot|stow

Then, when looking at the Top Conversion Paths report, filter by Channel Grouping Path.

To see where branded keywords are assisting, choose:
Include > Channel Grouping Path > Matching RegExp > ^branded organic

To see how much non-branded keywords are assisting in conversions choose:
Include > Channel Grouping Path > Containing > non-branded organic

Does that help some? Please feel free to follow up here.

Thanks,

Yehoshua

Nico | August 2012

I've been meaning to get my head around custom channel groupings for a while, knowing that they have the potential to solve the not-detailed-enough/too-detailed problem with MCF's. Thank you SO much for this post, you have saved me a lot of time with your clear thinking and clear explanations. Adding you to my short list of bloggers to follow.

Anonymous | March 2013

Best post I've read about GA MCF!

Tobias | April 2013

Hey Yehoshua,

really cool article. I read through it 3 times because it's really filled with a lot of information.

After that I played around a lot with MCF and the Google API to get the MCF Data. I'm trying to consolidate all the data based on the original marketing date.
So I'm taking the conversion date (mcf:conversionDate) and show additionally the time lag (mcf:timeLagInDaysHistogram) together with the first, assisted and last conversion values. Later then I can subtract the Timlag from the conversion date and can look on the conversions based on the original marketing date.

Most of the time it matches exactly the Google Analytics Diagrams in the web when I do some cross-checks. But sometimes not exactly. The weird thing are entries like this:

ConversionDate : Today
TimeLag: 24 days
First = Assisted = Last Conversion Value: 60$

How can that be? Does that mean that this customer clicked 24 days ago on exactly this ad and then today again and therefore the ad has both first and last conversion value?

How can I be sure that I sum up to the right total revenue? I think it should be right if i ONLY sum up First Click Conversion Values. Then I never sum it up twice right?

Would be cool if you could share your thoughts about this.

Thx, Tobias

Anonymous | July 2013

yeah, very interesting video
and a lot of useless information in analytics reports

say:
direct - direct - direct

what the hell is this really???

Jim | October 2013

Another great post about pan-session...It would be very helpful if you can add "pdf-this" feature like Avinash's..would like to read and re-read these posts on the road.

John | November 2013

I like the idea of this post. I'm curious how one actually creates this custom report. Perhaps GA has already changed the interface. I went to the assisted conversion report and clicked on advanced, but I don't seem to be seeing the same options that are shown in the above post. Any help will be appreciated.

Audrey | December 2013

Yehoshua,
Great post and evergreen.
Question on the suggested custom segment that reveals PPC's impact on rest of online: Why are the assisted conversions of all other media except PPC not zero, exactly as PPC's last-click conversions are zero?

For instance, how would you interpret the number 126 for Direct or 79 for Organic in the table shown above (according to the segment definition both should be zero)?

Anonymous | January 2014

How do I use google Analytics to track users' actions (which pages they visit) ON MY SITE till before they make the purchase? My cart is tagged but analytics only shows the incoming traffic from outside sources or funnel only lists couple of my pages.
I sell online video tutorials, I basically want to figure out from which video tutorial pages buyers come from/originate from.

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