Advanced Multi-Channel Funnel Analysis using Google Analytics
Multi-Channel Funnel has been in Google Analytics for a while. Although by searching "Multi-Channel Analysis" you could find a lot of great how-to articles to leverage this powerful function, but seldom of them have explored the opportunities in using it for better resources allocation decision based on ROI estimation, particularly, using Assisted Conversions. Hence, i have decided to put together my experience in marketing, analytic, and infographic to demonstrate the following analytic model. Enjoy.
(Disclosure: i am currently working at MRM Worldiwide, a Digital Strategy agency under McCann WorldGroup, and hopefully the following model will be used in our service someday.....so........ have fun ! XD)
A Closer Look to Assisted Conversion
Apart from the Multi-Channel Funnel view in Google Analytics, Assisted
Conversion Report is the one that we are looking for. If you have some
experience in Omniture, you would know that Assisted Conversion is
indeed having similar logic as Participation Variable,
a way to estimate the potential value that a particular entity, exists
within a funnel of process, has driven. For example, if a visitor
purchase a dress online after she visit a review on a forum (outside the
eshop), both our eshop and the external forum will be entitled to have
contributions, in terms of conversions and revenue, counted towards
themselves, either evenly (all entities have the same value) or linearly
(all entities gain the average of value gained, only in Omniture),
demonstrating that how the entity "participate" within the whole
conversion path.
 |
| A snapshot of Assisted Conversion Report of my groupbuying site: Cheapppy |
....
What makes Multi-Channel Funnel in Google Analytics more powerful (than Omniture) is, it has segmented that participation value for you, based on whether that entity, in the above graph would be "Channel", has contributed as the Last Interaction or not. Should that channel is not the last step before the visitor being converted, then it's "assisting" the conversion flow, and that counts towards as their Assisted Conversions instead.
The power behind this logic is the different between Assisted Conversion and Last Interactions Conversion. Traditionally we count conversion towards the "last stop" of visitors, but with the uprising of Social Media, where fans usually "engage" and "consider" rather than "converted", increases the complexity of the tradition conversion path as well as the evaluation process. Google does this tedious work for you, by introducing the Assisted Conversion, it is easier for analyst to tell if certain channels are good enough to support the conversion flow despite that they might not be the "last stop" of visitors. To make this concept even more clear, Google introduce the "Assisted / Last Interaction Conversions (Ratio)" which tells whether a channel could drive more Assisted or Last Interaction Conversions (>1 = "contribute more within the flow", <1 = "contribute more as last stop").
So, How to decide when we need Resources Reallocation?
Before drilling into details on how we could leverage such report for resources planning, let's talk more about how to determine a if a channel is "Good-or-Bad" under the new complexity of conversion cycle.
 |
| Knowing the Path is one thing, determine the effectiveness is another |
To answer such question, indeed, it depends on the "what you are looking for". In general, a channel with high ROI (relative to other channels) would always mean that they're performing better. With the help of Google for having segmentation in Assisted between Last Interaction makes this question more insightful: is certain channel better at assisting other channels for conversion? If i were Levis, should my f-commerce strategy more effective in assisting other channels or driving direct conversion?
Another frequently asked questions would be, instead of Channel level, how good would certain Ad Group performing? How good are we adapting our Sales cycle along with our SEM strategy (i.e. paid search traffics driven by targeting relevant landing pages based on Awareness-Consideration-PurchaseIntent model) ? An effective Ad Group should thus have a relatively higher Assisted ROI if it is targeting for Awareness or Consideration items, otherwise we should change the target to Purchase Intent or even Conversion pages if it's last interaction ROI is higher.
Simply speaking, to determine if we reallocating from one to another, we need to determine the characteristics of the items, either they're Channels or AdGroups, first.
Assisted ROI & Direct ROI Estimation
Normally Google suggests us to us "Assisted / Last Interaction Conversions" to analyze for how good certain channel perform in either Assisted or Direct way. But it's just a ratio based on "occurrence". If you have followed the whole logic so far, you should know by now we should look for a way to determine both the Assisted ROI (from Assisted Conversions) or Direct ROI (from Last Interaction Conversions) of Channels which provide a more business angle for us to handle our question. So how could we calculate such business metric based on what we currently have (the Assisted Conversion report)? Let's begin from the basic definition:
As for our case, we could easily fill-in-blank using the following formulas...
And here's how the data (fake one) presented in a spreadsheet:
 |
| Spreadsheet with mock-up data (already explained a lot of things!) |
Organ Part - Assisted Conversions Report from Google Analytics
Green Part - Cost spent a particular Channel (or items). Based on the participation concept, the total Cost spent on a Channel will be shared by all the Assisted and Direct Conversions, thus the Assisted Cost and Direct Cost of a Channel could be estimated by the portion of corresponding conversions achieved. (If you wish to know more about Costing Estimation,
Purple Part - ROI based on corresponding segment (Assisted or Direct)
Easy enough, to determine the channel characteristics, we simply put the data on a Relational Map, with x-axis as Assisted ROI and y-axis as Direct ROI. Almost done!
 |
| Looks like the Referral is under-preformed, sounds like a good action point to begin with! |
What Actions We need to Take?
I couldn't emphasize more enough that any chart or infographics without Action Triggers is simply meaningless., I have talked about how we could put an infographic onto an upper level and make it more actionable in the 5th steps. A relation map like above is no more than a graphics presentation based on data. So the key is to help readers identify the action right away after they read the chart. On the above chart, think about breaking down into 4 different sections:
1 | 2
3 | 4
Section #1 - Channels that are bad in Assisted Conversion (-ve Assisted ROI) but good at Direct Conversion (+ Direct ROI)
Section #2 - Channels that are good at both Assisted and Direct Conversion
Section #3 - Channels that are bad at both Assisted and Direct Conversion.... (simply under-preformed...)
Section #4 - Channels that are good at Assisted Conversion but bad at Direct one
Now identifying action is simple:
If you are looking for under-performed channel, read Section #3, those channels are definitely having some issue (which you have to investigate!)
If you are looking for performance of social entities, like Facebook, Twitter, Youtube, and even other online communities, where you are expecting high Assisted ROI, see if they're at Section #2 or #4, if not, well, you know what i mean.
How about reallocation of resources? Start from something poorly performed and move those resources, usually dollar-signed, to those well performed ones. Simply speaking, if you have decide to reallocate instead of optimizing or fixing problem, move resources from items in Section #3 to those in #2. Easy.
The point here is, in any business, or even down to a single business process (like SEM), all we're focusing is Return on Investment. If some operation couldn't bring return in any form (e.g. Impression is a kind of return, and we could easy convert it into dollar-signed using CPM), then it is either we need to fix the problem, or simply give it up and free the resources for those well-performed ones.
Always remember the if-this-than-that rules. It always helps in designing for actions triggers.
Looks Good, but only Channel level Analysis?
No. (Why stop here? XD)
The Assisted vs Direct ROI drives lots of potential dimensions in analysis, here's some other variation we could take a look based on the same logic as above:
1. Referral Analysis
Segment the Assisted report based on Source/Medium will give you a over view in all upstream traffics. It is essentially important if you have broad social media strategy which occupying different channels like facebook plus twitter plus linkedin plus pinterest and so on... then such break down will let you understand how good is your social media team is working and let them know if they need to tune their tactics in different media.
2. Campaign Analysis
A more aggressive approach is broken down by Campaigns. This angle provides marketers a more insightful view on how different campaigns are performing. Not limited to social media engagement, but also social ad. vs sponsored tweets, break-up email vs cart-abandon emails, banners ad vs display ad, (offline with qr code) etc. Make sure your marketing team have tagged the upstream URL correctly in order to fully leverage this powerful report.
3. AdGroup / Keywords Analysis (in AdWords)
The last angle we could have a look is the SEM performance, which, traditionally, we focus too much solely on the click-through rate and cost per click, and simply overlook the importance of how they actually generate goals or even leads to our business. With the Assisted / Direct ROI model we could now easily tell if certain AdGroup or Keyword are performing as expected, says, a set of retention keywords (e.g. "where to repair my iPad 2?") should be expecting a higher Assisted ROI as it helps satisfying customer and supporting future purchases. This will also help strategizing how each landing page should be doing as well.
“In other languages, there’s often less competition doing international search,” says 
