Empirical Path web analytics practice lead Jim Snyder was featured in another webinar with one of our agency partners and clients, Seattle search engine marketing firm Point It. The presentation covers:
- How to give credit to various sources across your sales funnel with the MultiChannel Funnels/Attribution Modeling Tool
- How to use Content Experiments to test different versions of your site to understand the most effective versions of landing pages
- How to track the same visitors across mobile to your website using Universal Analytics
What’s New in Google Analytics: New Features & What You Need to Know
1. Jim Snyder, Web Analytics Practice Lead at Empirical Path
Preston Parshall, Point It
September 11, 2013
What’s New in Google Analytics
New Features & What You Need to Know
2. Presentation Agenda
• Launched in April 2002
• Seattle’s largest independent
search marketing firm
• SEM, SEO, Social, Local, Mobile
• $30 MM in managed media/yr
• Servicing clients across all
verticals and revenue models
3. • Being recorded & will be uploaded to website
• This is an interactive webinar so it is okay to
ask questions during the presentation
Webinar Information
4. Presentation Agenda
• Web analytics, market research
and campaign measurement
• Founded in Washington DC in ‘02
• Atlanta, NM and DC offices
• Seasoned Web analytics
professionals
• Google Analytics Certified Partner
• Webtrends Agency
• Contacts
• www.EmpiricalPath.com
• @EmpiricalPath
• Facebook.com/ EmpiricalPath
• +Empirical Path
About Empirical Path
6. Test Ideas: Content Experiments
• GA lets you experiment with
ideas and use conversion
rate in key segments to pick
a winner:
• Landing Page
• Ad Copy
• Calls to action
• Offer
• Discount
• Button color
• Page layout
19. • The vast majority of customers have multiple
touch points before converting
• Overwhelmingly the last click gets the entire
credit for a conversion and first and other
contributing touches never get noticed!
• The CPA leads most advertisers to spend
money to a channel where the decision to buy
has already been made
Attribution Modeling: Background
20. Attribution Modeling: Menus
• Choose model
to see change
in CPA, Value
or ROAS
• Compare up to
3 models side-
by-side
• Or create
custom models
based on
business model
29. Attribution Modeling: Functionality
Last Interaction: Last interaction gets 100% credit (including direct)
Last Non-Direct Click: Last interaction before direct gets 100% credit
(what is used in GA by default)
First Interaction: First touch gets 100% credit
Linear: Each touch gets equal % of credit
Last AdWords Click: Last AdWords click get 100% credit
Time Decay: Touch closest to conversion gets more credit
Position Based: 40% goes equally to first/last. 20% is spread evenly
with other sources
36. • Newest version of Google Analytics that changes the
way data is collected
• Custom Dimensions & Metrics: Replaces Custom
Variables and stored right in GA interfaced). Allows
customization for anything (like profit margin for
metric)
• Moves away from visit-based and features tracking
individually across devices & websites
• Features the Measurement Protocol that allows you
to track across ANY device (ID Scanners, Refrigerators
– literally anything).
Universal Analytics: What is UA?
39. Universal Analytics: Custom Dimensions
Custom
Dimension
Choices:
Hit: Like page-level custom variable
Session: Like Session-level custom variable
User: Stays with user past the session and
for duration of cookie.
Replaces Custom
Variable. Info
entered directly
into GA interface.
Replaces Custom
Variable. Info
entered directly
into GA interface.
51. • Preston Parshall
• prestonp@pointit.com
• Point It Search
Marketing Agency
• www.PointIt.com
• support@pointit.com
Thank You
• Jim Snyder
• jim.snyder@empiricalpath.com
• @jimdsnyder
• 203-804-4509
• Empirical Path
• www.EmpiricalPath.com
• @EmpiricalPath
• Facebook.com/
EmpiricalPath
• +Empirical Path
Notes de l'éditeur
Thanks for joining us – just some quick background on Empirical PathAll we do is measurement - web analytics, market research and campaign measurementThat’s why it’s great to team up with PointIt on search and marketingWe have offices in DC, Atlanta and New MexicoWe have a very close partnership with google as a certified partnerThese are some of the companies we’ve worked with on Web analytics
So here is our agenda. These are features that if you’re not using now – you really need to be in GA.They’ll save you time, give you more insights than you imagined and make you a rock-star in to your C-suiteThere are some you may have heard of – some are very recently released - others have new parts to them that you’ll want to know about.Especially MultiChannel Funnels and the API.So we’ll start out talking about A/B testing (essentially what is merged from the GWO product)Then I’ll take a look at Multichannel funnels and what dimensions you should really look at.That leads into the API which feeds the abilty to see your GA results in Google Docs –so you don’t really have use the GA interface anymore if you don’t want!Then we’ll look at tag Manager – a just released product that controls all of your GA and marketing tags.Finally we’ll look at the remarketing feature.
So let’s start out by talking about a big change that happened Recently.Google Website Optimizer(which many of you might have used) essentially was retired and brought right into Google Analytics.What is content experiments and why should you care? This is a very useful feature product that essentially can help you increase performance of landing pages By letting you literally test various landing pages, ad copy, calls to action, offers, discounts etc.Even details down to button colors and page layouts. Each page is served up automatically an equal amount of time.You tell Google the pages you want to test and the action you want them to take and wa- lah you can see what is the most successful page.
So who’s actually using this tool and what did they get out of itHere’s an example of a client of our’s that was able to produce a lot of value through Cont. Experiments. You’re looking at the original landing page for a client of our’sKapitall. This company has an online stock trading platform. The call to action is a sign-up for a brokerage account. Each sign-up is huge to Kapitall so they need to have the most efficient landing page. So we set up a Content Experiment in Google Analytics.
So as you can see their landing pages feature a sort of gaming-like interface.we worked with the design team to create four different versions – classic gaming – Pac Man/Mario/Space Invaders and newer Angry Birds. each of which (like GWO) is served up equally over the course of about 2 weeks (usually around 1K visits to determine statistical significance).So some of you might be familiar with how A/B testing works – not going to cover a lot of detail but GA uses the ‘Multi Armed Bandits’ model. Google automatically checks twice a day to see how each variation has performed. Then they surface variations that are best statistically performing more often – pushing more traffic to the better performing variations. This cuts down on the time to find a winner and is still statistically valid. And Google does all of the work for you to serve up the pages.
This is what the set-up looks like in content experiments. Create different versions of the landing page ahead of time to a unique URL. There are some considerations for SEO that I’m not going to cover but suffice it to say that you can include ‘REL’Cannonical’ parameter is a signal to Engines that the pages are essentially the same. You can have up to I believe 9 variations
Jim: Each experiment must have one goal in which to evaluate a ‘winner’. They call this an ‘objective’ in CE but it’s a goal on your site that is in use. In this case the goal set up is triggered off of new account sign-ups. You can see there are other settings such as the % of visitors you want to see the experiment. In this case it’s 100% but lesser percentages can be selected.
Just like GWO there is a simple workflow where the code is spit out for you and you can insert right on your ‘original’ page. You then publish the experiment code to the live site. There’s some validation that takes place and then you launch the experiment.
And this is so cool. We can see after 15 days we have Google has served up each page equally and determined with statistical significance that Angry Birds won! We can see that it has a 99.93% chance to beat the original. There’s even an option to serve the winning version automatically.So this ended up being very big for Kapitall because every sign-up they get for brokerage accounts is critcal to their business.
So here is our agenda. These are features that if you’re not using now – you really need to be in GA.They’ll save you time, give you more insights than you imagined and make you a rock-star in to your C-suiteThere are some you may have heard of – some are very recently released - others have new parts to them that you’ll want to know about.Especially MultiChannel Funnels and the API.So we’ll start out talking about A/B testing (essentially what is merged from the GWO product)Then I’ll take a look at Multichannel funnels and what dimensions you should really look at.That leads into the API which feeds the abilty to see your GA results in Google Docs –so you don’t really have use the GA interface anymore if you don’t want!Then we’ll look at tag Manager – a just released product that controls all of your GA and marketing tags.Finally we’ll look at the remarketing feature.
So by default Google analytics defaults to what’s called a ‘last click’ model. On this example – Organic gets credit for 1108 purchases. In this way 100% of the credit goes to the last source someone clicked before converting (as example bought something, downloaded a white paper, watched a video etc.). This is good information but sometimes skews other sources such as paid that will get no credit for assisting in this model.
But first a bit of backround when you look at the whole picture – last click isn’t this simple. So when we look at visitors that come to your site - there's a series of decisions that people make before buying something, downloading a white paper things that we call 'conversions'It's almost never a simple 'see a display ad or adwords ad and buy'. There are very complex decisions that are made before say making a purchase.We can see some that go from a display ad to a search ad then purchase.Or Go from Organic to Social to purchase. Or some that don't purchase at all. Often times we see email and social and display typically far away from the last click. They introduce and ‘assist’ in the conversion process.And Google Analyitcs gives you a way to sort of unpack each of those visits by providing functionality through Attribution Modeling & Multichannel funnels to help expose what actually are the factors contributing to leads, purchases etc. on your site to help elaborate on the entire conversion funnel
What is Attribution Modeling:We'll start at a very high level. As discussed there are many interactions that take place before a conversion occurs through a website or mobile device. Google Analytics by default credits the last click - seen here as 'Email.
We can now see the individual interactions that lead up to making the cash register ring on the website (or what is called a ‘conversion’ in GA) This is the conversion funnel.
And how are these individual interactions valued?But also we’re interested to see not just how people move from say Social to Display advertising but also intra-channel as they go from generic to brand search and back? We want to know how that might impact ROI and
And how are these individual interactions valued?But also we’re interested to see not just how people move from say Social to Display advertising but also intra-channel as they go from generic to brand search and back? We want to know how that might impact ROI and
So you can find the tool located in the side bar under ‘Conversions’ section. You’ll also see the multichannel funnels reports listed in this section as well.Multichannel funnels has been out for almost 2 years – so you might be familiar with these sets of reports. The Attribution Model comparison tool used to be only available inside of the paid version of Google Analytics (GA Premium) and was recently release for free under GA standard.Attribution Model Comparison tool bulilds off the same data featured inside of the MCF but this tool is really about showing you the steps leading up to conversion on your website and mobile devices. When you open up the modeling tool you can compare models to see change in channel values Let’s dive in and take a look.
This is the last click model. It’s really the default view that you have with reporting in GA.It’s a good starting point because this is the perspective of what you have in GA when you start to analyze. So if someone started out with paid, did a Bing search and then bought – Bing gets 100% of the credit and paid search gets none.Outside of Attribution modeling in GA 100% of the credit goes to the last non-direct click in the last 6 months.
Then you can look at contributing traffic sources from a FIRST CLICK to see how these sources fit in. This tool lets you dive in and compare up to three of these models side by side from a different perspective. So you can see how Organic and Social contribute and in what order. And there are many models you can pick from as I’ll show you shortly
So I can immediately see how the value changes under the first click model. And the change compared to last click is shown. We can see how the first interaction compares with the last interaction – green arrows show increases and red arrows show decreases
In this case – we see CPA is much less - $8 and 58 cents compared to nearly 6 dollars more under a last click and it’s a 71% change in conversions for paid search if we look at a first click model. Then you can drill down inside of each area to get more detail.
Drilling into paid search I can see that Google CPC was mostly the key driver and accounted for 59% more conversions on the first click model than the last. I have a great Return on Ad Spend under the last click model Return on Ad Spend is $487% better on the first click model so I can even go deeper and look at specific keywords to see what are the key ingredients to this jump under first click so we can replicate it. I’ll show more of that in a minute
And also – Google recently changed so that your lookback window can change from a minium of 1 day to a max of 90 days! You’ll want to tweak this setting depending on your sales cycle. As example if you’re selling a higher consideration item like a car then that lookback window will be longer. Something that’s more spur of the moment like clothing can have shorter lookback windows. This really depends on your sales cycle.
And also – Google recently changed so that your lookback window can change from a minium of 1 day to a max of 90 days! You’ll want to tweak this setting depending on your sales cycle. As example if you’re selling a higher consideration item like a car then that lookback window will be longer. Something that’s more spur of the moment like clothing can have shorter lookback windows. This really depends on your sales cycle.
A deeper dive on choosing the last model. You can see there are a few last click models, first click as we’ve seen and some others that I’ll discuss. Plus you have the ability to create Custom Models that are even closer to your business model. We’ll see more of those in a minute.
In the Last Interaction attribution model, the last touchpoint -- in this case, the Direct channel -- would receive 100% of the credit for the sale. In the Last Non-Direct Click attribution model, all direct visits are ignored, and 100% of the credit for the sale goes to the last channel that the customer clicked through from before converting -- in this case, the Email channel. In the Last AdWords Click attribution model, the last AdWords click -- in this case, the first and only click to the Paid Search channel -- would receive 100% of the credit for the sale. In the First Interaction attribution model, the first touchpoint -- in this case, the Paid Search channel -- would receive 100% of the credit for the sale. In the 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. Useful for say a 2 step process or longer paths, or you want to understand the frequencency of interactions. If there’s a reason for a lot of returning, researching this is a very useful model for that. In the Time Decay attribution model, 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 or Email channels. Since the Paid Search interaction occurred one week earlier, this channel would receive significantly less credit. Very focused on the bottom of the funnel – to expose what In the 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.
So in this example – I’ve created 2 custom custom marketing channels - Branded Search and non-branded search. To give you some ideas on how to use Attribution Modeling. This is a super valuable distinction of how to look at how well your search traffic performs in a multi-channel context.On top of that I’ve also allied 2 custom models – that compare things at from the lower funnel and upper funnel compared to last click. Lower funnel gives more credit to searches toward the end bottom part of the sales funnel and the upper funnel that give more credit toward the early search keywords that contributed to conversions and de-emphasizing direct and branded search interactions.
I can also click on ‘Source’ directly at the top to get a different more finite view of different sources. I can quickly see that source ‘Critsend’ which is an email source – valued 202 conversions on my upper funnel model compared only 65 with a last click. This is a difference of 212%. So you can start to see that this tool really forces you to value interactions differently as you start to value different marketing channels.
Another view for specific campaigns shows me that the highled campaign drove most of that value. It gives 2 under last click and 27 on the upper part of the funnel. That’s a 1285% difference! HUGE!
Finally – you can see another example (Disgused Data btw) of the actual keywords that drove the value and how to compare them. Again upper funnel shows 101 conversions for the keyword ‘Google Analytics Audit’. So you you can start to really focus on what is working at the keyword level as well.
So here is our agenda. These are features that if you’re not using now – you really need to be in GA.They’ll save you time, give you more insights than you imagined and make you a rock-star in to your C-suiteThere are some you may have heard of – some are very recently released - others have new parts to them that you’ll want to know about.Especially MultiChannel Funnels and the API.So we’ll start out talking about A/B testing (essentially what is merged from the GWO product)Then I’ll take a look at Multichannel funnels and what dimensions you should really look at.That leads into the API which feeds the abilty to see your GA results in Google Docs –so you don’t really have use the GA interface anymore if you don’t want!Then we’ll look at tag Manager – a just released product that controls all of your GA and marketing tags.Finally we’ll look at the remarketing feature.
And now I’m looking at specifically which branded keywords are actually assisting.What role are they playing? This is huge for SEO to know this information.
Here’s a very simple exmple of the measurement protocol. I can literally take come code – slap it in a browser and it will register directly into GA> if I enter this entire URL – from anywhere on any device – I can see
And if you want to follow along at home – you can sub out that UA number with your own and watch it pop right into Google Analytics Real Time.
Like we discussed this is how to set up custom dimensions right inside of the GA interface. First you select Custom Defininitions, then custom dimensions. From there you decide (look on screen for choices).
And looking at actual reports I can see that if I wanted to create a custom demension – in this case User ID - I can literally see how different people buy across device to device. Keep in mind this has to still be non-personally identifable information so no emails or SSN.
And I can see that how revenue contributed across different devices. I can see that this person started on a desktop then bought on moblle.
And then later (or perhaps before) downloaded a white paper.