Peter Howley of Empirical Path delivered this presentation to the HTTQ web design and developers group in April 2013. Topics include counting conversions, segmenting audiences, testing ideas, and sharing insights using Google Analytics, Optimizely, Klipfolio and other tools. Video also available at TBD.
3. @EmpiricalPath
‣ Web analytics, market research and
campaign measurement
‣ Founded in Washington DC in 2002
‣ Atlanta, NM and DC offices
‣ Founder led web analytics at
washingtonpost.com
‣ Google Analytics Certified Partner
‣ Webtrends Agency
4. @EmpiricalPath
‣ Defend digital investment
‣ Compare online to offline
‣ Right-size digital investment
‣ Re-allocate budget to most valuable
activities
‣ Celebrate and reward successes
‣ Learn from mistakes
5. @EmpiricalPath
Reports with little context that measure
only how many people entered the
funnel:
‣ Impressions
‣ Visits
‣ Unique visitors
‣ Friends
‣ Followers
6. @EmpiricalPath
Measure how many times they complete
the funnel (& achieve campaign goal):
‣ Purchases
‣ Social shares
‣ Email signups
‣ Contacts via form & phone
‣ File downloads
‣ Video views
7. @EmpiricalPath
‣ Track user activity on pages
separately from pageviews
‣ Any Flash-driven element, like a
website or Movie player
‣ Embedded AJAX page elements
‣ Page gadgets
‣ File downloads
‣ Clicks on links to other domains
‣ Form submits without a unique
thank-you URL
‣ Load times for data
‣ Correct Bounce Rate by defining
visits with these user activities as
not Bounces
8. @EmpiricalPath
‣ Category
‣ Root level of event tracking and is the base level for sorting events
‣ Collect similar actions together, e.g., Outbound Links, Downloads,
Videos
‣ Action
‣ Descriptor for a particular event, described by any string you specify
‣ Verb that describes what user did, e.g., Played, Clicked, Paused,
Downloaded
‣ Not just one per category
‣ Label
‣ Optional descriptor to provide further granularity
‣ Often automatically passed based on code on page
‣ Don’t record Page where event occurred; tracked automatically
‣ Value
‣ Numerical variable can have explicit or inferred values
‣ Economic value of non-commerce action
‣ Estimated value of unmeasured commerce action
‣ Non-interactive parameter
‣ Tells GA that an Event was not initiated by visitor
‣ Counts single-page visits with only these Events in Bounce Rate
calculation
9. @EmpiricalPath
‣ Core events reports in Standard
Reporting > Content section
‣ Top Events lets you drill down from
Categories
‣ Overview rolls up all Categories,
Events, and Actions
‣ Metrics on total & unique events
‣ And any optional value tracked
‣ Standard tabs also available
‣ Site Usage: Visits, Avg. Duration, %
New Visits
‣ Ecommerce: Revenue, Transactions,
Avg. Order Value
10. @EmpiricalPath
‣ gaforflash is an ActionScript 3
API for Google Analytics data
collection
‣ 1st, add the gaforflash SWC to
Flash CS3
‣ Quit Flash CS3
‣ Navigate to the location where
you unzipped the GA zip and find
the swc
‣ Create a "Google" directory
‣ Copy the SWC file there
11. @EmpiricalPath
public function handlePlayTime(event:Event):void
{
//Tracking elapsed time for video for event and pageview
tracker.trackEvent("Videos", "AutoPlayed",
"[VideoTitle]",PlayTime);
tracker.trackPageview("/TDNB/[DomainName]/Videos/[VideoT
itle]");
}
public function handleComplete(event:Event):void
{
tracker.trackEvent("Videos", "Completed", "[VideoTitle]");
}
public function handlePause(event:MouseEvent):void
{
var button:Button = event.target as Button;
if (button.label == "Pause") {
video.stop();
button.label = "Resume";
tracker.trackEvent("Videos", "Paused", "[VideoTitle]");
} else {
video.play();
button.label = "Pause";
tracker.trackEvent("Videos", "Resumed", "[VideoTitle]",
ResumeTime);
}
}
‣ Example Tracking:
‣ http://stanfordhospital.org/events/awareness/stroke.html
‣ http://test.empiricalpath.com/ga-youtube2.html
12. @EmpiricalPath
‣ Example Reporting:
‣ https://www.google.com/analytics/web/?#realtime/rt-event/a12625662w34163983p33420663/%3Ffilter.list%3D40%3D%3Dyoutube-
video%25253A3Sk7cOqB9Dk%3B
22. @EmpiricalPath
Experiment with ideas and use
conversion rate in key segments to pick
a winner:
‣ Landing Page
‣ Ad Copy
‣ Home Page
‣ Offer
‣ Discount
‣ Button color
23. @EmpiricalPath
Above-the-fold
call-to-action
23% of next clicks
from the
homepage are to
promotions
23% of next clicks
from the
homepage are to
promotions
23% of next clicks
from the
homepage are to
promotions
23% of next
clicks from the
homepage are
to promotions
30. @EmpiricalPath
Downplay;
few click but
most of those
come back
Shrink;
pushes
products &
prices down
Shrink;
pushes
products &
prices down
Keep; many
entrances
click
Lose; draws
eye from
product art &
headline
32. @EmpiricalPath
Testing Tool Pros Cons Estimated Costs
Content Experiments
Free
No worries re: over-sampling
Easy to use
Part of existing GA login and code
base
No support but GACPs like
Empirical Path
No multi-variate testing
Non-intuitive splits between tests &
control
No visual WYSIWYG editor
No targeting
Free
Includes heatmaps & clickmaps
Multi-variate testing
Visual WYSIWYG editor
Includes targeting
GA integration
Customer support via email only
$49/mo for 10k UVs/mo
$249/mo for 100k UVs/mo
$1,700/mo for 1M UVs/mo
Live customer support
Multi-variate testing
Visual WYSIWYG editor
Forrester “strong performer”
Includes targeting
GA integration
Lacks heatmaps & clickmaps
Smaller early-stage firm
$71/mo for 10k UVs/mo
$359/mo for 100k UVs/mo
$1,999/mo for 1M UVs/mo
Deep support team at Adobe
Forrester “leader”
Includes targeting
Essentially requires costly support
team
Worst references in Forrester study
No GA integration
$900-$1,300/mo with substantial
incremental costs for each “mbox”
(the area where split-test is
executed, such as a graphic that is
being tested)
Deep support team
Forrester “leader”
Fixed fee lets you test ad nauseum
High total cost
Weak on integrations
“Six-figures” per year
The typical digital measurement reportLacks context: vs. plan, vs. prior period, vs. goalLacks focus on target audienceLacks impact in $
Conversions are obvious for direct response advertisersTake some thinking for other advertisersCan track phone calls using tools like IfByPhoneCan use proxies for offline conversions like looking up dealer location
This publishing client didn’t sell anythingInstead, engagement matters for ad salesBig commitment to social mediaNeeded to justify real estate and tech investment for sharing tools
Event tracking reveals user actions that don’t create a pageviewReports allow correlation with content, traffic sources, geography, technology
Now client can see Goal Flow for any key segmentsShows drop-off among various steps on checkoutShows how many visits from each source, keyword, state, and other segment start and finish checkout
Client can also see how some campaigns “assist” conversion.For example, someone who first uses organic search to reach the site, then visits directly to subscribe later.Both channels can get credit.
All the measurement in the world is useless if you don’t do something.Split testing lets you apply learning from measurement.A way to settle debates on the best user experience.A way to continuously improve conversion rates.
This home fixtures client had a high performing home page.But a key internal landing page was less than half as effective.We studied the homepage stats to see what was working there.
We also studied the landing page stats to see where people clicked next.We joined the client and agency in brain-storming ways to improve the page.
This client kept it simple with a pre-post test, versus a true split test.Promotions helped the home page, so they added them to the landing page.
We compared the period after this change to the same weekdays before.Revenue per visit from visits that viewed the landing page jumped.The bounce rate – visits that started and ended on the page with no other clicks -- dropped
Conversion rate jumped.Revenue climbed among these visitors even though there were fewer of them, for external reasons.
This solar client drove outstanding organic traffic to a landing page.But they wanted the page to yield more conversions.We analyzed the landing page stats and drafted a “straw-man” copy and layout.
This solar client drove outstanding organic traffic to a landing page.But they wanted the page to yield more conversions.We analyzed the landing page stats and drafted a “straw-man” copy and layout.
This solar client drove outstanding organic traffic to a landing page.But they wanted the page to yield more conversions.We analyzed the landing page stats and drafted a “straw-man” copy and layout.
This client kept it too simple and tested our straw man concepts.Phone conversions were also tracked but not part of the split test.Pricing per watt moved the needle slightly!But other changes made no significant impact.Lessons: invest in high-quality test elements!
Segments can be anything that is worth tracking and studying Analytics track some out of the box: state, browser, monitor resolution, new v. returning, search engine keywordOthers you can customize based on visitor behavior
Our university client serves varying constituencies
It has developed specific content, calls to action, and conversions for each.We customized Google Analytics to cookie visitors for a number of weeks based on which Guide to Emory they select
Now all standard and custom reports can include these SegmentsHere are Goal completions for each segment, for instance.Could do Goal Completions by Segment by Traffic Source.Could do Pages per Visit as an engagement metric by Segment.
The most important segment is, which campaign brought the visitor?Empirical Path is just a modest B2B advertiser, but we pay for campaigns.
We’re also experimenting in every social medium.And we send digital measurement tips to our mailing list quarterly.
But we get the most visits and leads from our software and staffing partnerships.That’s a dozen or more campaigns; how can consumer advertisers track hundreds?
We recommend a spreadsheet tool to speed up “tagging” campaign landing pagesOften a Google Doc to be shared by everyone in an organization who creates inbound links.One row for each link or links, as granular as you want to go.E.g., email > November blast > lapsed customer list > top story > image link.Requires rules on naming and capitalization to be most effective.Can power more than one analytics tool when agencies and clients use many.
The tool adds some parameters to landing page URLs These don’t affect page load
Google Analytics tags add three to five dimensions to each campaign link.Most important for email marketing, which is counted as either “direct” or “referral”; it is neitherOpens up campaign reporting drill-down capabilities in Google Analytics and Clicky, for starters.But “garbage in, garbage out.”
Google Analytics tags add three to five dimensions to each campaign link.Most important for email marketing, which is counted as either “direct” or “referral”; it is neitherOpens up campaign reporting drill-down capabilities in Google Analytics and Clicky, for starters.But “garbage in, garbage out.”
Google Analytics tags add three to five dimensions to each campaign link.Most important for email marketing, which is counted as either “direct” or “referral”; it is neitherOpens up campaign reporting drill-down capabilities in Google Analytics and Clicky, for starters.But “garbage in, garbage out.”
As with testing, all the metrics in the world don’t help if decision-makers don’t see them.
Every analytics tool has a dashboard for those willing to log into the system.Customizing it with only the important reports – not defaults – boosts usage.
Everyone has email, and you can automatically schedule reports or dashboards this way.Not everyone opens messages and attachments.Some are frustrated because they can’t link to learn more.
Everyone has email, and you can automatically schedule reports or dashboards this way.Not everyone opens messages and attachments.Some are frustrated because they can’t link to learn more.
Everyone has email, and you can automatically schedule reports or dashboards this way.Not everyone opens messages and attachments.Some are frustrated because they can’t link to learn more.
Everyone has email, and you can automatically schedule reports or dashboards this way.Not everyone opens messages and attachments.Some are frustrated because they can’t link to learn more.
Lots of analytics tools allow access via mobile devices, too.Generally limited drill-down capabilities for the analytics geek in you.Webtrends, ClickyTouch, ChartBeat, Analytics App
Real-time tools like ChartBeat are increasingly popular.
Woopra adds live chat to a dashboard so you can reach out to visitors real-time.
Clicky has a stripped down display you could post in the plasma screen in the lobby.
Of course, Google Analytics won’t sit still.They have a beta real-time display so you don’t test other tools.