In this series of articles I’m going to share with you our approach to clearing away the less fruitful pursuits in your web analytics and focus you on the ones which will yield the biggest insights with the minimum effort.

Install Google Analytics

follow the rules to install Google Analytics on your website at http://www.google.com/analytics/ (Little plug: We do that for no additional cost when you buy either our Search Presence or Search Light services.)

Define goals

The starting point for all good decision making is to decide what is a good thing.  It is no different with your website and advertising.  We always create two “Goals” in Google Analytics.  The first represents a good thing that happens fairly often — we always refer to this as “Engagement”.  Then we create the goal which represents the closest thing your website has to a sale and call this “Conversion.”  We’d always dig in to find the best choice for each of these goal definitions, however for the sake of example let’s say your website doesn’t have any ecommerce but you’re very happy to get sales leads on your site through your contact form.  In this case we’d setup these two goals in GA:

• Goal 1
• Give it a name like “Contact Engagement”
• Type: URL destination
• Goal URL: /contact (or whatever the path to your contact page is)
• Match type: regular expression (this allows the URL to match if you have /products/contact, /services/contact, /corporate/contact)
• Give it a value if you’d like to keep track of the value of these interactions
• Goal 2
• Call it “Contact Conversion”
• Type: URL destination
• Goal URL: /contact/thankyou (use your actual page address)
• Match type: regular expression
• Set the value.  This should be significantly higher.    Engagement is a precursor to the truly valuable interaction — completing the sale

So now you’re off to a good start.  Google will keep track of the goals on your website going forward.  You’ll now be able to see the engagement and conversion taking place on your site.

Truly, this is only the beginning though.  When your reports show you referrers that generate engagement you’ve really got something.  When you know which geographic regions generate engagement you’ve got a little more.  Identify the content on your site that is engaging and BAM!  you’re getting closer.  Keywords that generate engagement is the icing on the cake.  Of course, you need a custom report for all that.  I’ll save that for Part 2.

We wrap all this up in our Search Light service where you don’t have to look at any code or carry out these steps yourself.  Better yet, the report comes to your email as a spreadsheet with exactly what you need to know on your schedule!

I found Chris S. Penn’s excellent article on this topic at http://www.christopherspenn.com/2011/01/how-balanced-is-your-google-analytics-pie/

Good

• Make Google, Bing, and Yahoo happy
• Increase your ad’s quality score
• Decrease your Cost Per Click (CPC)

Bad

• Spend your budget faster
• Decrease the number of impressions your ad gets
• Increase the cost to keep your ads running for any given date/time schedule

Just from that list you can see there is a trade-off with increasing CTR.  Sometimes this is good.  Sometimes it is not good.  Here’s an example of each.

Good

You’re running a campaign and it costs $1.41/click (CPC = $1.41).  Your total spend is $10.53/day.  You have another $9.47/day budget to buy more clicks to your website.  You’ve checked to see that you’re getting a respectable engagement and conversion from these visitors… You are.  You’ve also been tuning your keywords for months and you don’t have any to add to the mix.  What are you going to do to get more qualified clicks on your website landing pages?  Increase the CTR by rewriting your ads to be BOTH more engaging AND stay true to the landing page.  There are many sources of wisdom on how to increase the CTR for your ads so I won’t touch that right now.  Go Google-Bing-Yahoo-Ask for them.

I’ll offer an example of  the wrong way to increase CTR in part 2, The Bad.

First, I’ll focus on just the creation of the query URI.  In part 1, our query URL looked like:

query_uri = gdata.analytics.client.DataFeedQuery({
      'ids': PROFILE_ID,
      'start-date': sd,
      'end-date': ed,
      'dimensions': 'ga:date',
      'metrics': 'ga:visits',
	})

and had the output:

ga:date ga:visits
20111121        3214
20111122        2692
20111123        2360
20111124        1537
20111125        2227
20111126        2171
20111127        2220

A query asks the Google Analytics API for all metrics for each dimension specified. In this case the query asks for all “ga:visits” for each “ga:date” dimension which are recorded over the date range from “start-date” to “end-date” for the “ids” specified.

Here’s a new query to show visits and pageviews for all referrers each day inside the date range:

 query_uri = gdata.analytics.client.DataFeedQuery({
      'ids': PROFILE_ID,
      'start-date': sd,
      'end-date': ed,
      'dimensions': 'ga:date,ga:source',
      'metrics': 'ga:visits,ga:pageviews',
        })

And its output looks like:

ga:date ga:source       ga:visits       ga:pageviews
20111121        (direct)        1029    7806
20111121        ask     8       44
20111121        austin360.com   1       12
20111121        bing    133     1130
20111121        en.wikipedia.org        2       40
20111121        facebook.com    2       9
20111121        google  1550    10642
20111121        google.com      6       27
20111121        home.myhughesnet.com    1       2
20111121        yellowbook.com  3       15
20111121        yellowpages.com 1       5
20111121        yelp.com        1       9
... many entries eliminated to protect the innocent
20111122        (direct)        844     4728
20111122        aol     5       11
20111122        ask     4       32
20111122        yahoo   124     1005
20111122        yellowbook.com  2       10
20111122        yellowpages.com 3       31
... many entries eliminated to protect the innocent
20111123        (direct)        793     4256
20111123        aol     9       62
20111123        ask     3       75
20111123        bing    118     867
20111123        yelp.com        1       7
... many entries eliminated to protect the innocent
20111124        (direct)        475     3526
20111124        aol     4       31
20111124        ask     5       37
20111124        bing    60      389
... many entries eliminated to protect the innocent
20111125        (direct)        676     4041
20111125        aol     6       42
20111125        ask     7       44
20111125        bing    99      588
20111125        yahoo   130     1152
20111125        yellowbook.com  3       14
... many entries eliminated to protect the innocent
20111126        (direct)        664     3524
20111126        aol     5       35
20111126        ask     9       34
20111126        bing    76      599
20111126        sxsw.com        7       17
20111126        yahoo   138     1002
20111126        yellowbook.com  5       18
20111126        yellowpages.com 1       19
20111126        yelp.com        2       48
... many entries eliminated to protect the innocent
20111127        (direct)        636     3507
20111127        aol     7       127
20111127        ask     7       67
20111127        bing    84      548
20111127        en.wikipedia.org        1       7
20111127        facebook.com    2       12
20111127        google.com      10      72
20111127        m.yp.com        2       5
20111127        mail.aol.com    1       4
20111127        yahoo   127     1074
20111127        yellowbook.com  3       10
20111127        yellowpages.com 1       1
20111127        yelp.com        2       38

Drop the date dimension and you get all referrers in your date range with no date grouping by the “ga:date” dimension. The query:

query_uri = gdata.analytics.client.DataFeedQuery({
      'ids': PROFILE_ID,
      'start-date': sd,
      'end-date': ed,
      'dimensions': 'ga:source',
      'metrics': 'ga:visits,ga:pageviews',
        })

It’s output:

ga:source       ga:visits       ga:pageviews
(direct)        5117    31388
aol     50      386
ask     43      333
bing    677     4705
bookmarks.yahoo.com     2       6
business.com    1       7
google  7937    53270
mws.ask.com     1       10
my.msn.com      1       41
my.yahoo.com    1       3
sxsw.com        33      89
yahoo   875     6789
yellowbook.com  25      104
yellowpages.com 7       68
yelp.com        6       102
... many entries eliminated to protect the innocent

Both of those outputs are valuable. First, the referrers, their referral volume, by date is quite valuable. I’d expect to use the second version to get an overview of aggregate referreral volume without the date segmentation to get a feel for a site’s referral health. I’d use the first version to find out when the referrals were delivered.

Here’s the full code for the Python program that produced the second output:

#!/usr/bin/python

'''

Use the Google Data python module to query Google Analytics

You'll get the "PROFILE_ID" from your Google Analytics account.  From the default
listing you click on the account name and then take the "Edit" action on the website
profile you want to find the profile id for.  It is listed near the top right under
"Profile Settings"
'''

__author__ = 'Leo Edmiston-Cyr '

import gdata.analytics.client
import datetime

USERNAME = 'yourAnalyticsAccount@gmail.com'
PASSWORD = 'youToughPassw0rd'
PROFILE_ID = 'ga:1234567' # the GA profile ID to query
SOURCE_APP_NAME = 'GAGettah' # anything you want to call it
sd = datetime.date(2011,11,21)
ed = datetime.date(2011,11,27)
#ed = datetime.date.today()

def main ():

    my_client = gdata.analytics.client.AnalyticsClient(source=SOURCE_APP_NAME)
    my_client.client_login(
        USERNAME,
        PASSWORD,
        SOURCE_APP_NAME,
        service='analytics')

    query_uri = gdata.analytics.client.DataFeedQuery({
      'ids': PROFILE_ID,
      'start-date': sd,
      'end-date': ed,
      #'dimensions': 'ga:date,ga:source', # enable grouping by date
      'dimensions': 'ga:source',
      'metrics': 'ga:visits,ga:pageviews',
        })
    feed = my_client.GetDataFeed(query_uri)

    # find out if this is the first run through the results
    # to build a simple header for the dimensions and metrics
    firstRun = True
    heading = []

    # we'll run through the data feed reutrned from our query
    for entry in feed.entry:

      # build each row of data from the feed
      row = []

      # pull all dimensions out of this entry
      for dim in entry.dimension:
        if firstRun:
                heading.append(dim.name)
        row.append(dim.value)

      # pull all metrics out of this entry
      for met in entry.metric:
        if firstRun:
                heading.append(met.name)
        row.append(met.value)

      # print the dimension and metric names as the header
      if firstRun:
        print "\t".join(heading) + "\r"

      # print all rows from the feed as they are built
      print "\t".join(row) + "\r"

      # don't print the dimension and metric names as the header again
      firstRun = False

if __name__ == '__main__':

        main()

Want more dimensions and metrics? Visit the GA API dimensions and metrics reference page http://code.google.com/apis/analytics/docs/gdata/dimsmets/dimsmets.html

These articles were aimed at the technical web marketer who either writes a little code or has a programmer they want to egg on.

Here is the simplest example worth trying to get you started

#!/usr/bin/python

'''

Use the Google Data python module to query Google Analytics

You'll get the "PROFILE_ID" from your Google Analytics account.
From the default listing you click on the account name and then
take the "Edit" action on the website profile you want to find
the profile id for.  It is listed near the top right under
"Profile Settings"
'''

__author__ = 'Leo Edmiston-Cyr '

import gdata.analytics.client
import datetime

USERNAME = 'yourAnalyticsAccount@gmail.com'
PASSWORD = 'youToughPassw0rd'
PROFILE_ID = 'ga:1234567' # the GA profile ID to query
SOURCE_APP_NAME = 'GAGettah'
sd = datetime.date(2011,11,21)
ed = datetime.date(2011,11,27)

def main ():

    my_client = gdata.analytics.client.AnalyticsClient(source=SOURCE_APP_NAME)
    my_client.client_login(
        USERNAME,
        PASSWORD,
        SOURCE_APP_NAME,
        service='analytics')

    query_uri = gdata.analytics.client.DataFeedQuery({
      'ids': PROFILE_ID,
      'start-date': sd,
      'end-date': ed,
      'dimensions': 'ga:date',
      'metrics': 'ga:visits',
	})
    feed = my_client.GetDataFeed(query_uri)

    # find out if this is the first run through the results
    # to build a simple header for the dimensions and metrics
    firstRun = True
    heading = []

    # we'll run through the data feed reutrned from our query
    for entry in feed.entry:

      # build each row of data from the feed
      row = []

      # pull all dimensions out of this entry
      for dim in entry.dimension:
        if firstRun:
                heading.append(dim.name)
        row.append(dim.value)

      # pull all metrics out of this entry
      for met in entry.metric:
        if firstRun:
                heading.append(met.name)
        row.append(met.value)

      # print the dimension and metric names as the header
      if firstRun:
	print "\t".join(heading) + "\r"

      # print all rows from the feed as they are built
      print "\t".join(row) + "\r"

      # don't print the dimension and metric names as the header again
      firstRun = False

if __name__ == '__main__':

	main()

Here is its output

ga:date ga:visits
20111121        3214
20111122        2692
20111123        2360
20111124        1537
20111125        2227
20111126        2171
20111127        2220

In the next post I’ll add some dimensions and metrics.

Advertising must change to sustain communities into the future.  I believe I’m on to something.  I believe every community can have a corresponding store (online and/or physical) which becomes indispensable for the community members.  The community is paid for by being the ultimate reminder for the members of where to shop for community related goods.  This is an improvement over traditional ads because the associated store is there for you only when you’re thinking about the community.  Think about most trail/bicycle/water trail stores — they’re equal parts busieness and social network.  This is a workable solution by my measure; one I like.

How about an example?  Try http://www.mdtrails.com  It is a real community I made several years ago as a testbed for some ideas I was putting together at the time — online community around local, Maryland trails, integrating a social component, and bring in a way to make money that is NOT traditional ad based.  I have used Google ads on MDtrails and many other sites in the past — I know it is a quick solution (Thanks Google).  I just think we need to dig deeper to push our communities into the mid 21 century.

So here is an offer and a challenge to a Maryland trail outfitter with a web catalog — let’s partner to tie MDtrails.com (hikemaryland.com too) to your online store (much like I’ve done with an amazon aStore) and let’s test monetization with a real store, with real profit margins, and a sizeable catalog.

Contact me @ digital mountain consulting if you’re interested.

There has been no other time in history when we needed more collaboration on how to make ends meet within our economy than now.  Wait, it gets sweeter.  There is no time in history when there were so many critical thinkers who can process the raw data of business as there are today.  Programmers with 20 years or more programming, managing, and business experience are not impossible to find.  Any good programmer will do if you are willing to put the time into the relationship.

Business men and women: I recommend you consult with a programmer.

Programmers and engineers: I recommend you consult with business decision makers.

We can use business intelligence to pull ourselves out of this economic mess right where we are right now.  Let’s get to work grafting and fusing that org chart!

PS (another teaser) I’m having a great time using the Python-excel libraries at http://www.python-excel.org/ More later.

In the next few blog posts I’ll share some how-to steps for the highly technical analytics/digital marketing professional to automate some of the most tedious steps in reporting.

Of course you want new customers.  In this statement is the assumption you are from planet earth.  If this assumption is incorrect please forgive my ignorance.  So, for us earthlings we believe in growth.  I have, what I believe is, a balanced approach to reaching, satisfying, and speaking to existing customers  and acquiring new ones to achieve growth.

Our gut reaction to spending money on marketing activity to existing customers is “Over my dead body” because it is new customers that directly account for “growth”.  I understand that and find the argument impossible to ignore.  However, I recognize the relationship between current customer satisfaction and the growth of positive word of mouth marketing.  Remember, few things are able to generate a sale like a trusted recommendation.

My approach is always to maximize growth through both new customer connections AND reminding existing customers why your business was the best choice.  Why should they compete?  In the process your advertising creates conversation pieces.

Here is a set of three ads that highlights our approach:

Acquisition focused, ready to sell:

Longhorn Network
Subscribe Today, Introductory Pricing
$69.99/mo. Includes Phone and Internet
CableInc.com

Soft seller:

CableInc TV is Smarter
TiVo Learns What You Like and Don’t
TiVo Suggestions Adapt
CableInc.com

Watch Longhorn Network
Check The Channel Lineup
Only CableInc Has It
CableInc.com

A well designed campaign will provide BOTH ad styles.

In closing, don’t miss the opportunity to BOTH acquire new customers AND enhance the satisfaction of existing customers with your marketing spend.  You don’t have to.

I was not convinced however, that the next step up to a “small” instance was the best path forward for this site.  A micro instance runs about $15/mo. + data transfer and storage and a small runs about $70/mo. + data transfer and storage.  Since the cpu was only exceeding 40% occasionally during short periods of about 5 – 30 minutes about 10 times a week I just didn’t feel good about that price jump.  That got me looking around at alternatives.  Here’s my previous post on that search.

So how did this switch go?  Very well.  I went from maxing out cpu occasionally with corresponding site sluggishness to smooth page loads at all hours of the day — so far.  It has been about four days.

Right now I’m not just happy with a little more horsepower for, believe it or not less money, but the UI of the cloud manager is beautifully simple.  It makes easy things easy.  Best of all is they’ve eliminated all the different kinds of storage — many ebs buckets mounted at different points.  As useful as that has been to me in a previous life (pennswoods.net/atlanticbb.com) for scaling massive legacy systems (email for example) it creates so much work for simply running a website that is fairly busy.

I haven’t formatted an EBS volume, fought with attaching it to my instance, nor tried to figure out an effective labeling scheme to keep track of my system wide (multiple EBS volumes) backups for rackspace cloud.  What about creating AMIs?  No problem.  The running server is persistent!  When you make changes to a server, stop it, restart it they’re still there!  Crazy; I know.  I’ve got my fingers crossed that they keep it this simple and it stays stable.

Here is a cool systematic comparison between the two http://www.thebitsource.com/featured-posts/rackspace-cloud-servers-versus-amazon-ec2-performance-analysis/