You are doing Google Analytics all wrong. Here is why

 

I have dealt with hundreds of Google Analytics accounts in my career and have seen lot of issues from incorrect tracking code, selecting wrong KPIs to analyzing data without custom reports & advanced segments.

All of these issues prompted me to write the post Google Universal Analytics Setup Checklist and Common Universal Analytics Mistakes that kill your Analysis, Reporting and Conversions. 

But these posts don’t really solve the biggest problem of all in web analytics:

‘Misinterpretation of Analytics data’.

Everyone seems to be making this mistake of crediting conversions and e-commerce transactions to wrong acquisition channel and that too again and again.

Many marketers can’t help themselves because they believe that the reports provided by Google Analytics (& other web analytics softwares) are

what you see is what you get’.

But it is actually

what you interpret is what you get.

Majority of businesses and marketers even today give credit for conversions and ecommerce transactions to the last campaign, ad or search that referred the visitor when he/she was converted.

This has resulted in marketers taking wrong business decisions and losing money.  

All of the data you see in Google Analytics Reports today lie to you unless you know exactly how to interpret it correctly.

Let us consider three different scenarios:

 

Scenario-1:

scenario-1

Majority of marketers looking at this standard ‘All Traffic’ report in Universal Analytics of the last 3 month will draw following conclusion:

Organic traffic is playing a second fiddle to direct traffic. Majority of traffic and revenue is coming through direct traffic. We need to speed up content development and link building.

 

Scenario-2:

One look at this monthly PPC report and many of you will declare this whole campaign a total failure.

Look at the first campaign, just one conversion in the whole month and cost per conversion is whooping $531. You must be kidding, right.

 

Scenario-3:

scenario-3

Do you really think your brand name generated revenue of more than $241k?

Welcome to the Real World

Let us analyse these three different scenarios once again but this time in the real world.

 

Scenario-1:

 scenario-1

Truth about direct traffic

All untagged or improperly tagged marketing campaigns from display ads to emails could be treated as direct traffic by Google.  

Whenever a referrer is not passed, the traffic is treated as direct traffic by Google.

  • Mobile applications don’t send a referrer.
  • word/PDF documents don’t send a referrer.
  • 302 redirects sometimes caused the referrer to be dropped.
  • Sometimes browsers don’t pass the referrer.
  • During http to https redirect (or vice versa) the referrer is not passed because of security reasons.

All such traffic is treated as direct traffic by Google.

So on the surface it looks like 618,199 visits/sessions were direct, but it may actually be only 25000 sessions which were from direct traffic and the rest were from display ads, email, organic,  social media and applications/campaigns in which the referrers were not passed.

But this analysis does not end here because you are still not looking at the complete picture.

Here is the complete picture:

scenario-1.1

Visitors do not always access your website directly before they make a purchase.

They are generally exposed to multiple acquisition/marketing channels (like display ads, social media, paid search, organic search, referral websites, email etc) before they access your website directly and make a purchase.

So if you are unaware of the role played by prior acquisition channels, you will credit conversions and e-commerce transactions to wrong acquisition channels, like in the present case to direct traffic.

If you look at the chart above organic search is playing a key role in driving direct traffic to the website which eventually resulted in conversions and ecommerce transactions.

But to get such type of understanding you need to understand and implement Attribution Modelling.

So the conclusion that organic traffic is playing a second fiddle to direct traffic is incorrect.

Related Post: Geek guide to Direct Traffic Analysis

 

Scenario-2:

Visitors do not always click on your paid search ads before they make a purchase.

They are generally exposed to multiple acquisition/marketing channels before they click on your ad and make a purchase. Sometimes visitors may click on your ads but make a purchase through different acquisition channel or medium.

For example a person may click on your paid search ad through laptop at work. Then later make a purchase via his home desktop PC through a branded organic keyword.

Sometimes your paid search ads play a bigger role in assisting conversions than directly resulting in conversions.

The ‘conversions’ and ‘cost per conversions’ (cost/conv) reported by Google Adwords in scenario-2 above, are all based on last adwords touch attribution model (people click on ad and buy) and hence provide poor analytical insight.

When I paused these campaigns, I saw a huge decline in revenue.

These campaigns are in fact very profitable and there assisted conversion value is also very high.

Related Post: Google Adwords Analytics – Complete Guide

 

Scenario-3:

scenario-3

This scenario is not any different from scenario 1 and scenario 2. Here too you don’t see the complete picture.

Visitors do not always search for your brand name before they make a purchase.

They generally start their search with a non branded and generic search term then they refine their search queries as they get better understanding of what exactly they are looking for.

Sometimes they make a purchase right after making a search but often they come back later to your site via a branded search term.

Since website/brand name is easiest to remember among all branded search terms, it often ends up being attributed lot of conversions and transactions.

 

Takeaways

1. Web analytics reports are not ‘what you see is what you get’. It is ‘what you interpret is what you get’.

2. Direct traffic is polluted. So find ways to clean it. The first step should be to correctly tag all of your campaigns URLs. Use Google Analytics URL Builder.

3. Visitors do not always access your website directly before making a purchase.

4.Visitors do not always click on your paid search ads before making a purchase.

5. Visitors do not always search for your brand name before making a purchase.

6. Understand the role, various website referrals, social media, display, email, paid/organic search etc played prior to conversions via Multi Channel Funnel Reports before you discard/label any marketing channel as ineffective or over invest in any particular channel.

7. Understand how different acquisition channels work together to create conversions and transactions.

No one acquisition channel is solely responsible for sales in the world of multi channel marketing. 

So do not overestimate or underestimate the impact of other marketing channels.

8. Understand that when you change the budget of one acquisition channel it will have impact on the performance of other acquisition channels. Nothing is black and white in the world of analytics.

Related Post: How to allocate Budgets in Multi Channel Marketing

 

Attribution modeling is programmatically very difficult to implement at present but Google has taken one right step in this direction through ‘Multi Channel Funnel Reports’.

If you wish to learn more about attribution modeling and attributing conversions and ecommerce transactions to the right marketing channel then check out the following articles:

  1. Google Analytics Attribution Modeling – Beginners Guide
  2. 6 Keys to Digital Success in Attribution Modelling
  3. The Geek Guide to implementing Attribution Modelling
  4. Advanced Attribution Modelling in Google Analytics
  5. Google Adwords Attribution – Complete Guide

 

Other Posts you may find interesting:

Join over 5000 subscribers!
Receive an update straight to your inbox every time I publish a new article.

 

About the Author:



My business thrives on referrals, so I really appreciate recommendations to people who would benefit from my help. Please feel free to endorse/forward my LinkedIn Profile to your clients, colleagues, friends and others you feel would benefit from SEO, PPC or Web Analytics.

 

 

  • http://www.njbsecurity.com iliyas

    Great insights. Really too much helpful and you describe in simply way.
    thanks

  • http://www.innovativeseogroup.com Amish Keshwani

    Hey Himanshu,

    Another stunning post :)

    This has really cleared my doubts about direct traffic, & I realize after reading this post that, recently I saw major improvement in direct traffic. This is a great set of information.

    Even earlier post on Multi Touch Attribution Model was also really super. I am now more exited about analytics part. Thanks again for sharing this great knowledge.

  • http://www.sparkinfosys.com spark

    well composed words good explanation it made me to think back . thanks for the post

  • http://www.imod.co.za Chris M

    Thank you very much for the post Himanshu, what you say is so very true and far too many people prefer the WYSIWYG approach. I’ve seen conversions increase by hundreds of percentages when your approach is taken and things aren’t taken at face value. Tagging every campaign is incredibly important, breaking down that ‘direct traffic’ is one of the most important pieces of advice for anyone getting into Analytics!

  • http://www.e-zen.be Béate Vervaecke

    Shouldn’t you add coming “from desktop apps”? “Google Analytics treats all the traffic that comes from untagged shortened ULRs on social media networks like Facebook and twitter as direct traffic”

    • http://twitter.com/Muyuan_Ma MuyuanMa

      Hi Béate,
      I totally agree with u on this point! Only if visitors are using social media mobile app/ desktop client tool, they will have untagged shortened ULRs recorded as direct traffic. Web browser based social media will take whatever social media platforms as referrer.

      D o you agree?

    • http://twitter.com/Muyuan_Ma MuyuanMa

      Hi Béate,
      I totally agree with u on this point! Only if visitors are using social media mobile app/ desktop client tool, they will have untagged shortened ULRs recorded as direct traffic. Web browser based social media will take whatever social media platforms as referrer.

      D o you agree?

    • http://twitter.com/Muyuan_Ma MuyuanMa

      Hi Béate,
      I totally agree with u on this point! Only if visitors are using social media mobile app/ desktop client tool, they will have untagged shortened ULRs recorded as direct traffic. Web browser based social media will take whatever social media platforms as referrer.

      D o you agree?

    • http://twitter.com/Muyuan_Ma MuyuanMa

      Hi Béate,
      I totally agree with u on this point! Only if visitors are using social media mobile app/ desktop client tool, they will have untagged shortened ULRs recorded as direct traffic. Web browser based social media will take whatever social media platforms as referrer.

      D o you agree?

  • http://www.barker.dj dan barker

    hi, Himanshu, this is a really well written post with some great arguments, but 2 of the 3 scenarios are wrong.

    I’m not keen on posting negative comments, but I thought it sensible to point out the misunderstandings.

    Scenario 1:

    In your example, you say that when visits come through ‘organic’ and then later return through ‘direct’, the sale will be attributed to the ‘direct’ traffic source. That isn’t right. Google Analytics means that those sales will be attributed to ‘organic’.

    Check this yourself: Visit your site through an obscure organic keyword. Revisit later via direct. You’ll see that you have 2 visits for the obscure organic keyword. In other words: if you visit ‘direct’, but Google Analytics can see a prior ‘known’ channel that you came through, it overwrites ‘direct’ with whatever the known channel was.

    Scenario 2:

    Here you say: “The ‘conversions’ and ‘cost per conversions’ (cost/conv) reported by Google Adwords in scenario-2 above, are all based on last touch attribution (people click on ad and buy) and hence are providing poor analytical insight.”

    This isn’t right. As standard AdWords does *not* use ‘last click’ attribution. The standard AdWords conversion tracking system has little/no knowledge of other channels, therefore it simply reports any conversions within the cookie period (30 days if I remember right).

    dan

    • http://www.seotakeaways.com/ Himanshu

      Hi Dan!
      Thank you for your comment.
      Scenario-1: Google Analytics by default attribute conversions to the most recent campaign (i.e. last touch attribution). http://support.google.com/googleanalytics/bin/answer.py?hl=en-GB&answer=55556

      Scenario-2: Cookies play no role in tracking a visitor in case of different devices/machines/browsers. So if you click on an ad from a laptop computer at work and then later make a purchase through another computer say your home desktop, the cookie set up earlier wont be able to identify you on other computer as the same visitor. We live in a world where people use different devices at different point of time while searching for a product and making a purchase

      • http://www.barker.dj dan barker

        hi, Himanshu,

        Scenario 1: In the case of ‘known channel vs known channel’ that’s right, but in your scenario ‘direct’ is the last touch. Where that is the case, and there are prior ‘known’ channels, the sale (and the whole visit) are attributed to those prior known channels.

        Scenario 2: You say “The ‘conversions’ and ‘cost per conversions’ (cost/conv) reported by Google Adwords in scenario-2 above, are all based on last touch attribution (people click on ad and buy) and hence are providing poor analytical insight.”

        This is wrong – AdWords conversions are not based on last touch.

        dan

        • http://www.seotakeaways.com/ Himanshu

          Hi Dan! GA tracks all the visits to your website as direct in which people arrived at your website directly i.e. by either typing the URL or via a bookmark. You say that conversions are not attributed to last touch in case of direct traffic. Even if this is correct, then the marketing channels which can be attributed conversions in case of direct traffic are either first touch or middle touches. Since GA doesn’t credit attribution to either the first touch or middle touches in stand reporting, your hypothesis doesn’t hold true. To keep reporting simple, GA credits conversions to the most recent campaigns. Same is the case with scenario2. I am afraid you are not looking at things the way i am looking at. Conversions paths are not that simple as you may think they are. It is not like someone clicked on your PPC ad without being exposed to any other marketing channel and then within 30 days made a purchase from the same browser and machine he used earlier to visit your website for the first time. Moreover i never used the word ‘all’ anywhere in scenario-2. That is your interpretation. Instead of dissecting my scenarios, please try to understand what i am trying to convey here.

    • analyticsNerd

      Hi Dan! I don’t seen anything wrong with these scenarios. Google does not treat direct traffic differently. To be honest you are the first person to say that. Had this been the case many analyst would have raised this issue by now. Google adwords doesn’t always provide the acutal picture. Often there are data discrepancies between the Google Analytics and Google Adwords reports regardless of cookies. And whenver this happens, i trust Google Analytics more than Adwords.

  • http://www.barker.dj dan barker

    hi, Himanshu, thanks.

    I was not trying to make any broader points about multi-channel/multi-touch attribution. I was simply pointing out 2 specific errors in the blog post (each of which could mislead readers who don’t know otherwise about how GA/AdWords report things). Those still both stand.

    “Moreover i never used the word ‘all’ anywhere in scenario-2. That is your interpretation” – I had copy/pasted you directly, so it can’t have been my interpretation.

    Apologies if my ‘dissecting’ offended you – I thought it worth pointing out the ‘wrong’ bits in an article about things being ‘wrong’!

    I hope life is good with you,

    dan

    • http://www.seotakeaways.com/ Himanshu

      Hi Dan! Your theory that GA does not credit conversions to the last touch in case of direct traffic but in fact credit conversions to the second last touch is incorrect. There is no such thing as ‘second last touch attribution’. This explains nothing is wrong with scenario-1. Now comes the second scenario. Google adwords reports by and large provide incomplete picture of conversions not only because of the discrepancy in the reported data between Google Analytics and adwords but also because of the way visitors make a purchase. And don not underestimate the readers of this blog, they are very intelligent people. They will rip my blog post apart if i try to mislead them :)

      • Gabor

        Hi Himanshu,

        I just wanted to make Dan’s point a bit clearer. I think he meant this:

        **Direct visits do not override campaign cooike information.**

        This means that if the direct visit was preceded by a visit from an Adwords or other campaign with autotagging on or properly tagged url’s, like an email campaign, Facebook, etc., the conversion will be attributed to the last campaign source, not to direct.

        In the video below, Avinash Kaushik talks about this special case, from 14:40.
        http://www.youtube.com/watch?feature=player_embedded&v=rMpEfqMp454#!

        Again, note that it only stands if the preceding campaign was properly tagged with the utm_campaign parameter.

        • http://www.seotakeaways.com/ Himanshu

          Hi Gabor! i would like to highlight one important statement from this video made by Avinash ‘in this particular narrow specific scenario’. What you have mentioned about ‘Direct visits do not override campaign cooike information’ is not always the case. But thanks for this video.

  • analyticsNerd

    Hi Himanshu! Love this post. I don’t think anyone can explain multi touch attribution better than this. I am looking forward to read more posts like this one from you. Keep up the good work.

  • Jitender

    Brilliant as usual. I love the way you explianed another invaluable topic of ‘multi touch attribution’ that isn’t covered well elsewhere. I have a question. We manage several websites of one client and the client want us to track traffic from these websites as direct traffic instead of referral traffic. What do you suggest? Thanks in advance.

  • http://www.sparkinfosys.com/ Niharika

    Really Great information. Thanks for the Great post

  • http://twitter.com/mikegracen Mike G

    Excellent info Himanshu. Particularly eye-opening “Google Analytics treats all the traffic that comes from untagged shortened ULRs on social media networks like Facebook and twitter as direct traffic. So if someone clicks on one of your untagged tweet and made a purchase on your website, Google will credit the conversion to direct traffic instead of the poor twitter.” I have a new action item :-)

  • https://twitter.com/#!/sajeetnair Sajeet Nair

    Visits from Untagged URLs from Twitter and other social media channels are attributed to Direct Traffic? Then what about the data that is shown in referral?

    Am I missing out on something here?

    - Sajeet

    • http://www.seotakeaways.com/ Himanshu

      yes you missed the word ‘shortened’. It is untagged shortened URLs :)

  • http://www.intellectsoft.co.uk/android_application_development.html Gordon

    Great hints. It’s really important to know how to do Google analytics in the right way. Otherwise you may get controversial results.

  • http://izoom.com Greg R

    Your take away point #8 cannot be over stated to people responsible for creating analytics reports; problem is those who receive the reports often don’t want to know the data’s fuzzy. If they do know, they often want it interpreted to satisfy their existing conceptions of what’s working/not working.

    Properly tracking links is one very important component, as you noted. The current Google Analytics link builder is located here (your link in #2 goes to a builder for deprecated GA code):
    http://support.google.com/analytics/bin/answer.py?hl=en&answer=1033867

    • http://www.seotakeaways.com/ Himanshu

      Thank you for the comment and the link Greg. I have updated the Google Analytics Link Builder.

  • http://twitter.com/brainresidue Quimosabe Brain Rez

    Great stuff… not sure how I found this site, but I’ll be back!

    • seohimanshu

      Glad you like it. See you soon :)

  • Fahmi Mohammed

    Hi Himanshu,

    I must say this is one brilliant post with regards to multi-touch world. Please keep it up.

    On another note, I was wondering why do we sometimes see duplicate transactions (e-commerce tracking) recorded under Google/Organic and Direct/None? Auto tagging is enabled and everything but can’t seem to connect the dots here…any ideas? If it was between google/cpc and direct/none it could be down to tagging..but this is not the case here.

    Thanks!

    • seohimanshu

      I am not completely sure about your question. Google analytics by default uses the last non direct click attribution model. So conversion is attributed to the last non direct click.

  • Victor – Absinthes.com

    Hi Himanshu,

    Thanks for your post.

    I’ve got a question for you that has been stuck in my head for a few weeks.

    It happens often to me that I see a conversion on Google Adwords, but don’t see in Google Analytics. Do you think it is due to the 302 redirection or do you have another explanation?

    I can’t find a reason why I somehow “lose” the tracking, except the 302 redirection.

    Any thoughts?

    I’d be happy to hear your thoughts.

    Victor

    • seohimanshu

      Google adwords record conversion at the time of click on an ad. So if you click on an ad on say July 1 and then convert on July 4, Google Adwords will report conversion occurred on July 1. Whereas in Google Analytics, conversions are reported at the time of conversion. So if you click on an ad on July 1 and then convert on July 4, Google Analytics will report conversion on July 4

      So what is happening is that the conversion you see in Google Adwords has already been reported by Google Analytics.

      • Victor – Absinthes.com

        Hi Himanshu,

        Thank you for this great answer.

        I will look into my data and see when those conversions happen.

        Have a great day!

        Victor

  • Gnosis Media Group

    You do a fabulous job explaining a complex situation with visual aids, Himanshu. Do you know of an attribution model (could be custom built) that takes into account when a paid search ad isn’t clicked but contributes to a conversion later via organic in the middle and then at last via direct? This is what we’re seeing a lot for one of our products.

    • seohimanshu

      Thanks for the kind words. Please look at the Top conversion path report in your Google Analytics account. That will solve your problem. No need for attribution model here. Also check this post: http://www.seotakeaways.com/8-super-awesome-google-analytics-conversions-segments/ and the section ‘How to determine the Effect of PPC on Organic Search Traffic’.

      • Gnosis Media Group

        The Top Conversion Path report will include a paid search (not display) ad even if it’s not clicked?

        • seohimanshu

          No

          • Gnosis Media Group

            Now that we’ve gathered more data, I think what is happening is what you mention above: “During http to https redirect (or vice versa) the referrer is not passed because of security reasons”. Referrer is getting dropped sometimes because this particular product page starts off http and then goes to https (on a totally different proxy subdomain, no less) for credit card checkout. We’ve added cross-domain tracking in Analytics, but still nothing registers in Adwords under Click Conversions. I spoke to 2 diff Adwords analysts on this, and they both think that the tracking parameters are being dropped when crossing over to the https subdomain. We know the ads are a necessary (but not sufficient) cause for conversions – because when we don’t run them, even though we get the same amount of non-paid traffic, it doesn’t convert. But measuring the precise connection and weight of the ads is very difficult to do.

            • seotakeaways

              Consider making your whole website secured with https. That way there will be no http to https redirect issue.

      • Gnosis Media Group

        The thing is, we are seeing what this article explains: organic search traffic is increasing when paid ads are running. We are trying to create a custom attribution model in GA to account for this: http://www.mediatwopointoh.com/the-affect-of-google-adwords-on-your-seo-efforts/#axzz2rKI2pbqP

  • Kingi Gilbert

    Thanks, great article, I will post about it on my site! :) This is essential stuff to know for any advertiser/marketer!

  • catmario online

    This article was highlighted in a most recent email from analytics and drew
    me toward it.

    Have to agree with the points, albeit point One, is far less effective today
    than it was 9-10 months ago when google started masking keywords for people who
    have signed into google. In fact that seriously hampers

    point One, and requires a lot of sophisticated work arounds to grasp this
    critical point. Unfortunately it will continue to get worse over time.

    We definitely focus on point 5 for a critical point. The above referenced
    business and some of our others are very attached to search and because of that
    naturally google. We work all the time to review how our off line efforts
    reflect bottom line results.

    Just reviewed a business re: August results this year and last. Noticed a
    marked difference and one reflection had to do with some offline advertising
    which may or may not have contributed to a significant

    change in revenues.

    Now here is a different perspective that we had using–not only analytics, but
    our in house analysis of results and adwords along with analytics.

    The business is local and regional. We noticed “impressions” was way down for a
    critical geo portion of the market in adwords. Other locations were consistent.

    We checked analytics for the two comparative time periods …and yep not only
    less people thinking about the topic via adwords…but the geo portion of
    analytics told us less visits to the site.

    We had less sales that corresponding month from that target geo region.

    Meanwhile in the comparative month…we had run a newspaper campaign. Definitely
    insights that helped us with regard to understanding off line results.

    catmario online

  • http://www.allmoh.in/ SEO Services

    There are some usable
    parts but it seems mostly geared sales and selling. Social media ROI also
    should include the savings obtained through efficiency of reach. It helps me nurture relationships that
    improve the value of the
    services we provide. The long term customers, clients, prospects and associates need multiple channels of
    communications that are easily accessible.

    go
    here
    thank you

  • http://seoexperimentz.com/ Suraj G

    Himanshu, as usual great post. I have a question about scenario 1. I know that my direct traffic is not what I see there and when I look at MCF reports I get the conversion paths that contributed to direct traffic. Now my question is looking at the MCF report, how can I re-arrange the data in a three column table of Organic, Direct and Referral to ensure I credit conversions to right medium.source. Can we do it by channel grouping?