John Lovett's Blog at Web Analytics Demystified - Latest Comments

Re: Bulletproof Business Requirements

Samantha Doyle — Wed, 10 Sep 2014 17:55:30 -0000

This is certainly a pain I can relate too. I introduced a brief to assist with the framework but a certain level of education needs to take place with the business user in order for them to understand what is available. Then follows a "I'd like to know everything" statement, followed by my "what will you do with most of that data?". The more savvy the team members become, initially the more they want but eventually, the "want" turns to "need" and the process becomes more succinct and efficient.

I have over supplied in the past, to create team engagement with analytics and educate, in order to achieve clearer briefs in the future, for the most part it works. I really enjoying the requirements gathering and often learn new aspects about the business whilst gathering info - it's great to talk to the teams, especially when it comes to adding extra value to conclusions. Some tips of improving the process would be fabulous and your workshops are of great interest, but I'm in the UK, are their web ex opportunities or plans for Uk workshops at all?

Re: Re-Examining Attribution

Ryan Green — Mon, 11 Mar 2013 13:28:51 -0000

Adobe Insight (Adobe Analytics Premium) is able to track across devices and or browsers for KNOWN customers. The other item you want to consider on top of that is soft metric integrations like email opens or banner views. If you just take basic web click data then you cannot truly do a full media mix allocation based on clicks. I have clients who say many customers open their marketing
emails then just go to the site without clicking...don’t you want to give email that credit? you are paying for that channel! Anyway, we still have a long way to go with these models. I like the way that Derek & John start to lay down the discussion on this which is never easy!

Re: Re-Examining Attribution

Nate Sidmore — Mon, 21 Jan 2013 12:09:03 -0000

Peter this is a very helpful discussion you bring out. I think you've distilled the core problem well (using John as your mouthpiece). At this point it sounds like we each need to decide/determine how misguided we are without cross-device tracking and evaluate if this reasonably informs the educated decision John describes above.

Re: Re-Examining Attribution

Peter O'Neill — Thu, 10 Jan 2013 07:04:41 -0000

Am still in agreement with you here John but my rule of thumb was intended to identify businesses who will or will not get value from campaign attribution. Putting together two of your comments:

"if you aren’t associating known users across devices, then your attribution scheme will largely be misguided" AND "an ecommerce site, chances are that shoppers would only log in when they are purchasing/saving/sharing items"

equates to campaign attribution not being the appropriate solution/approach for (at least the majority of) ecommerce businesses. I would prefer to see the resources and brain power of vendors, consultants and organisations focused on developing solutions for these organisations with an output that is not misguided but will deliver the most value.

Re: Re-Examining Attribution

johnlovett — Tue, 08 Jan 2013 16:31:36 -0000

Hey Ken, Step 4 is typically where you try to punch holes in the model. For marketing managers, it might be as simple as recognizing that paid search is getting undue credit, but that you will continue to invest because it provides a strong source of qualified traffic. I look at this step as a means to either justify what you're doing with business logic -- or -- alter course based on the data. In either case, your making a data-informed decision. But ultimately you are correct in that it helps to know how you will conduct analysis before embarking down the path of attribution.

Thanks for the comment!

Re: Re-Examining Attribution

toddbdac — Tue, 08 Jan 2013 14:52:59 -0000

Thanks, John, for the timely post and upcoming webcast. Looking forward to it! I am one of the analysts who thinks that all attribution is flawed (what isn't flawed if you're talking black/white?). I'd argue further, though, that as you reduce the flaws, or compare flawed attribution models against one another, there is much to learn/gain from the experience.

My favorite advice here is definitely the out. Determine whether advanced attribution logic is needed. If not, don't get mired up in it. Set a date to re-visit, or set up alerting, and check back in on advanced attribution logic at that point.

I think the spirit of Ken's comments is spot on, too. Anyone looking to combine and aggregate data from multiple sources needs to have an idea of how that data will be analyzed in order to be successful in the venture. That said, the flipside is that if you don't start you'll never get there. Each of these steps, especially step 4, I see as iterative... there will always be new data sources, new data within those sources, new ways to analyze, etc.

Re: Re-Examining Attribution

Ken Pendergast — Tue, 08 Jan 2013 11:51:48 -0000

John, I think you defined a great model. For Step 4 Do you have any examples that a marketing manager could use to overcome any obstacles of not having the super savvy data analyst on staff?

I am having attribution discussions weekly and find that in most cases the marketing manager types are at this point and unaware on what to do next or they are jumping right to step 5 and making some very large decisions with a minimal understanding of methodologies for step 4.

In most cases I will not see what they did for their analysis until they are trying to implement. The efficacy of attribution modeling is highly dependent on the area in which the model is applied. The guys over at simply statistics are always bringing up statistical incompetence; I think that is a good assessment of people using this model without truly knowing what step 4 looks like.

Within this model I feel that people should have to fully define the methods for step 4 before they move passed step 1. Especially if they have offline transactions using online promos, something like revtrax is crucial for aggregating data. It’s difficult to understand how they can be acting strategically on attributions without the full landscape of conversions.

Lastly, I think the “messy or flawed” perception is driven by analysis paralysis. You have 5 different models, a large marketing mix, channel variables, and publisher business models. Defining step 1 and 4, before even taking action will help unclear the air for attribution.

Re: Re-Examining Attribution

johnlovett — Tue, 08 Jan 2013 08:47:46 -0000

Hi Blair! Thanks for your comment. I too found in my research that many marketers get stuck on step 4 above when it comes time to choose an attribution model. The question, How do I know which model to use when? often surfaced as *the* burning question. While, nearly everyone I spoke with had a different answer, I found that the solution was to look at multiple views of attribution to answer specific questions. Now, it's challenging to do this without distorting (torturing?) the data to make it say what you want...that's where the ethical, unbiased analyst seeking the truth kicks in.

That said, even the best attribution models must be revisited at regular intervals. As product offerings change, consumer behaviors change, and macro-environmental factors shift, so too must attribution models. A best practice is to revisit your model every 6 months to ensure that it's still relevant. Hope this helps! (PS: I dig you blog btw :)

Re: Re-Examining Attribution

johnlovett — Tue, 08 Jan 2013 08:38:26 -0000

I'm wary of applying a rule of thumb here. Here's why:

1) If you operate an online bank where account holders log in to access their details > Then Yes, a majority of visitors would log in every visit

2) If you operate an ecommerce site, chances are that shoppers would only log in when they are purchasing/saving/sharing items > So, No

Like many things in digital analytics, it depends. Knowing your business (and your data) is key to success.

Re: Re-Examining Attribution

Peter O'Neill — Tue, 08 Jan 2013 05:50:29 -0000

Thanks for the reply John and I think I agree with everything you have said. Can you just clarify on an early sentence "if you aren't associating known users across devices, then your attribution scheme will largely be misguided". Would you agree a decent rule of thumb could be "the majority of your visitors logging in (or otherwise identified) in the majority of their visits?"?

Re: Re-Examining Attribution

ReevesJB — Mon, 07 Jan 2013 14:56:14 -0000

Attribution analysis is definitely a component of any mature marketing organization's bread-and-butter marketing analytics these days, but we've encountered a lot of bewilderment about the very basic models you mention here. "Should we use first-click? Equal credit? A custom rubric?" And so on. When the process hits #4 on your list above, you're already beyond the capacity of most marketing orgs, particularly in the mid-market. So marketing attribution tools must continue to simplify, because creating the automated attribution modeling that most orgs need isn't necessarily that complicated, but selecting the insight from the noise can be.

What is important is that your marketing analytics team is clear on what, exactly, they're trying to attribute, and for whom. I heard so much nonsense about last-click models being totally useless that I went ahead and wrote a blog post about it: http://www.smartercommerceb...

Looking forward to the webinar!

Re: Re-Examining Attribution

johnlovett — Mon, 07 Jan 2013 14:08:48 -0000

Hi Peter,

Thanks for your comment, and for pointing out a limitation of my Attribution Litmus Test. You’re right in that using this method without being able to identify users across devices would be limiting. My take is that if you aren’t associating known users across devices, then your attribution scheme will largely be misguided.

You point this out very well in your presentation and blog post. However, companies using advanced attribution should be thinking about associating their users across devices and utilizing login capabilities where feasible. Granted this isn’t always possible, but it’s my opinion that advanced marketing requires knowing and recognizing users across devices, because that’s the reality of the world we live in today. We will address this to a certain degree in the webcast, but my main point will be, despite the science behind your attribution solution, the result must be used in the context of your business. Understand the limits of what you’re solving for and how your tools consider these data, then make an educated decision in your marketing adjustments. While this is where you may cry, then why the need for attribution!?! I advocate that done right, this information is far better than propagating marketing efforts based solely on "first" or "last" click.

My $.02.

Re: Re-Examining Attribution

Peter O'Neill — Mon, 07 Jan 2013 12:51:13 -0000

Hi John - I am one of those that believe all attribution models are flawed. My simple example is similar to your Attribution Litmus test, based on the visit number in which a conversion was achieved. The scenario is simply:

-> a website has a high proportion of visitors researching before purchase - 80%

-> 75% of these researchers do so on their work computer before purchasing at home (first visit on that device)

-> any web analytics tool (assuming the visitors were not forced to login in order to research) would display 80% of purchases occurred on the first visit

-> the business strategy would be based on converting visitors on their first visit (as that is what the data informs is visitor behaviour) with this being a sub optimal strategy (if true behaviour was known)

Is this issue addressed in the whitepaper/webinar? I detailed out these issues, what I believe the real business questions to be and alternative approaches at the recent London eMetrics - presentation can be found at http://www.l3analytics.com/....

Re: Fortune 500 CEOs Aren’t Social — Ummm…Thank You!

jacqueswarren — Fri, 20 Jul 2012 08:29:17 -0000

You're not alone! And I would add that any C-Suite dude with a Klout score over 60 should be under their CEO's scrutiny as potential slackers...

Re: Fortune 500 CEOs Aren’t Social — Ummm…Thank You!

randyzwitch — Mon, 16 Jul 2012 11:46:49 -0000

Couldn't agree more John. When I saw the article (besides being turned off about the not-so-subtle advertisement of James' company), I wondered "Why WOULD I expect CEOs to be playing in social media?"

When you look at what companies make up the Fortune 500, many of them aren't B2C companies; why would they be interested in hearing the comments of a single person/non-customer? What benefit does having energy company CEOs on Twitter? To take a nonstop barrage about how oil is bad? Or non-branch banks/financial like Berkshire Hathaway and TIAA-CREF?

Like you, I'm glad most Fortune 500 CEOs have better things to do than Tweet (or, just have more discretion?).

Re: New Research on Social Marketing Analytics

Scott — Thu, 29 Dec 2011 22:34:29 -0000

Hi guys, I am a Phd student, trying to find the latest research ideas in Social media marketing, throw me as many ideas as you can, thanks.

Re: Counting ROI in Pennies with Social Media

deanshaw — Sun, 18 Dec 2011 23:59:33 -0000

Appreciate the debate. Here's the best long-form comment I can muster late on a Sunday night http://wp.me/pKLGG-6g #WhoopDeeFrigginDo ;)

Re: Google’s New Social Data Hub

datalicious — Mon, 12 Dec 2011 03:32:20 -0000

This sounds a lot like the fruits from the Postank aquisition http://techcrunch.com/2011/... Great platform, already had a GA integration

Re: Google’s New Social Data Hub

Justin Cutroni — Thu, 08 Dec 2011 07:08:35 -0000

John,

This is a great writeup. In fact, I just quoted you 5 times when posting my thoughts on this new development.

I absolutely agree that this will be a huge win for marketers and anyone looking to get the complete picture of social media performance. I think it really showcase the value of off-site/on-site data integration. Having one tool that will show me the trend of content distribution, then traffic from that content, and finally revenue from that content, is a very exciting thought.

I would love your opinion on PII. I expect this feature to include social media account names, which could very well be PII. This is something that I point out in my post.

Justin

Re: Google’s New Social Data Hub

KeithBurtis — Wed, 07 Dec 2011 17:27:54 -0000

John, My favorite part of that article, "Typically this yields low-hanging counting metrics which can be of some value, but more often than not end up as isolated bits of information that don’t provide business value."

Right on! Much of this is actually doable right now through referrer segmentation. It will be very interesting to see how Googles new and upcoming offerings will pump that up! Now, you talk about this social data hub being a potential boon for new and upcoming platforms. I'd be interested in discussing that more as it seems that lately hoarding network data seems to be the decision du jour.

Thanks for the great wrap-up. Looking forward to digging in more!
@KeithBurtis

Re: Moneyball Will Put Web Analytics on the Map

John Lovett — Tue, 27 Sep 2011 20:29:36 -0000

@Rudi ~ Me too. I gave my neighbor a copy of my book just to be nice, and I don't think he's opened his blinds since ;-)

@David ~ Better late than never!

@Alice ~ That's awesome! I went to see it with my wife on Friday too and she turned to me and said ... "Wait, are you supposed to be Jonah Hill?!" I guess she wanted Brad Pitt. Ahh well...

Re: Moneyball Will Put Web Analytics on the Map

Alice Watson — Tue, 27 Sep 2011 20:03:07 -0000

We had a team outing at work last Friday to see this movie, and I joked that it was work related as it explained what I did. Somehow my co-workers didn't buy it though.

Re: The Social Technology Spectrum

Joomla Developers USA — Fri, 23 Sep 2011 06:50:38 -0000

Hi really thanx for the article i am looking for the information got this article very informative .

Re: Moneyball Will Put Web Analytics on the Map

David McBride — Thu, 22 Sep 2011 17:35:40 -0000

John,
I just recently read Moneyball and had the same thoughts on its parallels to our industry. As is often the case, I was eight years late to the book, but I liked it so much I bought copies for my entire team. We're going to have an Analytics book club one of these days. Maybe I can even expense my movie tickets this weekend!
Thanks for the post.
David

Re: Moneyball Will Put Web Analytics on the Map

Rudi Shumpert — Thu, 22 Sep 2011 02:06:44 -0000

John,

I sure hope it puts our industry on the map. Or at least I'd settle for my neighbors to be able to grasp a better understanding of what we do ;)

-Rudi