Subscribe to Vincent Granville's Weekly Digest:
What is the difference between web mining and web analytics? What kind of degree or experience do you need to do web analytics? What kind of software do web analysts use? Do web analysts write code? Are they business analysts, but focused on web metrics? Is text mining / machine learning part of web analytics? What is the difference between web analytics and advanced web analytics?

Tags: web analytics

Views: 317

Reply to This

Replies to This Discussion

Hi, Vincent -- here's a months-delayed response to your post....

Web analytics is more about getting insight out of the data, whereas web mining, at least as I use the term, would be using some sort of technique to sift out specific information based on an a priori expectation of what you will find.

That is, WA may be more exploratory in nature, and so on-the-fly segmentation tools are very important. Web mining on the other hand brings to mind the integration of CRM purchase history and other non-web data to calculate segmented LTV (for example), because you have a specific email campaign in mind to boost customer loyalty to a specific segment.

Analysts spend much of their time enforcing process QA -- such as ensuring your web agency partners put the right tags on every asset, and that your SEM and email partners correctly tagged the links so you can track your traffic allocation metrics. Time is also spent helping stakeholders understand what the web analytics tools can do, and what they can't, then providing reporting, KPIs, dashboards, and other tools to map decisions and action to strategy.

Generally, text mining (content analysis) and machine learning are not part of web analytics -- despite vendor claims, the market is not that advanced. Certainly, the vast majority of customers have not yet identified the value that these tools might yield.

"Advanced" web analytics would, in my experience, include things such as multivariate testing, behavioral targeting, qualitative surveys, and other related tools, besides the traditional web analytics realm that includes measuring the online components of a typical marketing mix (email, banner / display ads, affiliate programs, SEM, SEO, plus the buzz around social networks, viral videos, online communities, blogs, podcasts, etc.).

While a business analyst mind set is useful for all of these, in terms of being able to prioritize stakeholder requests and allocate scarce resources, the technical skills useful to a data miner are really overkill for most web analytics projects that the typical marketing client might realistically consider.

I'd love to hear from others who have differing opinions about this, as well as how anyone thinks this may change in the next few years.

WDave
I actually know some stuff on web mining and web analytics. Old information can still be useful.

Web Mining is based upon techniques and finding patterns. I can tell you it is similar to Business and Financial Analysis. Based upon the business objective, you gather the relevant data. Then, search for the patterns. Lay them out. Analyze and forecast. It could be broken into 3 groups. For an easier understanding, you shouldn't ask me what they are, because I'm not an expert in the field. So my analogies would probably make you laugh. But I am sure that a Systems Architects, Programming Managers and other IT Professional would be more capable of going into the details with you.

Web Analytics is focused. It is the measurement and maintenance of a website Internet, Intranet and/or company online marketing campaigns, etc.... Web Analytics in can be captured immediately. It is ever changing too. Stats are readily available through third party servers. When the WWW first came into mass markets there were IT Departments and Traffic Departments. Traffic Departments worked directly in ensure data is properly captured, monitoring campaigns, consumer/b2b traffic, gathering statistics off of the servers. They also sometimes become responsible for the tagging of graphics.

I would say anyone educated as a Business or Financial Analysis would be experienced enough to do Web Mining. They would have to be open to a new language (computer jargons and mechanics). For Web Analytics, anyone with a Computer Science & Engineer background would do well.

Web Analysts read data off of servers. It could be downloaded on proprietary software, databases, or MS Excel with access. It would depend on expertise of the IT Departments and computer capabilities that have been authorized.

And yes, Web Analysts are capable of writing codes. So are Computer Graphics, Online Illustrators and other creative/marketing personnel. They should, however, it is not always part of the job. Size matters. How big is the company? What is the focus? How many people in the Traffic Department? Who does what? Which person covers which person?

What is the difference between web analytics and advanced web analytics?
Advanced Web Analytics comes with time. It is a senior position for those that are capable of handling the highly sophicated data. They'd be able to do collaborative filtering, audience targeting, and work in markets (horizontal and vertical), which means a business administration, advertising, marketing communications or organizational development/leadership/management background would be a necessity.

I hope that this helped you gain a deeper understanding of Web Mining vs. Web Analytics. And I am still on the career search, if you hear of any opportunities feel free to forward my resume.
Web mining is to web analytics what data mining is to data analysis. Just my 2 cents.
When I started learning "data mining" about ten years ago I remember thinking "hey, this is just what I learned as Exploratory Data Analysis in grad school, with a new fancy buzzword name". Exploratory = finding patterns. Finding = no a priori expectations. So I would disagree with some of what's said here.
But there's more than exploratory analysis, (in my opinion) it's also about real time processing of very large datasets using parametric, statistical or data-driven models trained on billions or trillions of historical transactions: web traffic scoring and ad routing, credit card fraud detection, automated high frequency stock trading, weather forecasting etc.

Some of the not so obvious steps are feature selection, building lookup tables with high predictive power that can fit in memory, and how to efficiency build a distributed implementation. Also identifying the right data sets (internal, external, vendors) and contructing predictive compound metrics (a combinatorial optimization problem).
I agree that "data mining" connotes very large scale, although I'm not sure when something becomes big enough to cross the line into "data mining" from "exploratory analysis." And I'm not sure that data mining necessarily means "real time" although the models that data mining produces are themselves often used in real time.

But my comment originally was just in response to the earlier definition that used the phrase "an a priori expectation of what you will find." I was only trying to make sure that hypothesis testing wasn't being implied, because data mining (in my mind) is quite different.

I could be all wrong. It's been a while since I studied data mining, if "studied" is the word.

Either way, it's all just labels and I think the discussions of individual methods are more interesting. For me, there's "reporting" which is simple descriptions and trending, and then there's "analysis" which actually attempts to be useful for business decisions, and that's kinda the end of my own little semantic road.
Web mining is ultimately no different than data mining. Though they employ different sources and data prep techniques, once you're beyond that step, there's little difference between the two (and I only said "little" as a caveat).

To me, web analytics are also analagous to more traditional exploratories and tracking systems ... a very infant but emerging field.
Check out the practitioner's web analytics blog:

http://webanalyticsnuggets.blogspot.com


Read about different ODG vehicles:

http://webanalyticsnuggets.blogspot.com/2010/06/comparison-of-odg-v...


and about Path Based Segmentation as an input to test design:

http://webanalyticsnuggets.blogspot.com/search/label/test_idea_fram...

RSS

Follow us

© 2013   AnalyticBridge.com is a subsidiary and dedicated channel of Data Science Central LLC

Badges  |  Report an Issue  |  Terms of Service