Five Predictions For Web Analytics in 2011

Web Analytics predictions by Brian Clifton

Predicting the future invariably means you will be wrong most of the time. However,  it is an interesting process to go through as even getting just one prediction right can have a significant impact - to me personally, my business or my client's business. So I was honoured when Daniel Waisberg asked me to look into my crystal ball for what may happen in the world of web analytics in 2011.

Luckily I am on safe ground - I have been involved in web measurement since the very early days (1997 onwards), so I figure most of my predictions will prove to be valid and Daniel will invite me back this time next year to congratulate my foresight :P

1. Privacy - A Threat To The Web Analytics Industry

Web Privacy and Security
Privacy is one of the greatest challenges the industry faces at present and one that has been steadily growing over the past couple of years. My prediction is that this will come to the fore in 2011, with privacy on the web being discussed beyond specialists groups and into mainstream media.

What are the privacy issues?

There are many - from data collection methods using 1st party cookies, third party cookies, Flash objects and data triangulation, to the highly personal IMEI number of your mobile phone. The issue is that the consumer (every web user), struggles to know what information is private, what is public and what to do in order to change the info they wish to share or keep private. Today the "average" web user is simply not in control of their own personal information.

A consequence of this, is that a particularly bad experience can lead to a complete shut down of sharing visitor information. This can be either at the individual level (e.g. mass closure of Facebook accounts) or at the government level (e.g. Germany recently stating that GA was illegal to use on German websites).

This is a real problem for the web measurement industry. If web users decide not to share their anonymous information in large numbers, tracking data becomes sparse, and the web takes a backwards step. That is, websites cannot be improved if there is no data to go by...!

What can we do to mitigate such a privacy challenge?

It is very important that vendors, consultants and industry associations firstly educate, but also go out of their way, to ensure that visitor privacy always takes precedent over the "feature requests" of advertisers, marketers and web site owners. Not an easy task when these are the people paying the vendor/consultant's fees...

2. Optimisation Rather Than Reporting

Business Optimization
I am a big fan of Avinash Kaushik's work and we have been friends since I first met him in 2005 - he even wrote the foreword for my Google Analytics book - and his presentations always get me thinking. This year he spoke at the Google Analytics Summit at Google, Mountain View, on the topic:

"We [web analytics professionals] don't know how to sell web analytics"

In a nutshell, he presented on the failure of selling web analytics as a reporting tool, rather than an optimisation tool, which of course is where the value lies. My prediction for 2011 is that this will finally change...

All consultants I speak with would much rather sell web/business optimisation services than implementation. The difficultly is how to do this. For example, organisations contact my company to get GA setup. Once done and they have good, solid, reliable data coming in, they sign the cheque and say thank you and good bye. They either don't see any value of insights/analysis, or they feel they can do this themselves, which invariably they don't get around to (usually due to a lack of expertise/experience). Coming back to these organisations a year later, I find they know no more about their online business than before we started - they just have more data!

How do we remove the blockage?

Optimising a business via the online channel, or any channel for that matter, requires the buy-in of the CEO/CMO - that is, people who can align different teams and gain a consensus of opinion within an organisation to drive web analytics in the right direction. That is, using it to improve the visitor experience. Unfortunately, to date web analytics has not sat at the top table, so it has not been discussed at that level - its an uphill battle for those that make their living helping organisations get better at what they do online.

Of course, people like myself and Avinash are constantly speaking and writing about this - targeting the CxO's. So my prediction for 2011 is that web analytics will finally graduate to the CxO level, at least for the most proactive organisations.

3. Predictive Analytics Will Remain Niche

Predictive Analytics Market
Predictive Analytics, in theory, has fantastic potential. Currently, web analytics is about looking backwards into the past. For example, "what happened with last month's campaign?", or "how did the product launch perform last week?". Being able to look forward and predict the effect of a new launch/campaign is very exciting.

However, from my scientific background, trying to predict the future doesn't sit well with me. Its fraught with uncertainty and caveats that often get overlooked and forgotten over time. Most people, even smart analysts, have difficulty in understanding probabilities when more than one event effects an outcome. Predictive Analytics has to contend with many multiple, interacting events. The result is that future hypothesis can easily get built on sand.

So while predictive analytics is of interest to me, I can't see it being used in anything more than very specific and controlled situations where rigorous statistics are applied and understood. That is, its not going to be an integral part of marketing anywhere in the near future. That's a dichotomy, as in my view marketing is where web analytics belongs.

4. Accessing The Web Will Be Different - So Will Its Measurement

Multichannel Marketing Optimization
Will mobiles overtake laptops as the main device for accessing content? And if so will apps overtake browsers in countries like India & China, where the growth of Internet available phones is rapidly taking over computers?

For 2011, I think we will reach an equilibrium where all devices become a part of the user experience. For example, using a browser on a large screen when at home, following up via mobile while commuting to work, using an app for repeat purchases when the relationship has solidified i.e. when a strong confidence and trust has been nurtured (also see article by Vicky Brock on Mobile Optimization for e-commerce).

Small, portable screens are very convenient on the go and when you know what you are looking for. However, for information and product research you can't beat a large screen where you can multi-task with multiple windows. So I can't see us only using miniature handheld mobile devices for all our web needs - it will be a mix.

Designing web sites that work well on a mix of devices is a headache for developers. Yet smart devices such as the iPhone, iPad and Android systems are still capable of executing the tracking code deployed on sites - such as Google Analytics, WebTrends, Omniture etc. So data collection is not an issue!

The challenge however, is how to stitch together the behaviour of the same visitor who uses multiple devices during their day (home PC, work laptop, smart phone and iPad). At present this is not possible - unless the visitor is forced to log in to use your website - and that's not an attractive proposition to users. So the current situation leads to an over counting of unique visitors. That problem is becoming increasingly significant and the following presentation shows:

5. WebTrends Will No Longer Exist

Facebook Analytics
When I first embarked on my web analytics career back in 1997, WebTrends *was* web analytics. It was the product I first used (alongside Stephen Turner's Analog) to look at what was happening on client websites. Unfortunately since those heady days of dominating the market place, WebTrends has lost its way.

In 2001, WebTrends became a part of enterprise IT services from NetIQ, just when web analytics was starting its move away from selling metrics tools to IT departments and towards marketing departments instead. Haemorrhaging customers, it thankfully managed to uncouple itself from that acquisition and became independent again in 2005. But the venture capitalists that bought it didn't understand the rapidly changing market place (Google Analytics launched in 2005). And so four CEOs later and WebTrends still struggles to hold on to its clients.

As I say, I cut my web analytics teeth on WebTrends Log Analyser, so I have a sentimental attraction to them. And their latest version is doing some pretty cool stuff with tracking Facebook engagement for large brands. However, the fact that is has been left on the shelf in the latest flurry of recent acquisitions doesn't bode well.

To survive against the likes of IBM (Unica & Coremetrics), Adobe (Onmiture), Google, and comScore (Nedstat), WebTrends needs to partner, or be acquired, by a major corporation. So for 2011 I predict their recent work with Facebook will stand them in good stead and  they will emerge as Facebook Analytics, or FaceBook Trends or similar...

Do you have any feedback for Brian...? Please leave your comments and questions, even your own predictions below and join the discussion.

Subscribe To Our Newsletter For Monthly Updates



Your e-mail will be kept private

Matthias | November 2010

Nice predictions, regardless you do not always believe in prediction analytics ;-) I actually like the time of the year with all the forecasts.

Very interesting thoughts about WebTrends. I would give it another year until it may change, the average customers are probably not ready for highly sophisticated services and still go on with "basic" analytics/reporting. You mentioned this behavior from your experience with a lot of customers. But things have quickly changed recently, so let's see.

Regarding privacy, I feel that this topic triggers concerns by nature as long as optimization (at least for commercial/sales websites) are explained as a service for the user.
This is possible, but the intention is clearly optimizing for increasing profit, savings, efficiency or reducing costs, etc. The user's comfort is more a "side effect" than the real goal.
But who counts the millions who are daily annoyed by (or worried about) e.g. targeting? And most of those are still not giving up to buy online ;-) What a patient and friendly customer segment!

I believe the major point is about missing transparency. If a website (or an ad-network, etc) would clearly tell what data they are collecting, and for what purpose, and who will have access to it, and for how long, and especially: How can one opt out (or at least adjust the ads/targeting etc to what really matters someone), and giving an complete answer if being asked what is actually stored about oneself (if applicable)?

That would certainly answer a lot of questions AND eliminate rumors (or isolate the not-so-good-practices which then could optimze..).
But as long these questions are not answered, or only on a superficial level ("we use cookies to optimize your user experience" and similar), it would be naive to hope that a user (literally) buys in to the given state of information.

By my experience the industry often sees privacy issues as kind of a senseless legal barrier from paranoid countries which actually hinders a customer to become happy. OK, I am getting sarcastic, sorry ;-)
What I mean is that taking only the position of the e-commerce does not serve the WA industry for long, because it simply ignores increasing doubts and fears by customers.
Also it ignores the fact that having access to data is power. Especially if being able to put different pieces together, which are by each not a matter of concern. This is a potential danger, if being abused. If not today, then maybe tomorrow? Maybe by someone who has a very different intention that the one who was collecting the data?
Saying that, the focus of the privacy discussion only on e-commerce or website optimization isn't going far enough. IMHO it's nothing less than the discussion about the ability of every person to control its personalized data.

So I fully agree with your prediction that privacy is a big, big topic for 2011. Finally!

Michael D. Healy | November 2010

Brian:

Excellent predictions, although privacy is still the largest.

Differential laws, regional expectations and technology have complicated what should be a fairly straightforward system:

Explain - What You Are Collecting, Clearly AND Comprehensively
Offer - The Chance To Opt Out
Report - Those Who Are Not Playing Nice

We, as the practitioners, should be warning the public about things like flash cookies; building credibility with the public. The fact that several students in a summer intern program conducted the research that received all the attention recently makes us look negligent, or worse.

If we explain to the public that tracking online usage is no different that trying to connect with customers entering a physical location they should get it.

The mistrust is driven when things are found out after the fact, obscured behind legalese or technological developments. If you sign up for the WAA code of ethics and see someone breaking it, would you call them out?

With regards to legality, much is written about European laws however few people seem to be as concerned about California law which, according to some legal experts, is directly on point.

Any website that is used by a resident of California, regardless of the location of business operating the site, may be breaking California law.

Read more here:

http://michaeldhealy.com/2010/11/does-your-website-break-california-law/

Michael D. Healy

Raghu Kashyap | November 2010

Interesting take on 2011. I believe that more and more people get analytics savvy they will get data hungry. The core site analytics provides information for the normal folks. I feel that products like QlikView and Omniture Insights have a nice area and focus which will be harvested by data savvy business folks.

As far webtrends I totally agree with you :-). Companies such as ours are heavily focused on predictive analytics and believe that the future it holds is very promising

Nigel Bradley, University of Westminster | November 2010

Brilliant ideas

Again I see privacy as the biggest... so the WAA code of Conduct will be fascinating as it develops

Matt Gershoff | November 2010

Thanks for putting your predictions up, very interesting. One thing that has me thinking is your joint take on optimization and predictive analytics. ‘Yes’ to optimization, ‘not so much’ for predictive analytics. Perhaps it is just the term ‘Predictive Analytics’ that you find disagreeable because I am not too sure how one can optimize anything without out having some predicted estimate of the impact of some potential action.
I normally think of optimization as a prediction/control problem and web analytics reporting as a form of adhoc prediction/estimation about both how users are distributed and how these users might respond to some experience. All of the talk about segmentation by Avinaish et al. is really just pointing out that the user distribution is not really uniform or drawn from some single Gaussian (or pick your favorite distribution) but is better modeled as some form of mixture of multivariate distributions.

This type of data partitioning (segmentation) is of interest because we have expectations that users sitting in different regions will respond differently, in the average, to various treatments/exposures/experience ect. - so there is some implicit notion of conditional probabilities/value here. We also have expectations that the segments we see today provide some prior on our expectations about our users tomorrow, in other words the environment, at least in the short term, is in some sense stationary -at least within a given time frame.
From that perspective, I don't see why you think predictive analytics per se has no large place in the future, especially if we think optimization is finally going to be moving to the top of the agenda. How can one take actions that will optimize without some sort of model of the environment in which those actions are taking place?
Now if you are really just saying, 'Hey, just throwing some regression modeling (support vector machine, whatever) at the problem, in of itself isn't going to be too useful", then I agree. However, either the data that is collected from web analytics is useful or it isn't and by useful I mean helps use make decisions. If it is, then I could easily see that embedding some intelligent technologies into applications that take advantage of that data would lead to improvements. If not, then why bother with web analytics reporting in the first place. If past experience is no guide for the future then there can be no learning and certainly no optimization. The value of knowing how we did is that it to some degree informs us of how we will do in the future.

I would think the main issue is not whether or not more formal methods are or are not used for prediction, but it is that prediction and control are often decoupled, so you have web analytics folks trying to get all sorts of learning (predictions) but with no mechanism to link them to actions.
Cheers and thanks again for the post.

Brian Clifton | November 2010

A great discussion happening here :) Keep them coming...

Responding to comments...

Matthias
I agree transparency is important but don't think is the issue here. For me, it is the lack of control and complexity that web users experience when trying to control their privacy settings that is the problem. Privacy needs to be centralised - either on the browser, or through a trusted 3rd-party, where the user manages everything in one place. Similar to how SSL certificates work...

Michael
ditto my comments above. Having simplified, clear language privacy statements is obviously a great step forward. However very few people (including myself!) ever have time to read these so I don't feel that should be the focus for improving privacy. I agree that education has a big role to play here.

Nigel
I think the WAA Code of Conduct is an excellent approach. However, though the WAA is a voluntary, dedicated and hard working group, they simply cannot be as proactive on such issues as I wish they could. For a CoC to mean anything, it needs to be policed, which logistically it cannot. Unfortunately, that makes it redundant IMO.

Matt
I thought my comments on predictive analytics might cause a stir ;)

Essentially, I view predictive analytics like I view the weather forecast - there are some insights to be had, but they are very short term and always come with a probability factor i.e. likelihood of being wrong.

If you are building a business on predicting the weather, then you need a strong understanding of the methods employed and caveats that come with it. I just don't see that as being a part of marketing in the foreseeable future (many, many years away) - which is where web analytics needs to be...

Matthias | November 2010

Hi Brian,

seeing transparency as a duty of the browser, to be configured by the user, seems not to solve the issue that there are growing concerns about webanalyics methods.
As a customer I may agree in a company storing (some of) my data. But how can I configure my browser if I only can activate/deactivate tracking in general and per domain? I'd like to define this in more detail depending on the purpose of the collection. I still see a strong need that the industry shows more transparency on their intended data usage. Nevertheless the browsers should provide tools which can be configured a bit more specific as only on/off. Seems not so easy.

Apps in Facebook are a good, bad example: I have to agree "access to my profile" when installing an app, but I have no idea which of my profile data is actually read out and what for (and stored for how long, etc). I'd like to have more transparency here by the app telling me exactly what data it wants to have and why.
Whats the problem in simply telling whatfor the user data is used (and what data is not collected or will not be used)? I believe this would increase the trust of consumers significantly. On the other hand not serving this need may cause much more problems (and derived from that worse regulations) as currently existing.

btw: Nice discussion indeed!

Matt Gershoff | November 2010

Brian,
Thanks for the response, and since you are asking for more I will gladly oblige ;).
Maybe some of the confusion is due to the under defining of both 'optimization' and ‘predictive analytics' within the context of digital marketing? So how do you (Brian) define both, and perhaps give some examples?
To make it easier I will give you my take.

Optimization –> How to make the best decisions
When I think of applied optimization more abstractly, I visualize some sort of controller that is tasked with making decisions in a given environment such that it maximizes the value of a specified objective(s). The controller is what takes action (makes decisions) in the environment, and monitors how the environment has changed based on its action. Over time the optimizing controller will be able to estimate which actions lead to the best outcomes (highest value) if it is able to learn how the environment responds to each of its actions.

Predictive Analytics -> How to estimate/learn how the environment behaves
Generally, I would think that predictive analytics refers to the application of formal methods to currently available data in order to make inferences about a domain, environment or process. I guess people generally mean using statistical inference when they talk about predictive analytics, but honestly, I am not sure.
Also, marketers have been using statistical models for several decades. Database marketing does tons of work with clustering (e.g. K-means), value scoring (e.g. regression), and classification/targeting (e.g. logistic regression, tree classifiers, etc.). I don’t think it would be too controversial to say that digital (web) marketing can be seen as an extension of database marketing and should be able to inherit many of DM’s methods.
Of course, you can’t get water from a stone. So if you have limited data availability, your ability to make inferences based on that data will be limited regardless of your approach. None of this stuff should be thought of as magic (including web analytics), but rather as just different methods that are more or less efficient at allowing you to extract useful information from available data.
Cheers - Matt

Neil Mason | November 2010

Hi Brian

Hope all is well. I think that your predictions are pretty much on track, though I would probably be happy to discuss your thoughts on predictive analytics! For me, predictive analytics is not so much around trying to accurately predict outcomes in the future but to reduce risk and uncertainty in decision making. I think that given the inherent nature of digital data with all its noise and confusion that sophisticated analytical techniques will be needed more in the future to extract the signals from all the noise. Data mining and predictive analytics are part of the toolset and, for example, it will be interesting to see whether someone like IBM, with its recent acquisitions of SPSS and Unica, will start to create momentum in this area.

Anyway, if anyone’s interested in some more of my views around the utility of predictive analytical techniques in the digital space, then hop over to the eMetrics videos part of this site for a video of a recent presentation given in Washington DC.

Regards

Neil

Brian Clifton | December 2010

Matt/Neil: To clarify, I am a big fan of predictive analytics - beta trialling the Google Analytics Intelligence reports in 2009 was what really whetted my appetite.

What I am saying/predicting in this post is that it will remain a niche area of digital marketing for some time to some - certainly next year and me thinks a lot longer...

At present I would describe the vast majority of web analytics implementations I come across as basic - and thats from large brands with intelligent web analytics teams behind them. But at least they are using the data in some way. Predictive analytics won't break out of niche until that bar rises...

Matt Gershoff | December 2010

Sure, I agree that organizations that have digital applications with scale/value will be the early adopters to more sophisticated methods. I would argue that PA, in a way, is really just web analytics coupled formal methods. I know this is a bit off topic but I think we will see the turn when rather than focusing on reporting/prediction, the industry refocuses on decisions/control. I find the following to be a nice framework to think about, as well as implement, optimization (or at least continual improvement).
http://goo.gl/BSxth

Cheers and thanks for the responses.

Amitha Singh | December 2010

Very interesting predictions. What specifically gave me goosebumps was the one about WebTrends... Coincidentally I began using WebTrends in 1996/97 as well! There was a tool called HitList (a whole lot lighter and faster than WebTrends) which generated very interesting marketing-friendly reports. If I am not mistaken that one got bought over by Microsoft and included in their IIS server suite. That's the last I've heard of it :(
Unfortunately for web analytics junkies, beside GA, there aren't too many good accessible tools available to continue drilling down on data and play around with the results. Will definitely be back same time next year, Brian to check on how your predictions turned out! I'd lay my bet on 2 out of the 5 (WebTrends Will No Longer Exist & Accessing the Web Will Be Different).

Regards,
Amitha

Post new comment
The content of this field is kept private and will not be shown publicly.
Refresh Type the characters you see in this picture.
Type the characters you see in the picture; if you can't read them, submit the form and a new image will be generated. Not case sensitive.  Switch to audio verification.
Online Behavior © 2012