As Chief Marketing Officers develop their social media marketing strategy for 2010, they are demanding business results. In 2009, 89% of CMOs tracked social media’s impact by using standard metrics such as site traffic, pageviews, and number of fans (as discussed in a recent survey).  However, CMOs expect that in 2010 top metrics will track more closely to P&L business goals––not just Web-related goals. The study forecasts the growth of adoption of the top three metrics in 2010, as follows:

• A 333% increase in tracking revenue
• A 174% increase in tracking conversion
• A 150% increase in tracking average order value

Such a shift in measurement expectation is significant. CMOs indicate a 300% year-over-year increase in 2010 in the number of companies that plan to measure social media’s impact on conversion and a 400% increase in the number of companies that will track social media’s direct impact on revenues.

The 2009 financial crisis probably did social media some good. Not that you need a ton of budget, but competing for scarce budget alongside more traditional projects is a creative constraint. As social media marketing matures, it will become a facet of marketing overall and it will be harder to spot a campaign that isn’t social. Part of that maturing is - you can’t manage what you can’t measure. Part of it driving activity that drives traditional metrics, like how you can drive conversions with LeadShare.

Social media marketers should get ahead of this curve. If you know you will eventually be accountable for traditional metrics, start iterating as soon as you can to find models that work. And volunteer to report these metrics before they are volunteered to you. This will require that you actively engage other parts of the marketing organization and give them stakeholdership in your outcomes. Take a look at your 2010 campaigns, reconsider your metrics, and incrementally realign your activities with the core of the marketing function.

My thoughts:

1. Social Media is bringing marketing and service people new opportunities of understanding their customers and also tracking/trapping events and customer needs. This is in addition to formerly understanding customers and their behaviors through the use of advanced data warehousing and Business Intelligence investments aimed at processes and techniques that (demand and) drive smart marketers actions. Teradata has been providing such solutions for over 20 years. But now the new era of faster communications, WEB 2.0 applications, and customer-driven interactions has catapulted many countries and companies into a totally new era.
2. Original Web Analytics are normally collected from logs or applications on  websites which focus on their sales activities and orders (and even revenues). But most companies have invested millions or ten of millions of dollars (or your local currency equivalent) and rarely are tracking and trapping the customers entering point, movements on your website, pages read (and how long), products reviewed and subsequent searches/views which define what customers seeks and would desire. If you had a sales-person on their shoulder watching them go through your website, wouldn’t he first be asking questions and second finding ways to fulfill the customer (purchasing or service) needs?
3. Investing in web analytics has now moved beyond the most obvious metrics (usually self-aimed at the business, not about the customers). The new age of tracking and understanding customers has enabled the opportunity of compiling data into a Teradata Data Warehouse and then using sophisticated analytical techniques and models to do what is necessary to take immediate actions and complete the sales or service cycle. In addition, Teradata has partnered with numerous companies, serving many industries, which extrapolate and move the web activity data to the Teradata Data Warehouse, then other companies provide applications and analytics that give INSIGHT to managers and executives who need information to manage their resources. This new area is known as “Interactive Web Intelligence” or “Integrated Web Intelligence” (IWI). This means integrating web data with your detailed customer data from all of your other channels. This is now (sort of) mandatory; if you plan to be successful in the electronic (PDA) world.
4. There are some excellent Teradata software alliance partners that provide modern-enabling tools for such gathering and analysis. They are WebTrends and SpeedTrap, along with others who provide additional infrastructure support and loading of data into the Teradata DW’s. 

SUMMARY AND RECOMMENDATION
Reporting at the end of the month, or even the end of the week, is no longer sensible or even useful. Latent “Post-Action” tracking and reporting, with delays in analyzing and then acting, provide little economic value.  In today’s world, using the enabling technologies and smart people to go with them, you should be seeking an ACTIVE Data Warehouse with ACTIVE Enterprise Intelligence. Your competitors are in the integration mode and now gathering web data, and then moving quickly to learn and use such data to manage customer retention, customer sales, customer services and customer satisfaction. Are you?

My best recommendation is to consider how much you plan, or have, invested in your customer marketing and/or web site. Then evaluate what it would mean if you took just ten percent (10%) and reallocated it to Integrated Web Intelligence (IWI). No one, including your customers and competitors, will find your reallocation to be less than magnificent in terms of ROI. BI and DW along with IWI are part of the new world of Web 2.0 and subsequently understanding your customers and prospects. Address them with your best RELEVANT messaging and you will win in the world of intelligence and revenues. What do you need to know? Let me know…

Ron Swift

www.teradata.com/ronswift

The report is a serious one because CIO’s everywhere will read it, see the quadrants and how they are populated – then make more informed decisions. Those decisions shape our future business. I read the report’s fine print, too.

Like always, the report focuses on the relevant capabilities of vendors. However, the analysts remarked this year that Teradata’s customers play a role in our product development. I’m glad they did, because it’s an important factor. I commented on it in the press release because I believe that much of the credit for our success as a business comes from our enthusiastic and engaged customers.

We all know that data warehousing is a sophisticated computer science. However, the excitement, enthusiasm and ongoing involvement of our customers (a big differentiator for Teradata) resemble a team sport. Our customers, our employees and our partners talk about data warehousing – the way a lot of us talk about sports.

These high-energy conversations take place at our conferences, in our meeting rooms and in our online forums. And when we hear things like “Teradata is a central player in a high-stakes IT arena,” I think of the Super Bowl. I can’t help it.

Of course, data warehousing is hard science, but for the Teradata community, using enterprise-class intelligence to do dramatic things for a business is serious fun. We talk about big plays, close calls, the best players and every dimension of the game.

And when the Gartner data warehouse reports come out every year, I am glad to see they are keeping score ... and reporting the team standings. I’m proud that Teradata has been in the Leaders Quadrant since 1999, a position we all work together to keep.

Darryl

Following-up to my blog 'Just Tell Me What I'm Doing', I'm starting a series of posts that define the key concepts and terms that make up my analytic world. Everything I do is coloured by my experience actually doing analytics in commercial organisations. So while I believe these posts will present practical definitions that will be actionable in the business world, I know that there are other worlds in academia and science where they are less relevant. At the very least, people in these areas will gain a better understanding of how business regards analytics.

Bennett's Analytica  
   A Practitioner's Guide To Analytics

Data, Information and Knowledge

Information is a collection of related data – often transformed and aggregated – about a topic. In business, that topic is often insight about an operational area or a performance question. In analytics, information is often used interchangeably to mean ‘data’ but data is actually best thought of as something that on its own carries no meaning. The main differences are in the degree of meaning and the level of abstraction being considered. To explain:

Degree Of Meaning

Data, information and knowledge all have some degree of meaning. Even data has meaning at some level. For example:

• data: 99.9 is a number (you know it is probably not text). There is still a possibility that 99.9 is code for a text string or value. 
• information: 99.9 is the percent of transactions successfully processed by an application.
• knowledge: 99.9 is 0.05 below the acceptable level for failed transactions with our customers.

Level Of Abstraction

Data is the lowest level of abstraction, information is the next level, and finally, knowledge is the highest level among all three.

Be careful: abstraction is not the same as summarisation. Summaries may only be the sum of individual pieces of data. This doesn't change the data into information in and of itself. An example:

A list of amounts 5, 8, 5, 2 can be summed to 20. Is 20 information?

Sources

In the business intelligence world data is extracted from fixed sources (batch or in real time, it doesn't matter). Sources are usually either transactional applications or reference data. All sources have meaning. Transactional data has meaning because:

• each transaction is stored in one or more records and this gives context to the individual data items of the record.
• the source application is known and that is information that gives additional meaning to the data.

Reference data also has meaning as the table(s) within which it is stored has an internal meaning due to the relationship between the table rows. Typically this meaning is either hierarchical (for example an organisational structure or products grouped into categories) or group (for example a list of product codes or currencies).

In order for data to become information, it must be interpreted and take on a meaning.

Analytica Illustration

An example (care of Wikipedia):

"The height of Mt. Everest is generally considered as "data", a book on Mt. Everest geological characteristics may be considered as "information", and a report containing practical information on the best way to reach Mt. Everest's peak may be considered as "knowledge"."

Related Terms and Concepts

Refer also to Data

Refer also to Metadata

Link to original post]]>Steve BennettIndustry PerspectivePredictive Analyticsanalytics,metadata,abstraction,fixed sourceshttp://smartdatacollective.com/Home/24822#0Sun, 07 Feb 2010 19:43:53 GMThttp://smartdatacollective.com/Home/24822http://smartdatacollective.com/Home/24822http://feedproxy.google.com/~r/smartdatacollective_allposts/~3/pcHsbL2ZdIc/24818With the continued growth of both the pubic social web and private collaboration, communication and social business tools we are creating an explosion of social data. As businesses get more deeply ...<img src="http://feeds.feedburner.com/~r/smartdatacollective_allposts/~4/pcHsbL2ZdIc" height="1" width="1"/>

With the continued growth of both the pubic social web and private collaboration, communication and social business tools we are creating an explosion of social data. As businesses get more deeply involved in the social business movement and as software vendors create more and different social tools there is a compelling case for tools to help businesses make sense out of all this social data. I wrote about the idea of social analytics last year. This year in our IDC Top 10 Predictions we included a prediction about social analytics: "Business applications will undergo a fundamental transformation — fusing business applications with social/collaboration software and analytics into a new generation of "socialytic" apps, challenging current market leaders." Here's a simple model that shows the concept of socilaytic platforms / apps and how they might be applied:

Some characteristics of socialytic apps might include:

• Aggregate social data from public and private social data sources (user / company selectable)
• Real-time search and monitoring
• Social metrics dashboard
• Natural language processing (NLP) with linguistic analysis capabilities
• Visualization and simulation
• Social trending
• Human search
• User configurability

I'm sure there are lot's of other characteristics and use cases for socialytic platforms and apps that will emerge as more businesses start deploying social tools over the next few years. Once the analysis capabilities are in place, businesses will undoubtedly start to look at automating decisions and building / deploying automated decision architecture-based solutions to help social analysts focus on higher value interactions. A few vendors are offering socialytic apps but the market is far from crowded at this point. An upswing in demand as companies get deeper into their social business projects should drive market growth and the development of new and varied offerings to address emerging needs.

Several month's ago a short video appeared on YouTube with an interview of LinkedIn's Chief Scientist DJ Patil. In it he discusses how 'Big Data' impacts the practise of analytics. I've only just got around to posting about it but I am doing so now because he has some insights that I agree with and would like to share as they are still relevant. 

Big data is today most often associated with the internet superstars like Google, eBay and Amazon. There are 3 other areas with lower profiles where big data is important: intelligence (spooks, the military, etc.), scientific and academic research, and the financial markets.

Big data's future is much bigger than this because more and more areas of human activity are going to be faced with vast data sets. When you hear people talking about the growth of knowledge and statements like 'if this data were printed then the stack would grow faster than NASA’s fastest rocket', you have to remember that there is a good chance that each page of new data is adding to someone's analytic data set.

I'm not quoting the guy verbatim but here's what I heard and my takeouts to his comments:

• Open source 'big data ready' technologies like Hadoop (see my earlier blog or here) have come into their own now. Look to people with these skills over those only with SQL if you are facing big data challenges.
• We have reached a tipping point in the use of open source for commercial solutions to big data problems.
• If you want good analysts then the best place is to look is in occupations where people will already have the practical skills in manipulating big data sets: scientific fields like meteorology, oceanography and the like. I agree but this is not the only place as in my experience I also need analysts that relate well to business decision makers - i.e. those people that make commercial decisions based on the analytics. This is perhaps less important in pure tech plays like LinkedIn.
• Open source will transform the practise of analytics in the next 3 - 5 years. I think it will take longer than this to really impact the more traditional industries. I'm not happy about this but I am realistic about the difficulty in convincing business leaders that open source is a superior solution to proprietary ones. The money behind the big vendors will keep them going for a number of years yet.

One potential qualifier to DJ Patil's perspective is that although he has a very impressive big data background as a mathematician, US Department of Defence analyst ('Threat Anticipation'), and former eBay Director of Strategy and Analytics, his current employer is LinkedIn.

The core of LinkedIn's big data is structured and fairly static: profiles of people. So I'm not sure how similar their big data challenges would be to, say, those faced with processing, understanding and predicting large streams of real time data from financial markets or very large sensor arrays. On the other hand, the growth of LinkedIn communities and their related activities must generate large amounts of semi-structured data.

I also have no idea what LinkedIn's own analytic goals are beyond what DJ mentions on his own profile where he says his analytics drives product features like:

• "People You May Know"
• "Who Viewed My Profile"
• "Groups You Might Like"

Maybe somebody reading this blog knows more?

The video is on YouTube and I embed it here for convenience:

Or you can download 'DJ Patil on How Big Data Impacts Analytics' directly from this blog.

I am by no means an expert in enterprise sales. Personally, I vastly prefer marketing (one-to-many) versus sales (one-to-one), hence only start companies making consumer or small business products (advertising based or sub-$5000 price tags). But I have been involved in a few enterprise companies over the years. Here’s the main thing I’ve observed. Almost every enterprise startup I’ve seen has a product that would solve a problem their prospective customers have. But that isn’t the key question. The key question is whether it solves a problem that is one of the prospective customer’s top immediate priorities. Getting an enterprise to cough up $100K+ requires the “buy in” of many people, most of whom would prefer to maintain the status quo. Only if your product is a top priority can you get powerful “champions” to cut through the red tape.

My rule of thumb is that every enterprise (or large business unit within an enterprise) will, at best, buy 1-3 new enterprise products per year.  You can have the greatest hardware/software in the world, but if you aren’t one of their top three priorities, you won’t be able to profitably sell to them.

One final note: enterprise-focused VC’s sometimes refer to products priced between (roughly) $5k and $100K as falling in the “valley of death.” Above $100K, you might be able to make a profit given the cost of sales. Below $5k you might be able to market your product, hence have a very low cost of sales. In between, you need to do sales but it’s hard to do it profitably. Your best bet is a “channel” strategy; however, for innovative new products that is often a lot like trying to push a string.

The day started with a keynote from Harvard economist Susan Athey. Her research focuses on the design of auction-based markets, a topic core to the business of search which largely relies on auction-based advertising models (cf. Google AdWords). Then came a session focused on learning and optimization. One paper proposed a method to learn ranking functions and query categorization simultaneously, reflecting that different categories of queries leads users to have different expectations about ranking. Another combined traditional list-based ranking with pair-wise comparisons between results to separate the results into tiers reflecting grades of relevance. An intriguing approach to query recommendation treated it as an optimization problem, perturbing users’ query-reformulation path to maximize the expected value of a utility function over the search session. Another paper looked not at ranking per se, but rather at improving the quality of training data for using machine learning for ranking. The final paper of the session, which earned a best-paper nomination, modeled document relevance based not on click-through behavior, but rather on post-click user behavior.

The next session was about users and measurement. It opened with another best-paper nominee: a analysis of over a hundred million users to understand how they re-find web content. Another offered a rigorous analysis of the often sloppily presented “long-tail” hypothesis: it found that light users disproportionately prefer content at the head of distribution while heavy users disproportionately prefer the tail. Another log-analysis paper analyzed search logs using a partially observable Markov model, a variant of thehidden Markov model in which not all of the hidden state transitions emit observable events–and compared the latent variables with eye-tracking studies. An intriguing study demonstrated that user behavior models are more predictive of goal success than models based on document relevance. The final paper of the session proposed methods for quantifying the reusability of the test collections that lie at the heart of information retrieval evaluation.

The last session of the day focused on social aspects of search. Two of the papers were concerned with modeling authority and influence in social networks, a problem in which I take a deep personal interest. Another inferred attributes of social network users based on those of other users in their communities (cg. MIT’s Project Gaydar). Another analyzed Flickr and Last.fm user logs to show that users’ semantic similarity based on their tagging behavior is predictive of social links. The final paper tackled the sparsity of social media tags by inferring latent topics from shared tags and spatial information.

Not surprisingly, a disproportionate number of contributors to the conference work at major web search companies, who have both the motivation to improve results and the access to data that is needed for such research. One of the ongoing research challenges for the field is to find ways to make this data available to others while respecting the business concerns of search engine companies and the privacy concerns of their users.

The extension command mechanism provides a way for users of traditional SPSS syntax to run Python programs written by someone else without needing any knowledge of Python.  But the program must have been written as an extension command.  While creating extension commands isn’t hard, it does require some extra knowledge and work.  (An article on how to do this is available on Developer Central.)

I have posted a new extension command, SPSSINC PROGRAM, that allows ordinary Python programs to be run with traditional syntax without the author having created an extension command: easy on the author and easy on the user.

Often someone writes and shares a Python program for use via BEGIN PROGRAM that requires some input parameters.  The BEGIN PROGRAM syntax does not allow for parameters, so the user must modify the program itself to specify these.  If you know Python, this is not a problem, but many users are uncomfortable doing that, since the Python language is quite different from the traditional SPSS language.

A Python programmer would typically define a function or class with the requisite parameters and just modify the function call.  An SPSS user might not know how to do that.

SPSSINC PROGRAM solves this problem.  (Of course, extension commands solve this problem more generally, but they take extra work to create.)   Suppose I have a program saved as mymodule.mypgm.py that does something to a pair of  SPSS variables , and I want the variable names to be  parameters.  Using SPSSINC PROGRAM, I would write

SPSSINC PROGRAM mymodule.mypgm firstvar secondvar.

firstvar and secondvar would be the parameter values passed to the program.  SPSSINC PROGRAM ensures that mymodule.mypgm is called,  makes the parameters available to the program, and handles various error conditions.

Instead of passing the parameters as function arguments, the parameters are set up as if they were a command line.  The author of the program would access these in the traditional Python way via sys.argv, with the first parameter being the module and program name.  It’s just as if the program were being run from a command shell, except that the parameter values have been passed through the SPSS Universal Parser.  Comments in the implementation module details the (small) differences this can produce in what the program sees.

So using SPSSINC PROGRAM is very easy on the Python programmer while still letting the user of the program work in the style he or she is comfortable with.  The package also includes a dialog box built with the Custom Dialog Builder where the user can enter the program name and any parameters.