Oz Analytics

Review of Dresner's Profiles in Performance

Howard Dresner, a fellow ex-Gartner Business Intelligence and Information Management Research Director (but with a far more distinguished research record than mine!) has published his second book at the end of last year. The book is called "Profiles in Performance: Business Intelligence Journeys and the Roadmap for Change". His first book "The Performance Management Revolution: Business Results Through Insight and Action" was published back in 2007.

This book is similar in intent and size as the recently reviewed "Analytics at Work: Smarter Decisions, Better Results", the new book by Thomas Davenport, Jeanne G. Harris (both of Competing On Analytics fame) and Robert Morison. This is not a bad thing as I like both of these books and while they share a common belief (crudely put: Analytics for performance management is a good thing) both make their own unique and valuable contributions.

In Howard's case, he has created a Performance Culture Maturity Model that resonates. This model seems to me to be a much better descriptor than the widely used Information Maturity Model originally created by the now defunct Meta Group.

The bulk of The Performance Management Revolution contains four case studies on:

• Denihan Hospitality Group
• Cleveland Clinic
• Northern California Public Broadcasting
• Meuller Inc.

If the book falls short anywhere, it is here. Maybe it is my background, but these are hardly houshold names. They are also all US based and I think that this could reduce the impact the book will have on readers. Why? In my case I didn't have an existing impression of any of these companies. If the examples were, for example, Coke or Unilever, then I would have had other knowledge to draw on as I read the book - and this would have enhanced the stories.

It is the PCMM that I will benefit most from. To show how this contributes to our thinking on enterprise performance management, remember the IMM from Meta (via the MIKE2.0 website):

The PCMM captures the multifaceted nature of information and performance maturity and the fact that my organisation can be at different levels of maturity at the same time. I reproduce Howard's model here as it is published elsewhere on Google and book sale sites:

So Howard's post-Hyperion life was been productive so far. Let's hope that he continues to develop the PCMM. I for one will happily use it in my professional life.

Here are the details, and of course you can order it online:

Profiles in Performance: Business Intelligence Journeys and the Roadmap for Change (Hardcover) by Howard Dresner
174 pages
Wiley
ISBN: 978-0470408865
US$32.85

A Data Quality Fairy Tail

I get it, and admire Platon for at least trying but the end result is still naff. Sorry.

And just in case Platon reads this: I still think this earlier attempt is a great video to get your point across that Information Management is vital:

Kudos to Platon. Here it is to download as an mp4 file.

Who is Platon? Essentially a mid-sized IM consulting firm formed and run by a bunch of Copenhagen Business School graduates. Based in Copenhagen but with offices in the US, etc. with a small operation in Australia. They have some good content on their website - with a dozen of so small case studies you can read here.

Mining Social Media

 The finest example of social media mining that I have found to-date is the superb site Please Robe Me.

In case you can't guess, it mines the twitter stream in realtime to republish anyone who is telling the world that they are not at home. Who would want to know this? Take a guess.

 So all you Twitter, Foursquare, Brightkite, Google Buzz, etc. users - keep that good info coming!

Check out my new site I'm-bored-and-broke-so-who-can-I-rob?.com. We feature full integration with Google Maps and their latest realtime street stealview technology^TM© - so you can watch your place being robbed as it happens via our iPhone app (a bargain at $599)!

Defining Analytics: Analytics

Following-up to my blog 'Just Tell Me What I'm Doing', this is the third in 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. Data Warehouse puritans should perhaps stop reading, take a pill and lie down instead.

---

Bennett's Analytica  
   A Practitioner's Guide To Analytics

---

Analytics

---

The simplest definition of Analytics is "the science of analysis". A simple and practical definition, is that it is how a business arrives at an optimal or realistic decision based on existing data. Business managers may choose to make decisions based on past experiences or rules of thumb, or there might be other qualitative aspects to decision making. This is not analytics. When the decision is based on data, then this is analytics.

Doug Lautzenheiser, a fellow blogger on the SmartData Collective recently gave us the following definition:

“Business Intelligence isn't just about seeing how many cans of beans you sold last month.  It's about figuring out how to sell even more in the future.”

I like this, but even more when I swap out 'business intelligence' for 'analytics'. So, just like data warehousing and business intelligence, these are all terms used interchangeably in practice.

Another favourite definition is:

"analytics is reporting with an intent to inform a decision that you want to make." 

or even just:

"analytics is reporting with intent."

But I use this only when I am talking to someone who really doesn't understand much about the subject.

Remember - my world is commerce. I haven't given much thought to other areas like academia and medicine, etc. where analytics is also used extensively. With that in mind, I define the analytic process as follows.

Analytica Illustration

The Analytic Process

To me, it is the process you go through to analyse a problem and make a decision. Here are what I think are the common steps involved:

1. Define the Problem. Example: "We need to improve by 20% the success of our Product X marketing campaigns." If you don't know what the problem is, then 99 times out of 100 you should stop the process and go and do something else that is useful!
2. Generate Hypotheses. Use your experience and knowledge to guess ways in which you can use data to solve your problem. Often you will have several candidate hypotheses. These are 'educated guesses'. Example: "We can better predict which existing customers will buy more products if we can understand better what existing Product X customers have in common."
3. Determine what information you need & gather it. Traditionally this information came from a data warehouse or some other sort of database or electronic file. In recent years unstructured textual data, multimedia and other sorts of data stores are being used.
4. Test sources. Run some basic reports to see if the data can support one (or more) of your hypotheses. If you are lucky, you have access to a Business Intelligence tool like Cognos, Business Objects, QlikView, etc. Often you don't and in these cases you can use SQL to extract data from their sources and import it into Microsoft Excel instead. Then test how good the source data is - or more precisely - determine how suitable the data is to our analytic purpose.
5. Test Hypotheses. Run further reports that are built to show how successful (or not) your hypothesis actually is. It is important to question your results and there are many way to do this. Here are a couple of the most common ways to do this:
6. • Find Root Causes. Look extra hard at the results that are the exceptions that do not support your hypothesis. Apply root cause analysis - this is analysis done in order to understand why the exceptions exist. Exceptions are often the data that do not fit your hypothesis. Example: "70% of the existing customers analysed had purchased a car in the 2 weeks before they purchased a tow bar [US: tow hitch]. But for 30% the towbar was their first purchase - but they only work in cars we sell". Concentrate on understanding the 30%. Example answer: "The tow bar purchaser for the exception 30% of customers, was 90% of the time either the spouse of the car purchaser or their employer".
   • Experiment. Identify a set of customers with something in common but that haven't purchased Product X. Example: "30 - 40 year old women with 2 or more children under the age of 18". Also, randomly select a second group of customers (i.e. those with nothing in common) to be your control group. Then run the same test campaign using both groups and compare the results for both sets of customers.
7. Review the results with your peers to check that the result really is valuable. Is it fit for purpose? Is the effort worth the benefit gained? Can you identify enough customers meeting the criteria to meet the uplift target?
8. Package the process. If you regularly need to analyse the same or similar problems then you need to make the process repeatable by putting it into 'production' and packaging it with additional information about how it works, any known issues, its purpose, etc.
9. Get user feedback and evaluate the results by either tracking the decisions made based on the analysis or more informally. If you do the former, then congratulations - you are operating Decision Management. This is a holy grail for my work. More often you have to settle for informal feedback from the decision makers (or their colleagues). Example: "Analyst to Campaign Manager: So whadya think?"
10. Learn. Use this knowledge to improve the hypotheses, sources, data and process.
11. Repeat the process again until you get the result you want - or give it up as an unattainable goal.

Well that's how I see the process - at a very high level. I went looking online to see what other people thought and there was very little to find in the commercial context. Even wikipedia let me down although it does have interesting things to say about Analytic Hierarchy Processes and the Intelligence Analysis Process used by NSA and the like. There is an old discussion thread on LinkedIn that's relevant.

Analytic Techniques

There are a large number of statistical, mathematic and management techniques that can be used. These will be discussed in a future Analytica entry.

Common Uses of Analytics

Predictive Analytics

Predictive analytics encompasses a variety of techniques from statistics, data mining and game theory that analyse current and historical facts to make predictions about future events.

In business, predictive models exploit patterns found in historical and transactional data to identify risks and opportunities. Models capture relationships among many factors to allow assessment of risk or potential associated with a particular set of conditions, guiding decision making for candidate transactions.

Web Analytics

Web analytics is the measurement, collection, analysis and reporting of internet data for purposes of understanding and optimising web usage – not just measuring web traffic.

Related Terms and Concepts

Refer also to Data Mining

---

Comments? Via form below or send feedback to [email protected]         version 0.1 201002

---

Now That's Ubiquity

Google's ad during the recent Superbowl really drove home to me how search has become part of our lives:

You can also download here in HD the 2010 Google Superbowl ad.

I also reminds me of how limited the impact has been of analytics in my professional life. What is it about the tools, technologies and processes we use in business analytics that are preventing me/us from delivering applications as powerful and ubiquitous as Google?

Staying on the subject of Google, here is an interesting way to look at the company:

As gigaom says:  Google is an advertising company.

"Ask your friends what business Google is in and the answer you’ll most likely get is “search.” And they would be wrong. Google is, first and foremost, an advertising company."

Analytics at Work

Hot off the presses jumps Analytics at Work: Smarter Decisions, Better Results, the new book by Thomas Davenport, Jeanne G. Harris (both of Competing On Analytics fame) and Robert Morison. This book is very much the sequel to Competing On Analytics and this is not a bad thing. CIO Insight magazine named Competing on Analytics one of the all-time “Top 15 Most Groundbreaking Management Books”.

What I like about the book is that it is going after the big picture where better decisions drive outstanding performance, improve competitiveness and help organisations differentiate themselves from the competition.

This time the authors have completed new research on companies in the US and Europe. They include Humana, Tesco, Air France/KLM, Best Buy, Progressive Insurance, and Hotels.com. Using these organisations as illustrations the book is intended to be an implementation-focused guide that describes a five-step model for deploying and succeeding with analytical initiatives.

In the first half of the book five factors are presented as being necessary if an organisation is to succeed at 'doing analytics'. The authors group use the acronym DELTA (the Greek letter that signifies “change” in an equation):

• D for accessible, high-quality data
• E for an enterprise orientation
• L for analytical leadership
• T for strategic targets
• A for analytical talent

This all makes sense to me and then in the second half of the book the five-step model is explained. I will not spoil the authors work by discussing that here.

To me the book is presented as a primer for doing analytics. How well does it achieve this objective? The answer is both 'very well' and 'not really'. Let me explain:

• If you are a C* level executive trying to understand why you need analytics, and you want a high-level understanding of how to achieve this, then the book delivers - and delivers well.
• If you are a Senior Manager directly responsible for making analytics happen, the book is less successful. Not through any fault of the authors. I am one of these managers and I know that analytics is hard. Even a 1,000 page tome and a dozen CD's would not be enough for that.

So I hope that the book will inspire C* level executives that are interested enough to read a 200-or-so page book on the subject. If they only read one book on the subject then this is the one I would give them. For those that can't be bothered, then I can sit waiting for the next crisis to hit and the inevitable change of management to occur!

To the author's credit, I notice scattered throughout the book discussions (on pages 14-15 for example) and insight on how the Financial Industry got risk analytics so wrong - contributing to the financial crisis of 2007. Finance's use of analytics was featured as one of the successes in Competing On Analytics.

Although a Harvard Business Press title, the book is 'tied-in' to Accenture (Harris works for them and Davenport is an Accenture Fellow) and the company has a microsite within their Institute for High Performance consulting practice. Here's a nice brochure from Accenture on Analytics At Work.

Business Intelligence Beans

Doug Lautzenheiser, a fellow blogger on the SmartData Collective has just posted the following definition:

“Business Intelligence isn't just about seeing how many cans of beans you sold last month.  It's about figuring out how to sell even more in the future.”

I like it.

Defining Analytics: Data Warehousing

 Following-up to my blog 'Just Tell Me What I'm Doing', this is the second in 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. Data Warehouse puritans should perhaps stop reading, take a pill and lie down instead.

---

Bennett's Analytica  
   A Practitioner's Guide To Analytics

---

Data Warehousing

---

The overall process of providing Information to support business decision making is referred to as either data warehousing or business intelligence. The terms are used interchangeably and often as acronyms like DW/BI, DWH, EDW, ODW and BI.

EDW stands for enterprise data warehouse. This is a warehouse that stores data about activities in many areas of an organisation. An EDW or DWH is typically used to support the strategic business requirements of an organisation.

ODW stands for operational data warehouse. This is a warehouse whose primary purpose is to analyse and report on operational (day-to-day) activities.

Data warehousing refers to not only the data warehouse itself (see below) but also the technologies and activities to put data into it, extract data from it, analyse the data, and inform users through reports, dashboards, alerts and data extracts.

(image copyright unknown)

ETL

Is the acronym of Extract, Transform and Load data into and out of a data warehouse. ETL, also referred to as data integration is an important area of analytics and deserves it's own section in any Analytica. Stay tuned for a fuller discussion in another post.

Briefly, ETL is the activity of:

• Getting data into a data warehouse:
• • Extract data from source system files. Source systems are usually the transactional systems of the organisation or are files that contain metadata required to assist in the analysis of the transactional data.
  • Transform the data ready to place it into a warehouse. Often multiple sources of data are involved and similar data from different sources must be modified (i.e. transformed) so that they have the same format and/or meaning. When this is completed the data is said to be conformed.
  • Load the conformed data into the tables of the data warehouse
• Getting data out of the data warehouse:
• • Extract data from the tables of the data warehouse.
  • Transform the data so that it is better organised to support the analytic purpose that the business requires to be performed.
  • Load the transformed data into another database (called a data mart) or file ready for analytic or reporting use. It's called a data mart because it conjures-up comparisons with a supermarket - a place where you can easily pick what you want to consume.

Just to confuse you, ETL is now a popular term for loading and extracting data from any data store or application - especially in IT middleware buses where data is moved between applications.

Data Warehouse

A data warehouse on the other hand is generally defined as a database where the data is stored in such a way that it facilitates reporting and analysis. Other databases in business tend to be places to store data for transactional use. This difference in purpose has a surprisingly fundamental impact on the way that data is stored and the tools used to get data into and out of the database.

Another way to think about the difference is to think of transactional use to be operational where the speed of recording a transaction is important. Data warehouses on the other hand are used to support decisions where speed of recording is less important than our ability to analyse. This is useful but just to confuse you, data warehouses can also be used for operational reporting and analysis.

Most data warehouses store data in rows within tables but a growing number use columns (see Columnar Data Warehouses below).

Data Mart

Data marts are data stores whose data covers specific business functions for a specific community within an organisation. For example a finance or sales data mart. Data marts are often subsets of data in a data warehouse. Data marts simplify analytics as they:

• Remove data that is not relevant to the intended type of analytic insight required by the business.
• Pre-calculate measures that are frequently needed. This means that the measures can be standardised, reports are simpler to design, and run more quickly as they don't have to do as much calculation.

Data Warehouse Appliances

A DWH Appliance is an integrated set of servers, storage, operating system, database and software that is-optimised for data warehousing. Most appliances today use massively parallel processing (MPP) architectures to provide high query performance and platform scalability. In this respect appliances use the same architecture of the traditional very large scale data warehouses from Teradata. Appliance vendors include Netezza and Greenplum.

Normalised versus dimensional Storage Models

A storage model is the way the data is stored in a database.

A normalised model is one where the tables are grouped together by subject area. Each area containing a data category. Examples are data on customers, products, and the like. The advantages of normalisation is that it is easy to load information into the database. The disadvantages are that the database can end up with a large and confusing number of tables. This often makes it difficult for users understand the data and it also slows down the retrieval of data.

A dimensional model is one where transaction data are separated into 'facts' and 'dimensions'. Facts are typically numeric and transaction data. Dimensions are the reference information that gives context to the facts.

An example (from Wikipedia) helps understand the difference:

"A sales transaction can be broken up into facts such as the number of products ordered and the price paid for the products, and into dimensions such as order date, customer name, product number, order ship-to and bill-to locations, and salesperson responsible for receiving the order."

The advantages of the dimensional approach are very big advantages. The first is that the data in the warehouse is easier to understand and use. Secondly, the retrieval of data from the warehouse is fast.

The main disadvantages of the dimensional approach are that making the data easy to understand comes at the cost of a more complex ETL load into the warehouse. This is because maintaining the integrity of the facts and dimensions is complex. The second disadvantage it that it can be difficult to change the data warehouse structure if the organisation changes the way it does business.

Another tricky aspect of the dimensional approach are Slowly Changing Dimensions (SCD). These are dimensions that have data that slowly changes (in case you didn't guess). An example best illustrates the problem:

A common dimension in a data warehouse is one that tracks the sales records of your company's salesforce. Reporting and analysing sales performance should be straightforward but what do you do when a sales person is transferred from one sales team to another. How do you record such a change in your sales dimension so that you remember the past and the present composition of your salesforce? This is important when you do all sorts of performance analytics.

Solving these (and a range of similar) issues involves methodologies referred to as Type 0, 1, 2, 3, 4, and 6 SCDs. The solutions to handling these types are outside of the scope of this post. Stay tuned for a fuller discussion in another post.

Alternative and Emerging Ideas

Columnar Data Warehouses

Wikipedia says it best:

"A column-oriented DBMS is a database management system (DBMS) which stores its content by column rather than by row. This has advantages for databases such as data warehouses and library catalogues, where aggregates are computed over large numbers of similar data items."

The Bus Architecture

A bus in computing is a subsystem that transfers data between computer components or between computers. It has a more specific meaning in data warehousing. Ralph Kimball, one of the founders of data warehousing explains it best. Here is my mashup of his ideas:

You can think of the conformed dimensions and facts of an enterprise as a standard set of connection points for applications — in other words, as a data warehouse bus architecture.

The corresponding bus matrix identifies the key business processes of an organisation, along with their associated dimensions. Business processes (typically corresponding to major source systems) are listed as matrix rows, while dimensions appear as matrix columns. The cells of the matrix are then marked to indicate which dimensions apply to which processes.

In a single document, the data warehouse team has a tool for planning the overall data warehouse, identifying the shared dimensions across the enterprise, coordinating the efforts of separate implementation teams, and communicating the importance of shared dimensions throughout the organisation.

Data warehouse bus architecture. The bus architecture is based on standardised dimensions and facts that let separate data marts, fact tables or OLAP cubes coexist and integrate. In this context, the term bus is not a large motor vehicle. Instead, it refers to the term's early meaning in the electrical power industry (a conductor for collecting electrical currents and distributing them) and commonly used in the computer industry to describe the standard interface specification that lets peripheral devices usefully coexist.

Ralph Kimball is credited with associating these fundamental bus concepts to the data warehouse delivery and presentation environment.

Related Terms and Concepts

Refer also to ODS, Operational Data Store

Refer also to Metadata

---

Comments? Via form below or send feedback to [email protected]         version 0.1 201002

---

Defining Analytics: Data, Information and Knowledge

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

---

Comments? Via form below or send feedback to [email protected]         version 0.1 201002

---

Big Data

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.