Business Intelligence Notes

Lecture 25 & 26 Visualization and social media monetization

2012-04-21T12:21:00.001-07:00

Key takeaway from the video, Hans Rosling shows the best stats you've ever seen

Observe data in different dimensions trend/moving pattern over time, record the moving paths to perceive past trend and predict future pattern (Use data to do projection based on the past pattern)
Visualize data based on reasonable measure, such as population
Show graphs by both Aggregate data and split data

Dangerous to only use average data (can't see the big picture), since the details can be much different from average. Also, average can be affected by outlier and misrepresent the truth.

Clearly show over time and regional pattern
Link data with internet the draw the whole picture and tell story about the data
Explain the behavior of the data - not just how data is moving but why
Volume of data makes things different - "big data" to be representative
How to Access to data and Cleaning of data are important
Present data in a visually appealing way
Data is not always in/about a company, rather, data outside the company, such as industry trend, is also important to predict future.

Monetize social media

make $ out of social media, not about networking
taking the data from end users and process the data, then selling back to end users
Followings are example of social media monetization

Google: advertising

Adsense

Google plus

to get more information/user preference to refine their business/Ads

Primary use search data

IP, how long you spend on the page

Facebook

Business model: advertising

Know your individual personal information (everything about you)

Acquire "instegram" - easier to access to your mobile photo

Facebook insights

Twitter

Firehouse allow people to access tweets on a real-time basis

Sell you data back

Sponsor users/brands/tweets

Premium accounts (Premium means that initially you provide free service and then sell additional services)

Recruitment from HR: lots of recruiters subscribe premium account (indirectly sell user data)

Advertising

Background check/references

ZYNGA

Gaming

Few levels for Free first, then sell premium

FOURSQUARE

Location based

App on mobile

Check in for the location

Lots real time data: Who/time/ location

Partner with AE/twitter, working with local small business

If people check in enough, you get reward

Sell data to business

If you use specific hashtag about the business, you can earn credit in AE

Provide incentive to make end users voluntarily Syncing your card with twitter and foursquare, providing your information about purchase tendency and frequency, who your friends are

Dropbox

Social - share

Media - storage

Sell more storage space

Individual: use different devices to access

Easily share files

They get information about who you are sharing files

Everybody use it, high switch costs, so buy more space

Amazon

Computation

Visualize me (infographic resume generator)

Partner with, for example, linkedin

Sell you back these visualization information

Groupon/social living

Partner with business

Act as a consolidator, like a whole seller working on social media

Users tell them what they are interested, Groupon send you coupon about that area

Group buying/group discount

URL shortener

Use it and post on twitter, since there is word limit in twitter

Free service for users

Track the URL to see how your post diffuse, show your statistics

Initially Partner with twitter, which has its own URL shortener now

Now partner with other, most are news

Trip adviser/yelp

Subscription from business

Attributes needed to consider before starting a new social media application

What is the data I can get

Unique Value proposition

What is going to sustain you over time

Primary audience

Takeaway from GOMC Competition

2012-04-21T11:31:00.000-07:00

From a company's perspective, performance of campaign is the most important thing. Therefore it's critical to identify goal of campaign and estimate what the costs and benefits are to measure the overall ROI for the whole campaign. While comparing the performance of campaign, not only compare to pre-campaign but also to post campaign, to show the contribution of the campaign. If the conversions are much higher than in both pre-campaign and post-campaign period, that means that campaign did increase conversions, helping organization achieve goal.

For the pre-campaign comparison, we should not only compare the campaign period with previous month but also the same month in the past year. Comparing to the same month in different enables us to capture the seasonality that is not able to see in only compare months in the same year. We will know whether the increase of conversion is due to campaign or just in hot season. On the other hand, comparing the the months just prior campaign allow us to get the overall scale of the visits and conversions. We can perceive that whether the increase by year is actually attributed by campaign rather than just business expansion.

The timing of campaign is also a crucial factor to consider for campaign success. That's the reason why we need to do market analysis before build campaign strategy. Thoroughly understanding the rush and negative seasons help us to decide when and how to execute campaign. For example, if you promote game outfit out of any game season, it's not hard to expect the ROI would be low, since people don't have incentive to buy those game related products when there is not game playing.

Lecture 23 & 24 Dimensional modeling and balanced scorecard

2012-04-20T10:45:00.003-07:00

Dimensional modeling Miscellaneous details

Slowly changing dimension

Purpose: Changes happen in attribute's value over time (e.g. brand, department, …), but don't happen frequently, rather happen once a while

Type 0: nothing is ever going to change (this is not typical)
Type 1: directly update attributes, completed lose history but simple
Type 2: in order to preserve history, split tables

Every time value of attribute change, add a new tuple with a new surrogate key, add start time and end time to the new tuple

Type3: make changes to the table, Add a new column/new attribute called "new" territory, and change the previous one as "old" territory

Rule playing dimension

Junk dimension

Put data altogether

Semi-additive

Non-additive: average, minimum, maximum, can't add together, e.g. balance of bank statement

Balanced scorecard

use a bunch of factors to evaluate company
Balancing different measurements to come up the overall performance

Can be applied to any level of company

BI is all related to performance management, which is the core function of balanced scorecard

4 perspectives (interacting with each other, the rest 3 factors are all linked back to finance)

Finance: how well are the company doing respective to shareholder

Customer

Internal process: align with business

Learning & growth

After evaluating the company via balanced scorecard, we can clearly know performance of company in different fields, allowing us to draw Strategy maps to understand How we are going to achieve goals

Lecture 19 & 20 Develop dimensional modeling for AYFG case

2012-03-28T12:29:00.001-07:00

AYFG is a gym with several branch nationwide. Provided ER diagram, we are going to develop a dimensional model step by step to answer the business questions and address business problems.

The first idea is star schema. Figure 1 below is the conceptual schema of the business. We can see there are several primary entities, including members, memberships, salesinvoices, merchandises, etc. An interesting thing we noticed is that there is only merchandises sales has quantity attribute. The reason of this is based on the assumption that sales quantities of other products, such as memberships and oneday pass, are defined as one for every record.

To create a star schema, we need to thoroughly understand how business works and define dimensions and fact first. The purpose to analyze this data is to figure out how to boost sales and profits, depending the past record to target most valuable customers. The information that can help us to understand customer's behavior is purchase pattern and trend, which can be extracted from transaction records. Therefore we define fact as transactions, including event, membership, and any purchases. Attributes in the fact table include surrogate keys from dimension tables, including times, products, members, etc. The common shape of dimension table would be short and fat, consisting of a lot attributes but not tons of data. On the other side, fact table is usually skinny and tall with few attributes and uncountable records. To achieve the objective here, we will need to execute ETL process ( Extracting tables, Transferring into different tables, Loading new tables which are different from the original extracted tables).

In addition, followings are a few terms that are frequently used in dimensional modeling.

Factless Fact table: a table contains nothing but surrogate keys. Frequently used in promotion to associate relationship among dimensions.
Transaction ID (degenerate key): an attribute in fact table just to order and count records.
Market Basket Analysis: using a model to figure out what products are more likely to be purchased together.

Lecture 17 & 18 Dimensional Modeling

2012-03-28T11:51:00.003-07:00

Data cleansing and data profiling take around 80% of data analysis process, playing an extremely important role. The first thing we need to do is to look at the schema and try to understand what information the schema and attributes tell you. We can come up a graph to visualize the relationship among attributes. A few things we can check before analyzing data, are investigation of each attribute on each table, business rules that govern every attribute, such as referential/data integrity, combination of attributes, and carnality. Most common situations that we need to profile are data from different tables in different format because of merger and acquisition or different information system.

We will create system catalog in the data warehouse. Each database has a list of tables. Each table has a list of attributes. Further, Each attribute has a list of constraints. For constraints, it's a good practice to define a default value for each constraint. Otherwise, once you want to merge multiple tables with null value for the attributes associated with constraints, we will need to turn off the constraint to make data merge. The common case would be people forget to turn on the constraints, which might be used to follow data integrity, again, causing a mess of data.

Numerous vendors alleged that they provide MDM (Master Data Management) tools to help companies extract valuable information from the Big data in the information system. Things that companies usually not know is that MDM is largely a human efforts, that is, what we are looking for is not about tools but services. Profiling is a large part of the MDM, extracting, transforming, and loading data.

The core idea of dimensional modeling is star schema with a few factors, primarily including fact and dimensions. From the graph below from Laynetworks, Facts is like the heart of the whole data and all dimension tables connect to fact, so the referential relationship is one to many between dimension and fact. The size of dimension tables are not always fixed, depending on different types and characteristics. Only time dimension table has stable table size since it only relates to the granularity of fact. The maximum size of fact table is product table times customer table times time dimension table, which is the best case that every customer buys every product every time unit.

Source: http://www.laynetworks.com/Diamension-Modeling5.htm

Lecture 15 & 16 BI infrastructure

2012-03-19T12:07:00.001-07:00

OLAP is Online analytical processing, which is used in most BI infrastructure to look at the trend of purchase, aggregating query data/large amount of data and analyzing them.

OLTP is Online Transaction Processing, which is regularly used in operational database, analyzing single query data and small amount data.

There are two primary BI infrastructure. First is data warehouses, which was first brought up by Bill Inmon. The concept is starting from a big enterprise system, eventually divide into small division data marts. All supply chain systems are pulled into data warehouse. The other one is data marts, which was proposed by Ralph Kimball. The idea is starting from small data marts and come together to a big enterprise system. The two actually generate the same result, but different approaches. In the modern age, Kimball is more common.

Big data is generated everyday, but dirty data is everywhere, so data profiling becomes a very important part to help make good decisions. The fundamental assumption of data profiling is that the data is not in the form you want to use, so we need to profile them and extract useful information by data quality analyzing tool. There are a lot methods to cleansing data. The regular process is ETL (Extraction, Transformation, Loading). The common problem of data cleansing is the latency. Nowadays, people try to shorter the latency to a few hours to achieve real-time, but still can't get the real "real-time". Furthermore, the more real time model you need, the more price you need to pay, so maybe capability is enough to do it but it's to expensive.

Some companies will create an ODS (operational data store) between srouce and data warehouse to do profiling and cleansing by reorganizing the data to fit the form of data in the data warehouse.

Lecture 13 & 14 Network Analysis

2012-03-04T14:10:00.002-07:00

Before Analyzing network, we need to have to know basic structural properties. Centrality measures are used to examine how close the relationship in the network. In the centrality, we use different measures to how the network connection.

Degree Centrality: counting the number of link, i.e., how many people can a particular person like directly. In- and out- link are counted as the same in the degree centrality. If every node in a network has both in- and out-link, we call that fully connected network. Clique means that every node connects to every other node in a network. If a network is not "clique", there must have some bridges between unconnected directly nodes.
Between Centrality: How likely is a node to be in the direct route between two nodes
Closeness Centrality: the distance of a node to link to all other nodes in a network
Eigenvector centrality: How well is a person's network overall. It bases on the influence of nodes to assign score of each node.

In addition to centrality, we can use size to measure a network, calculating the number of nodes in a network. Through diameter, which is the longest shortest path of the entire network, we can see how big a network is and what the distance among nodes. Reciprocity is another aspect to interpret a network by directionality and clustering coefficient. The higher clustering coefficient, the more density of network (more connection). Nodes that are similar to each other tend to connect to each other, that is, people connect to the same person are more likely to connect to each other in the future.

To learn how to analyze social network, we can start from analyze our own personal Facebook network using the netvizz app on Facebook. Then we can use Gephi (a network analysis tool, download here) to visiulize our personal network, following the quick start guide. You will be able to see the similar graph like below. Different colors represent different communities in a group.

Eller MBA Facebook Network

Lecture 12: Graph Theory

2012-02-22T15:28:00.001-07:00

Like 7 bridge problem, graph theory is a study of how nodes in the network connect to each other in the most efficient way. Two primary factors in the graph are vertices and edges. In this modern age, we are interested in information networks, which is the network moving information from one node to another. There are some terminologies that we might need to know before we go further to the detail of networks. First, collaboration graph visualize how people cooperate with each other, linking every pair together if they cooperate with each other. Path is a sequence of path that link nodes. In the Internet, we choose the shortest path/minimum path, which goes through the least number of nodes.

We also interpret graph by degree of network, which is the number of link connecting to a node. The more degree the more strength the connectivity is. Divided by network direction, degree include indegree, request from others, and outdegree, request to others. The number of nodes in the network is also an important indicator to interpret graph. An inference of degree is the small world phenomenon, which is also known as six degree of separation, defining the whole world graph as a giant component. Using the least numbers of nodes to define the distance between two nodes, the small world experiment found the average degree is around 6 between a person and everybody in the world, that is, we can connect to everyone in the world through 6 people. Interestingly, Facebook revealed that currently the average degree became 5, which means the world becomes closer. Small world phenomenon can be observed when you are looking for a job, but surprisingly, weak ties with less strength connection can help you more than the strength ties.

Lecture 11: Long Tail Keywords

2012-02-22T13:27:00.002-07:00

On 2006, Chris Anderson published a book called "The Long Tail". He stated that the society is moving from a high demand market portion toward a number of small "hits", accumulating a huge portion in niche market tail.

Source: http://www.longtail.com/the_long_tail/about.html

We can utilize the same concept in the keyword advertisement. Instead of biding high price broad keyword with lots competition, we can bid on numerous specific keyword phrases with much lower prices to bring more conversion. The reasoning is based on the customer buying cycle. When a customer start to have an interest in buying a product, say camera, he begin to search online using board keyword "camera". After he figured out what function he want, he start to add more specific keywords, such as "full HD", "NIKKOR Lenses", etc, narrowing down to a list of particular products. Then he will do more search to compare among the list. At this point, the customer is at the end of the buying cycle and is more likely to purchase. We want to use this characteristic to target customers at the end of the cycle, placing specific keyword phrases related to the conversion we want visitors make to increase conversion rate. The assumption under this thread is that the keyword phrases are longer/specific in the long tail. The strategy is suitable for small business whose conversion is selling product, increasing cost-efficiency.

Google provides several keyword matching options for us to tailor our ads. For the broad keyword, people can use broad match. As long as one word user entered is matched to one word in your keyword phrase, the ad will showed up. Phrase match is that the ad will show up when your keyword is fully match full or part of the keywords user entered. For the long tail keyword, it's recommended the use exact keyword, which will only show up when your keyword is exactly matched to user's keyword. One another special function is negative keyword. We can utilize it to avoid the undesirable visitors.

To improve ClickThrough Rate (CTR) and conversion rate, we would like to lead visitors to the page that match their demand. Keyword assertion can help us achieve it using dynamic feature of ad. Several user might enter different keywords in the same category, so they will see the same ad box. After clicking, they are led to the different pages, which is related to the particular keyword they entered.

Lecture 9&10: Pre-Campaign Report & Optimized Publishing in Social Spaces

2012-02-15T21:01:00.001-07:00

Generally, pre-campaign report is used by judge to understand the campaign structure/strategy, company background…etc. With the current situation of company, we can measure the baseline case and compare with expected benefits to show the value of campaign. Avoiding abstract phrases, we should provide precise quantitative numbers to measure campaign and make sure you know what you really mean about objectives. For example, how many visitors will be attracted by campaign, or what’s the conversion rate. With concrete figure targets, we can examine performance on real time and adjust movement depending on different situations. While analyzing client industry and current situation, we can not only research industrial data but also incorporate traffic information of website to predict future trend.

For campaign budget allocation, timeline and different campaign are two major dimensions that we need to consider. Some campaigns are related to current or future event. The timing of events will affect the timing and frequencies of ads. Also, we should prioritize all campaigns, identify segment that client most focus on or group that provide more benefits/profits. It’s better to use table to show your priority decision making process and budget allocation, since people will be easy and clear to compare side-by-side. When measuring market competition, remember to use Google AdWords keyword tool and traffic estimator to see how's the competition status in the industry and cost per click. You can type keywords either from product information or Google Analytics into the link below. The tool will show you the critical information about keywords.

Google Adwords Keyword Tool: https://adwords.google.com/select/KeywordToolExternal

SocialFlow
A scientific approach to measurably increased engagement.

How to use information to optimize publishing to get attention?

Publish is not problem, getting intention is a problem

Scope your audience

What do my content tends to be better for them

Large size does not equals more engagement.

Lecture 8: Design an Ad Campaign

2012-02-09T20:21:00.000-07:00

Advertisement is part of marketing. The fundamental idea of marketing is to convey your message to customers and expect conversion. The same idea applies to either traditional ads or online ads. Initial analysis is non-trivial for a marketing campaign. We need to thoroughly understand client's business, industry, and products by marketing researching and constantly communicating with client. The most important is to know business objective clients want to achieve and the potential customers/audiences. Industry information is critical as well. Understanding the industrial trend and competitors help us to design the content of ads.

There are two kinds of online network advertising. First is search network, which is the ads show up after you enter keywords into the search engine. Google will base on the keywords people input to dynamically decide what ads are going to present. Another is display network that ads present on any kind of websites. The content of these ads depend on the content of the website. One of the advantages of online advertising is that they actually present ads to potential targeted audiences, increasing conversion possibility. To design a online advertising campaign, after market analysis, we should decide campaign structure, dividing customers into several groups and targeting distinct groups by different approaches. Since the value provided by each group will be different, how to allocate budget is important. We need to make sure we emphasize on the right group, which might be the highest value group.

To know position of your website, Google provide a page for people to check their page rank. Then sponsors can have a measurement to inspect whether the campaign they design is effective. Commonly, if you can get a rank 8 or more than 8 is really good but hard to achieve. For example, ebay only gets 7 out of 10 and Amazon gets 9. An interesting one is that Google evaluate itself only 9. I think they might need to modify their standard. Go and check the rank of your website: http://www.prchecker.info/check_page_rank.php

Lecture 7: Advertising

2012-02-09T16:30:00.002-07:00

Traditional advertising, such as TV, radio, billboard, print ads, usually use impression as a unit to measure price and value. Impression is the frequency of advertising showing up. Traditional commercial media charge sponsors based on cost per thousand impression or mille (CPM). Super bowl is an extreme case that cost per impression is very high. All these traditional advertising focus on the frequency of ads appearance rather then the effectiveness of the commercial. Sponsors can spend a great deal of money but do not obtain expected return, since the budget spent on advertising is not positively related to income, placing a huge risk for a firm when spending millions dollars to advertise via traditional media.

In the current age, Google introduced a totally different view of advertising. Instead of requiring a fixed fee up front, Google charge by clicks rather than by impression, so sponsors will not be charged because of their ads showing up on the search list. To maintain the quality of its search engine, Google ranks ads by a function of charge per click (CPC) and quality score, so people can not raise their ads' rank by simply direct tons of traffic to website. Capturing irrelevant traffic just for increase in CPC is not rational. First, it will cost you a lot of money, since Google charge you by click. Second, people will earn a opposite effect, lowering rank of ads, since based on visitors' behavior in the website, Google will interpret the website is not effective and lower its quality score. Google does not reveal the factors of quality score. The only information we know is that click through rate (CTR) is positively related to quality score. Actually, it's a mutual beneficial mechanism. Increase in CTR means that more visitor conversion in the website, helping sponsors move toward business objectives. Also, Google improve its search quality by increase in CTR, that is, users get more effective information they want.

To sum up, although online advertising seems more appealing than traditional advertising. We can't deny that traditional advertising market is still necessary, especially TV commercial. People spend time watch TV show, sport games. Online advertising provides an alternative for companies but is not able to replace the traditional ads in short term.

Lecture 6: Network Analysis

2012-02-01T16:11:00.001-07:00

Social media networks grow rapidly. Business, especially marketing folks, want to utilize Big Data to distribute customized message to potential customers. People want to connect to each other to strengthen their individual social networks, building relationship. To achieve those objectives, more and more demand of networks analysis increase. To do network analysis, first we should think about what kind of network we are going to construct and what are we going to define node in the network. Then, how do we connect those nodes? Social media connect people as nodes. This post will focus on two current popular social media in the United States: LinkedIn and Tweeter.

LinkedIn is a professional social media, allowing users to upload resume and connect to each other in related fields.
A special application of LinkedIn is LinkedIn InMaps (inmaps.linkedinlabs.com/). Inmaps graphically shows how your own networks looks like, using distinct colors and clusters to group your social networks. We can observe the structure of InMaps to recognize the density or degree of the relationship, growing networks. By tracking the map over time, we can see our behavior and how we build network.

Example of InMaps: My InMaps

Second, Tweeter is a short message posting social media, allowing people to follow others. The most obvious difference between LinkedIn and Tweeter is that Tweeter show the direction of relationship among people by indicating followed or following. Interactions among people are also easily to see by observing tweet and retweet. Users use hashtags to show a group discussion for people in the group easily to find the thread. A lot of data are out there. How to retrieve useful information from data is a critical issue, so now several social media analysis websites are built, such as the Archivist (http://archivist.visitmix.com/), saving and analyzing tweets.

Example of Archivist: Search for "MIS587"

Lecture 5: Website Analysis Process

2012-02-01T12:39:00.000-07:00

Ten years ago, websites are only for web presents. Now, we not only want people to visit our websites but want conversion, that is, people visit our website and do what we want. To improve the website, we should analyze the website and see what we can do. Before starting to analyze visit traffic, there are some critical steps we need to proceed.

First of all, thoroughly understand the website in both content and structure. Knowing the content helps us to familiarize with organization's business and core competences, and what message it wants to convey. From the structure of website, we can see how an organization navigate visitors within website and generate improvements that make website more friendly and goal oriented.

Second, it's important to know the objectives and targeted audiences of website. With objectives, we have direction to make improvements, inducing visitors moving toward our goal, such as increase in our sales. Different organizations have different properties and targeted customers. For example, firm selling kitchen accessories may target married women rather than teenagers. So we identify targeted audiences, and focus on them to design website and marketing events, making website work more effectively toward our goals. Therefore, blindly directing traffic to website by inserting irrelevant keywords may not be a good idea. If you only want high volume traffic, that might be fine, but if you want visitors to do what you want, such as printing online coupon to purchase products in your store, you'd better to hit their needs. But how to know the goal? We can collect requirements from our clients, knowing what they need, or we can observe business/operation and discuss with clients to narrow objectives aligning the mission of organization.

Then, we can start to log into the Google Analytics to observe the characteristics of traffic visiting our website, but remember that the first thing to do as log in is to setup a long period to see a long term trend, seasonality, and how the traffic fluctuated. Only from a long time period, we can see some periods have much higher or lower volume of traffic than the rest of periods. Focus on these specific periods to figure out why and what bring visits in or out. By doing that, we may get visitor's incentives to come to our website and make specific recommendations to achieve the goals.

With the Google Analytics, we can monitor visitor's behaviors by using combination of metrics to interpret situations, say, why people doing that after entering a particular page. Marketing folks can track how their promotions work by comparing prior- and post-event traffic and other KPIs, taking lesson-learned to improve future marketing events. Finally, with these empirical metrics, we are able to make recommendations supported by solid quantitative information.

Lecture 4: Analysis Presentation of Eller MIS department website

2012-01-29T17:40:00.000-07:00

Presentations are focused on recommendations to the alumni and partnership sections of the Eller MIS website to improve each section. Most teams merely introduced the general findings in the Google Analytics and made recommendations without considering the targeted audiences and the goal of the website. As Dr. Ram and Ms. Anji Seigel mentioned, we should identify the goal and the targeted audiences before we start our analysis and making recommendations. Otherwise, we actually don't know what we are working on, since recommendations should be tied to objectives.

In addition, most teams did not identify the timeline of analysis. Long timeline can see seasonal trends of the website and make long-term strategy. Short timeline is usually used to control a campaign or event to see whether the event effectively attracts users to the website and help the company achieves the goal. Therefore, using different timeline can observe different types of information from the statistic numbers.

Lecture 3: Google Analytics

2012-01-29T12:13:00.000-07:00

Before access into Google Analytics account, we are better to know the goal that we want to achieve and what the KPI for the goal is. When you first time log into the Google Analytics, setup a long period time slot to overview the fluctuation of traffic over time. We should be curious about what reasons or events caused the fluctuation, since the factors made a peak in previous period can an important element of our future element. Several critical terminologies of Google Analytics as followings were introduced during the lecture, helping web/data analyst to interpret visitor's behavior.

Landing Page: The first page that visitors come into the website. It tells us the page that leads visitors to the website, so we can base on this information to improve pages and lead visitors to where we want.
Exit Rate: only for sections or pages in a website but not reasonable for the whole website. Since visitors ultimately will exit website, the reasons they exit may not be website but personal factors.
Bounce Rate: Visitors go to the website and exit without visiting any other page (single page visit to the website). It's important but can't interpret isolated. For example, if a purchase page, which is defined as a final page for buyers by the website, possesses high bounce rate, it may not be a bad sign. However, if a homepage has low bounce rate, mostly it's not a good sign.

I felt it amazing when I first time used the Google Analytics and thought that I can get all information I want in the account. After the lecture and actually using it frequently, I realized that metrics can just assist us to interpret some phenomena and support our arguments, but there is no absolute number fit every situation. That's why humanity is a key factor in BI.

Lecture 2: BI and Web Metrics

2012-01-20T17:50:00.000-07:00

-What is BI (terminology)?

Business Intelligence (BI)

Tools, technologies and techniques used for collection, measurement, understanding, analysis, and prediction using data (internal and external) in some way for performance management.

Performance Management

It’s more than measurement. First identify target and define key performance indicators (KPI), some probably can quantify but some might only qualify, to measure performance. Then based on data analyzed and metrics, we can take appropriate strategic action to reach the target and manage performance. (Monitoring, Managing, Metrics)

KPI is the most important thing manager wants to measure, so what kind of data needed to collect depends on elements affecting KPI.

-BI Life Cycle

-Intro to web analytics and social media analytics

-PARADIGM SHIFT: BI, social media

The purpose of network analysis/web analytics (measurement) is to make better prediction (about trend) to improve performance, by understanding what happen/why happen/who make it happen/ pattern.

OLAP: online analytical processing

Data Warehouses/Data Marts

Big Data

Real-time: immediately

Billions of users make dynamic interaction and generate a large amount of data in many different sources, such as BLOGS, MICROBLOG, FORUMS, REVIEW SITES, FOURSQUARE, WIKIS. Therefore, instead of sampling, now we can observe the whole population interacting. Web is like a laboratory for pulse of humanity. Since data out there is unstructured, the quality of data or whether the data can generate useful information relies on data profiling and cleansing.

Web Metrics (6W): what, why, who, when, where, which