Week 2: Science of Analytics
Data
- Business captures moments of transactions to deliver economic value
- Identify problem
- Who’s customer? Frequency? Products bought? Loyal? Valuable
customer? How to learn more? Strategy to bring customer in?
Principles of Problem Framing
- Tell interesting and complete story
- Meaningful, solution reused for related issues,
assumptions/boundaries
- Appropriate solution framework
- Break down problem, iterative agile way, analytical/modeling
techniques
- Routinize procedure
- Document, next similar questions solved, build a system
Analytics
- Process of developing actionable decisions or recommendations for
actions based on insights generated from historical data
- Combination of computer tech, management science techniques,
statistics
- Agile approach
-Perform business discovery→data discovery→prepare data→model→ score and
deploy→evaluate and improve
Basketball Analytics
- Numbers to show where to shoot ball
- Show data and make decisions during half time
- Camera to show every player’s movements
Cost for each bad decision
- Tinder model = reduce cost of bad decision
- Data is new oil and Analytics is new engine = gain competitive
advantage
- Bring insights and create value
- Data = collect, clean, analysis, communication
-
- Clean = take out outliers, Analyses of day to day, Communicate
to those responsible of operations
Data collection/methods
- Primary = survey, interview (marketing research)
- Secondary = proprietary database, internist data (crawlers, screapy),
stock/capital market (compustat, CRSP), accounting disclosure
(10k,10Q)
- Simulated = assumption and simulation (scheduling, routing,
queuing)
Hypothesis/Data/Info
-
Data preparation for analysis
- Extraction = extract from primary/secondary source
- Transformation = clean into proper format/structure for querying and
analysis
- Load = load into final database, operational data store, mart or
warehouse
- ETL = time consuming, done in parallel
Types of Data Analytics
- Descriptive = sales since last year? Customer frequenting more? What
buying?
- What happened, standard reporting, dashboards, visual, well
defined problems and opportunities
- Parameters describe data (mean, median, standard dev)
-Look at all the data in hand
- Explanatory = why sales surge in Nov? Customers not buying
product?
- why/how it happened, inferential statistics, visual,
discover/understand relationship of outcome
- Correlation, different variables, causality vs correlation
-Look at sample of data
- Predictive = likely rev in 5 years? Who will respond to ad?
- Happen next? Data mining, text mining, forecasting, accurate
projections of future outcomes/events
- Demographic factors
- EX: predict - water leakage, depression, criminal activity, stocks,
demand
-Use sample representative of population but never 100% certain
of end results
- Prescriptive = certain product not shelling good? Manage inventory
with low demand?
- Done about it?optimization, simulation, expert systems, best
possible business decision and outcome
- Best course of action with complex decisions
- EX: credit score, asset management, wine company with grapes
- Develop analysis with prescriptive analytics
What’s next for data - Fast forward labs
- Business analytics = interpret and package data
-Decisions are evidence based, data driven, better
- Machine learning = what questions to ask
Sports Analytics
- Gather data, Analyze data, Visualize data, Explain significance of data
VLOOKUP
- TRUE = approx match
- FALSE = exact match
Powered by TCPDF (www.tcpdf.org)
