Please enable JavaScript.
Coggle requires JavaScript to display documents.
HUMAN OBSESSION WITH PREDICTION - Coggle Diagram
HUMAN OBSESSION WITH PREDICTION
PREDICTION
2000 year history
Ancient Greeks thronged to hear predictions from the Oracle of Delphi
prediction has bounced around among common sense, nonsense, and wishful thinking
Prediction turned scientific only about three hundred years ago with the emergence of statistics in England
Hyper-progress around World War II (1939-1945)
The bread and butter of analytics is predictive models
relied on historical data
Firms got better at collecting more data
Spreadsheets gave way to data warehouses, and predictive models increasingly became a reality check for managerial intuition
Good predictions needs sound data and sound models
led to a flurry of innovative practices in all functional areas
segment markets more finely, tier customers on their lifetime value, and forecast demand for a particular product on a particular day well enough
GOOD PREDICTION NEEDS
Sound data
Data is the raw material of analytics
It is raw facts, which become information when they are sorted, condensed, and contextualized
Insight, or knowledge, is simply actionable information
e.g temperatures in Fahrenheit for your zip code tomorrow makes them information
Insight which is Actionable information
e.g need to wear a warm coat to work tomorrow
Unreliable data produces untrustworthy insight
Collecting, cleaning, organizing, and making data flow are largely the IT unit’s job
Not all information it produces is competitively valuable
Sound reasoning
Software apps transform data into business insight
Insight does not emerge spontaneously from data without effort to transform it
underlying reasoning is based on functional domain knowledge, which can come only from non-IT managers
classifying a customer in a high-maintenance group requires a marketing manager’s insight into what predicts membership in that group
Turning raw data into actionable insight must begin with questions
non-IT manager, it can help identify what business outcomes are competitively important to predict, within what time constraints, and which model would use to predict