Want More Information?
Need Tech Support?
1. You don't know what actually ‘‘causes” anything, and you probably never will. The sole product of prediction is reliably precise and accurate calculation or derivation of values at some future point in time. “Causality” is a worthy, long-term goal which may not need to be established before improvements are made, and actually bias predictions toward current dogma. We don’t know the exact cause of Migraine Headaches, and yet we already have remedies to provide relief.
2. Prediction is always a probabilistic assertion that behavior, as measured, will equal reality as perceived. It is never certainty and only approaches reality in large numbers and with homogeneous measurement.
3. The first, middle, and last focus of any task is exhaustive understanding of the question you are trying to answer. Fascination with our own technology and methodology often clouds the focus of task. You are ineffective if you provide the right answer to the wrong question, no matter how pretty the answer is.
4. Predictive analytics must be an unbiased perspective, not an exercise to validate existing hypotheses...that is statistics. While it is always critical to understand the subject matter under study, predictive analytics is its own disciple, which when employed correctly, provides data with a seat at the decision table and thereby true information value.
5. All slop cancels. Collect and treat rigorously, especially that which you cannot measure precisely. Don’t avoid data sources that are quantifiable but intangible. Just because a source is fuzzy, don’t treat it as a lesser potential predictor. Truly innocuous data with no actual prediction value will fail to predict reliably anyway; vital fuzzy data may overcome the noise to predict reliably.
6. Static models are useful for showing what happened. To predict, models must constantly adapt as environments change. Information value erodes rapidly and erratically. Life doesn’t wait for formulas to be tweaked, tested, and tweaked, because most of life, business, and their systems are dynamic. Real-world Predictive Analytics should not be measured by accuracy alone, but rather speed of adaptation to environmental change.