Predictive analytics and machine learning take information that is readily available to you and make predictions about future events based upon your data, existing theories, algorithms, and probability. These approaches may help identify future opportunities and threats in ways that allow you to have as much control and time to react to imporant events as possible.

Good research design is well balanced, usually incorporates a mixed method approach combining quantitative and qualitative processes and is fully informed by known assumptions and limitations given constraints, targeted outcomes, and informed input by requisite stakeholders.

There will likely be some weaknesses in each of the various steps of your data processes. It’s important to be aware of these from the research design stages in order to identify problems that are likely to arise and/or compensate or adapt to avoid known pitfalls. Sampling bias, technological issues/disparities, missing data, open surveys without forced responses, nonnormal distributions and many other individually small issues can be problematic in the data process when added to others.

It is essential to collect accurate information that is validated and reliable from the appropriate target audience in order to apply the feedback of the analysis to the intended problem or situation. Failure to correctly identify and assess any of these components will undermine the veracity and magnitude of findings.

Non-normal data can be analyzed using non-parametric analyses or other analysis that require less strict assumptions. Additionally sometimes it is beneficial to transform non-normal quantitative data into categorical data which can be normally distributed by using ranges of values for categories.