Limitations of Predictive Analytics

why don't we use big data all the time?

Incorrect data:

Most of big data is unstructured and unstructured data is difficult to make sense of. It is very likely the system will not be complex enough to understand it fully. It may be interpreted wrongly resulting in nonsensical output or it may contain incorrectly spelled words or grammatically incorrect sentences which we cannot use easily.

Incorrect/missing out data:

Another problem is that one person out of the data set may have recorded information incorrectly and so it is not a true data set therefore making analysis of the incorrect data meaningless for real life.  Data produced from people can be biased, this leads to inaccurate predictions. People may have been paid to give out certain information which can be incorrect e.g. sponsored content is often ingenuine and this can reduce the effectiveness of the predictions as we can't tell if the opinions were true or not.

***************************************************************************
Correct data:

Another thing that affects big data analytics is that an unknown third variable may come into why the data is the way it is. If we have a graph with two variables we are assuming that only those two variables will be affecting the result when in real life there are often multiple variables affecting an outcome. Predictive analytics cannot magically decide what the third variable is without knowing. There is almost always unexpected things we would not be able to predict. A major limitation of predictive analytics is that data cannot predict human feelings - human feelings often lead to the outcome. Human emotions can also lead to data being taken out of context which affects the reliability of the data and therefore can lead to inaccurate predictions.