This is a very common misconception about Predictive Analytics. Many people think they do not have enough data to do Predictive Analytics. But the reality is all companies store data in one way or another. If you have data, you can do Predictive Analytics. It does not matter if you have a Data Warehouse, a Data Base or raw text files. Also, do worry about the amount of data.
Tip #1: quantity is not quality
Even companies with a huge amount of data need to break it into smaller pieces. Small groups of data are easier to be analysed and you can do parallel processing. The key is to know what data are relevant to solve your problem. Selecting the right data is far better then use all data available without a previous analysis.
But what is “enough” data ?
Of couse it depends on your objective, but to illustrate check out some examples:
If you are doing Time Series prediction, which is forecast events in a timeline (for examples, orders, abandoned carts), make sure to input sufficient data to train your Predictive Model. A good starting point is to input at least 50% of the predicted amount of days. If you are willing to predict 30 days, input at least 15 days. If you are willing to predict 30 days, input at least 15 days, and so on. This is not rule. It does not work for everybody. It just a reference.
Tip #2: use entire groups of data for categorization predictions
For categorization predictions, you can use all your available data, but consider key factors like price changes or seasonal periods. Once you have the right subset, the more data, the more accurate the prediction will be. But it does mean you can not do it with few data. It is better to predict even with low accuracy, then take decision randomly. Your prediction precision will get better gradually.