The simplest, easiest and spreadsheetble (I just invented this word) information which can be used for predictive and forecasting is defined data over a period. For example, date or time driven patterns which keep repeating till the cows come home. We will not discuss this here and hope you are relieved of some scepticism about this article at this stage.
Now, if you are a large e-commerce site or a banking or credit card company you would have more complex data. And the good news is I am not going to address this complexity as you would have heard more about this then the cows who are returning home. Yes, we all know that the e-commerce sites are so fascinatingly fast that they predict what exactly you are going to look for next. If they are so smart good for you. If they are learning, still good for you.
Ramshankar Sharma is selling buckets in the wholesale plastic market, Kishanlal Pandey a manufacturer of plastic storage items for home use, and Rasiklal Shah is a man committed to selling nuts and bolts. These guys matter as the pillars of Indian GDP, what are they going to do about Predictive Analytics?
Analysts could write 20 pages on this. Someone could write a book. I am going to attempt two paragraphs.
1. Associating the known and accurate data: To run accurate predictive analysis, one would need base data which is accurate and pristine. Your internal data, from your ERP or accounting system, is one such source. The other is officially reported data for manufacturing, petroleum, compiled balance sheets of public companies for an industrial sector and in some cases macroeconomic data. This data is sold by many companies in the business of selling data and is also publicly available at times.
For example, if you know your business is connected to the rainfall (monsoon) before you get into predicting rainfall one should understand the correct relationship between your sales and rains. With your region wide sales data and regional rains connected with timeline, if you can build a proper model to incorporate your stock and team strength along with marketing spends, you have something going. The Met department can be a bit off in predicting the rain, but you will in getting your sales right. Similarly, if you are in the business of selling decorative paints, there will be a relation to connecting it with cement sales. Get this one with other variables right, and you have got past the first step.
2. Once you have the model, you are going to do some very simple things. So simple, that we have a full industry devoted to it. Predict! The software industry and our products are ready for this.
a. Where do I need to focus?
b. Which customer will buy more and who should I focus on?
c. Where do I need to augment my sales team?
d. Which regions will sell more?
e. Which dealers should we focus on?
You have successfully managed this predictive with our gut feeling, your relationships, newspapers and sales reports. Now you can try this with data.
Without attempting to insult your intelligence, I will give an example. For those who have figured out what this is about, you can leave this web page or put your email in the box below if you want me to implement this predictive analysis.
But here is an example how you used to do it with your gut and newspaper. Due to heavy flooding in Assam, there will be more construction. But, people in Assam are not so rich. The government is unlikely to spend. Let us focus on organic growth in Gujarat.
Here is how data will show, sales of cement spiked in Assam 2 months after floods; which drove both your dealer’s performance twice the normal and you did not have a sales person there. And the only reason it was limited was that there was no warehouse there nor was there stock in your Kolkata warehouse.
Predictive analytics will give you a list of dealers with who you the best opportunity to make big sales.
