ERP, Machine Learning and demand forecasting

ERP,  Machine Learning and demand forecasting

I am going to speak on the impact of artificial intelligence and ERP As ERP practitioner I have dealt with multiple projects for multiple clients in multiple geographies including us Europe Middle East and India. I find that ERP packages are excellent for maintaining the record in a systematic manner and have almost all the feature which will require any Enterprise to work in a standard method but then the piece which is it is missing is on actionable suggestions managing a large on structure or unstructured data this is by design as we wanted to keep ERP very focused on basic business processes.

let’s take a simple example of now if you want to do generate demand forecasting or you want to calculate the number of vistas which will which will come to your site and which who will  buy and based on that you can do supply chain planning or manufacturing plan MRP planning put requisite resources whether its human manpower or capital investment 2 forecast the demand eventually leading to better sales or more revenue or  more profit. another example could be as no you can have access to all kind of data of your prospect and you can collect the data analyse the data and process the data as per your business plan. For example Amazon collect the detail of their customer check which kind of product they are looking for and not only provide that project that product but also provide suggestions or  substitution, They have gone further and you are providing suggestions to customer that if they buy some product or they have bought some product what else can complement product.

For example, if you are buying  DSLR camera and you have chosen Nikon then you will have a suggestion that you should buy a case of the camera, additional lenses, tripod et cetera.   the suggested product will have multiple options and based on the parameter such as your type of camera you have bought they will make an intelligent suggestion to you. if you look at the model of the Amazon they run a complex a logarithm to make the suggestion which involves various parameters. search price, seller ranking,  product description, the sale of the product which is the historical data and so and so if you look at it the external factors determines the buying patterns of an individual. So you can visualise why companies like Amazon have become super successful in a very short span of time.Whereas traditional companies are struggling to stay afloat. The amalgamation of Technology and customer have tried to remove brick and mortar stores.

So if you look at it the data is available and and and experience resource can use the data to protect the eventual sale or outcome but here is the issue to accurately do this prediction any person needs to do a plethora of calculation We can do the gut level calculation.  The typical the gut level calculation involves sales organisation from sales director to sales executive, where is sales executive are the people on the ground they collect demand from the distributors or primary customers. They can sense what is going to sell, at what price it can sell,  what competitors are doing, what is the Social factor which will influence sell. the sales manager can take the input from sales executive to generate expected sales figure for the year.

The organisation has different means they expect to grow by 30% but a projection of sales team amount to less than that percentage.  so to increase the sale stimuli is required whether it is in term of sales promotion or discount et cetera.

If we look at there are various outcome based problem,  which involves multiple stakeholder so ml helps you determine the outcome based on various parameters or we can say problem statements.  

The limitation of ML

  • Problem statement should be clear.
  • It requires good quality of data,  which should correspond to problem statement. the quality of data if not available then it can create an issue of data bias or overfitting of data. In layman’s term, Pisces or something like guesses which we tend to do and we learn it on the accuracy of  our guess.

So in future when when prediction is made the errors of our guesses are taken into account,   so each sales executive projection is validated by their managers. So what we are looking at the breakdown of each activity which determines the final outcome.  all these activities either should have the parameter or processes. Interestingly some of these activities are you and beyond organisation, some will determine by the market,  competitors any other local factors.

Now  coming back to Initial point ERP systems do not capture all of these data , the data which is stored in ERP table pertains to fact ( actual sales or revenue)   The supervised learning combined with linear equation / deep learning will provide result which will be much higher accuracy and will help organization to streamline their business process to very level

This is one of the main reason why we see the stupendous appetite for ml in the enterprise today.


Leave a Reply

Your email address will not be published. Required fields are marked *