What will be the best method to carry spend analysis for a company with 1 Million parts and orders are getting created using free text.
You will need some clever formulas, not an impossible task, but you will want to create a list of searchable words by searching the cells for the words they contain, then normalise that list to remove duplicates and codify into categories before you start cleansing nonsense data.
Alternatively you can pay for someone to do this for you, I would recommend https://www.theclassificationguru.com/
Seems like a challenge, free text is always hard. Is it English then you can works with NLP, stemming, text mining using some libraries, work with Levenstein distances. To build a supervised machine learning model you will also need a training set.
Power BI is a powerful tool to model and serve.
We developed a machine learning tool that automatically cleans and categorises spend data in seconds/minutes (SPAT) to a high degreee of accuracy (rubbish data in - accurate data out). It works off invoice data, various fields, but primarily the free text data. We use MS PowerBI to present the data in an interactive dashboard. We have successfully used this across a range of large organisations
Most organisations use free text for POs - to automate this is very complex and hard to maintain an exhaustive list (especially if you alter service/good suppliers regularly). Suggest good master and sub categorisation of POs at point of order. However, this needs resource to periodic review and cleanse as the users will ignorantly or lazily use the incorrect categories. Any cleansing needs to take place in co-operation with Finance to avoid any budgetary misalignments. The way to 'sell' this to the business is that the more in control of spend and trend analysis then greater intelligence, savings and innovation can be identified and delivered with the business benefits far outweighing the resource costs.