AUTOMATED Competitive MARKET ANALYSIS TO DETERMINE BUYBACK PRICES FOR A RE-COmmerce firm
exit10 marketing
http://gamesinc.com | http://bootstrapp.com
An e commerce firm which aims to organize pre-owned electronics
sector via listing and reselling old gadgets along (with new ones)
Alexander Souter http://in.linkedin.com/in/alexandersouter
Saptrishi Nath http://www.linkedin.com/in/saptarshinath
WHAT is re commerce
maintaining an inventory of pre-owned products
which is crowd-funded in exit10's case
i.e. people can sell devices to exit10
PROBLEM
Determining value of the product.
But this value keeps changinG
And depends on factors like :
Macro Level
- Local Market
- Global Market
Micro Level
- Charger working ?
- Screen Scratched ?
HOW TO DETERMINE THIS VALUE ?
HOW TO REPEAT THE PROCESS FOR MULTIPLE DEVICE MODELS ?
HOW TO DO IT PERIODICALLY ?
SOLUTION
Index data from other e commerce sites
(just like google indexes the internet)
Analyse product sentiments on social media platforms like twitter.
Use a general algorithm that can determine current value
INDEX OTHER WEBSITES AND STORE RELEVANT DATA IN PROCESS ABLE DATA STRUCTURES
PYTHON URLLIB2
-
Serves as the http agent
-
Retrieves relevant data (to be parsed)
-
Handles Errors and Exceptions
Beautiful soup : XML STYLE parser
XML style : data stored in <open>data</close> form
XML parser : Program to extract data tacking hierarchical nesting
(Initially used regex patterns, but later realized that xml parser was better)
BS is a python library
MYSQL PERSISTENCE layer
AND pymysql SOCKET CONNECTOR
Store parsed data in logical relations.
ex. the price of a product needs to be present with name of product
{8000, "Nokia Asha"}
{32000, "Iphone 4"}
...
PROCESS this data
(PROPRIETARY ALGORITHM)
GIVE PLUG AN PLAY DATA WHICH CAN BE UPLOADED DIRECTLY TO THE WEBSITE.
PYTHON Xlwt, xlrt
Current exit10 architecture needs excel files(for buyback price injection)
XLWT and XLRT : Read/ Write excel files
PROCESS : Get data from mysql, apply algorithm, write results to excel
GIVE RELEVANT STATISTICS
we have more data that we need to determine prices
manually analysed and used to generate product statistics
INput FILE SAMPLE
blackberry z10:case, covers, cover, buy, want to purchase, duplicate, copy, chinese, china, thailand:0.25,0.75,0.8
Sony Xperia Z:case, covers, cover, buy, want to purchase, duplicate, copy, chinese, china, thailand:0.25,0.75,0.8 OUTPUT FILE SAMPLE
Blue fields marked 'p' are proprietary to the company
DEMO SCREENCAST
Other Branch of the project
Twitter sentiment analysis
The software is capable of automatically determining a products sentiment's but needs a plug in layer as companies algorithm didn't have provisions for this