The impact of external word-of-mouth sources on retailer sales of high-involvement products 2014.10.15.

Presentation on theme: "The impact of external word-of-mouth sources on retailer sales of high-involvement products 2014.10.15."— Presentation transcript:

1 The impact of external word-of-mouth sources on retailer sales of high-involvement products

2 Overview Problem Process Scheme

3 Problem Investigate the effect of WOM on sales of different sites(Amazon,JD,Tmall), different type of products(DSLR camera, digital camera, cell phone) Investigate the effect of external WOM on sales of different sites, different type of products Investigate the effect of specific event on sales of different sites, different type of products 1. 电商网站、商品的种类、商品品牌影响均需要分析 2. 除了利用计量经济学中假设检验的方法，还可以利用数据挖掘中的方法进行实验

4 Process Data Preparation
Use data crawler to take product information from JD, Amazon, Tmall, PCOnline and ZOL every day in about 3 months Create the mapping relation of products in the 4 sites and find the their intersection Save the data into database and update the database once a week Python, MongoDB Be integrated into trendata open platform

5 Process Data Preparation
Extract and arrange the data into excel table according to analysis demand The search interest Google trend Baidu trend 1. 百度指数，没有开放API，并且数据采用amf协议（可以利用charles抓包，获得相应flash文件），但还采用了固定秘钥+随机秘钥的方法，估计很困难

6 Process Analysis the correlation among key independent variables
AmazonSalesRank, AmazonRating, AmazonNum, AmazonPrice, AmazonNumOfDaySinceRelease, AmazonHelpfulRating, JDSalesRank…,TmallSalesRank…, PCOnlineRating, PCOnlineNum, ZOLRating, ZOLNum, SearchInterest Variance inflation factor(VIF)

7 Process Analysis using Stata
−𝐿𝑛 𝐽𝐷𝑆𝑎𝑙𝑒𝑠𝑅𝑎𝑛𝑘 𝑖𝑡 = 𝛽 1 𝐿𝑛 𝐽𝐷𝑅𝑎𝑡𝑖𝑛𝑔 𝑖𝑡 + 𝛾 1 𝐿𝑛 𝐽𝐷𝑁𝑢𝑚𝑅𝑒𝑣𝑖𝑒𝑤 𝑖𝑡 + 𝜃 1 𝐿𝑛 𝐽𝐷𝑃𝑟𝑖𝑐𝑒 𝑖𝑡 + 𝜃 2 𝐿𝑛 𝑆𝑒𝑎𝑟𝑐ℎ𝐼𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝑖𝑡 + 𝜃 3𝑖 𝐿𝑛 𝑁𝑢𝑚𝑂𝑓𝐷𝑎𝑦𝑆𝑖𝑛𝑐𝑒𝑅𝑒𝑙𝑒𝑎𝑠𝑒 𝑖𝑡 + 𝜌 𝑖 + 𝜀 𝑖𝑡 (1) −𝐿𝑛 𝐽𝐷𝑆𝑎𝑙𝑒𝑠𝑅𝑎𝑛𝑘 𝑖𝑡 = 𝛽 1 𝐿𝑛 𝐽𝐷𝑅𝑎𝑡𝑖𝑛𝑔 𝑖𝑡 + 𝛽 2 𝐿𝑛 𝑍𝑂𝐿𝑅𝑎𝑡𝑖𝑛𝑔 𝑖𝑡 + 𝛾 1 𝐿𝑛 𝐽𝐷𝑁𝑢𝑚𝑅𝑒𝑣𝑖𝑒𝑤 𝑖𝑡 + 𝛾 2 𝐿𝑛 𝑍𝑂𝐿𝑁𝑢𝑚𝑅𝑒𝑣𝑖𝑒𝑤 𝑖𝑡 + 𝜃 1 𝐿𝑛 𝐽𝐷𝑃𝑟𝑖𝑐𝑒 𝑖𝑡 + 𝜃 2 𝐿𝑛 𝑆𝑒𝑎𝑟𝑐ℎ𝐼𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝑖𝑡 + 𝜃 3𝑖 𝐿𝑛 𝑁𝑢𝑚𝑂𝑓𝐷𝑎𝑦𝑆𝑖𝑛𝑐𝑒𝑅𝑒𝑙𝑒𝑎𝑠𝑒 𝑖𝑡 + 𝜌 𝑖 + 𝜀 𝑖𝑡 (2) −𝐿𝑛 𝐽𝐷𝑆𝑎𝑙𝑒𝑠𝑅𝑎𝑛𝑘 𝑖𝑡 = 𝛽 1 𝐿𝑛 𝐽𝐷𝑅𝑎𝑡𝑖𝑛𝑔 𝑖𝑡 + 𝛽 2 𝐿𝑛 𝑍𝑂𝐿𝑅𝑎𝑡𝑖𝑛𝑔 𝑖𝑡 + 𝛽 3 𝐿𝑛 𝐴𝑚𝑎𝑧𝑜𝑛𝑅𝑎𝑡𝑖𝑛𝑔 𝑖𝑡 + 𝛾 1 𝐿𝑛 𝐽𝐷𝑁𝑢𝑚𝑅𝑒𝑣𝑖𝑒𝑤 𝑖𝑡 + 𝛾 2 𝐿𝑛 𝑍𝑂𝐿𝑁𝑢𝑚𝑅𝑒𝑣𝑖𝑒𝑤 𝑖𝑡 + 𝛾 3 𝐿𝑛 𝐴𝑚𝑎𝑧𝑜𝑛𝑁𝑢𝑚𝑅𝑒𝑣𝑖𝑒𝑤 𝑖𝑡 + 𝜃 1 𝐿𝑛 𝐽𝐷𝑃𝑟𝑖𝑐𝑒 𝑖𝑡 + 𝜃 2 𝐿𝑛 𝑆𝑒𝑎𝑟𝑐ℎ𝐼𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝑖𝑡 + 𝜃 3𝑖 𝐿𝑛 𝑁𝑢𝑚𝑂𝑓𝐷𝑎𝑦𝑆𝑖𝑛𝑐𝑒𝑅𝑒𝑙𝑒𝑎𝑠𝑒 𝑖𝑡 + 𝜌 𝑖 + 𝜀 𝑖𝑡 (3) …… 1. 参考使用不同公式，比如加入评分的方差等

8 Scheme 10/15-10/23 search interest & data preparation
10/24-10/30 analysis on JD 10/31-11/05 analysis on JD & PCOnline & ZOL 11/06-11/13 analysis on JD & Amazon 11/14-11/20 analysis on JD & Amazon & PCOnline & ZOL