基于滞后主成分回归的电商平台广告效果评价研究

Research on the Advertising Effectiveness Evaluation of E-Commerce Platform Based on the NBD-PCR

基于对广告投放相关变量的细分,考虑负二项分布中成功概率值在处理广告信息累积性方面的优势,提出负二项分布和主成分回归相结合的滞后主成分回归模型,并以天猫平台某口香糖广告投放为研究对象,采集该品牌真实广告展现和点击数据,进行了实例验证。实证结果表明,销售峰值的出现滞后于广告大力投放日,这体现出广告效应的滞后性和累积性。负二项分布对广告变量的滞后处理效果良好,所提出的滞后主成分回归模型能够真实地模拟天猫平台广告效应,并在提取不同渠道广告变量主要信息的基础上,较好地反映出广告投放各维度对销售额的贡献。作为点击计费广告,口香糖直通车和糖果类直通车的投入产出比普遍好于钻展,其中,竞争产品关键词、长尾词直通车的投放对销售额促进效果最为明显。星级店铺贡献率接近10%,不失为是一种很有效的电商广告。

Based on the subsection of variables relative to advertisement launch,considering the advantage of success probability value in negative binomial distribution which can deal with the cumulativity of advertisement information,the lag principal components regression model was presented which combines negative binomial distribution and principal components regression.Also,aiming at advertisement launch of a certain brand chewing,the real advertisement show and hits data were collected and verified.The demonstration results indicate that the appearance of sale peek value lags behind advertising day,which reflects the hysteresis quality and cumulativity;that the negative Binomial Distribution has a good deal with the lag advertisement variables,and that the proposed lag principal components regression model can truly simulate the corresponding Tmall platform advertisement effect.And on the basis of extracting the main information of different channels of advertising variables,it can better reflect the contribution of various dimensions of advertising to sales.As the click charging advertisement,the input-output ratios of gum and candy through chain are generally better than drill show,thereinto,the promoting effect of launch of competitor words and the long tail brand through chain on the sales is the most obvious.The star return of investment is close to 10%,which can yet be regarded as a kind of very effective e-commerce advertisement.