Development of a forecasting model for the prediction of product sales using search engine query data

预测交货期和销售量是物流计划中的一大挑战。在大多数情况下，销售方面的波动加剧了物流计划。销售波动可能会因广告或促销活动而出现。这种不确定性可能会导致延迟交货或有关材料的质量问题。其结果是，随着服务水平的降低，后勤绩效会下降。由于有效的销售计划日益重要，许多公司依赖于销售预测的量化方法。这些传统预测方法的缺点尤其是数据的及时性低、细节程度低和预测工作量大。这项研究计划的目的是开发和评估一种创新的产品特定销售预测方法。预测模型应基于互联网搜索引擎查询数据。搜索引擎查询数据包含与消费意图相关的关键字信息(例如，时间、数量)。因此，与现有的预测模型相比，它有望减少预测误差，增加预测范围。这些潜在的搜索引擎查询数据的销售预测应该被分析和揭示。此外，还应产生关于已实现方法的局限性的知识。这项研究为开发基于搜索引擎查询数据的适合行业的预测模型奠定了基础。

Forecasting delivery dates and quantity of sales is a big challenge concerning the logistical planning. In most cases the logistical planning is aggravated by fluctuations concerning sales. Sales fluctuations may arise due to advertising or promotion events. This uncertainty may cause delayed deliveries or quality issues regarding the material. As a consequence the logistical performance decreases such as the service level diminishes. Giving the growing importance of effective sales planning many companies rely on quantitative methods of sales forecasting. The shortcomings of these conventional forecasting approaches are in particular a low timeliness of data, a low level of detail and a high forecasting effort. The objective of this research proposal is to develop and evaluate an innovative approach for product specified sales forecasting. The forecasting model shall be based on internet search engine query data. Search engine query data contains keyword information (e. g. time, quantity) which are related with a consumption intention. Thus, it is expected to decrease forecasting errors and increase the forecast horizon compared to existing forecasting models. These potentials of search engine query data for sales forecasting shall be analyzed and revealed. Also knowledge shall be generated concerning the limitations of the achieved approach. The research forms the basis for the development of industrial-suited forecasting models which are based on search engine query data.