#	Pagina
attuale pagina	/open-h2020/projects/207028/results.html

Report

Teaser, summary, work performed and final results

Periodic Reporting for period 1 - EW-Shopp (EW-Shopp - Supporting Event and Weather-based Data Analytics and Marketing along the Shopper Journey)

Teaser

Summary

EW-Shopp aims at supporting companies to gain customer insights and build innovative data-driven services for their customers. Such objectives can be achieved through an increased capacity of factoring events and weather information into analyses performed on companiesâ€™ data. The resulting improved capacity of building data-driven services will enable the European business ecosystem to increase their efficiency and, ultimately, regain lost positions in competing against global internet service giants that over the last decade made intensive exploitation of integrated big data using exclusive and proprietary solutions. To achieve this objective, the main technical challenge for the project is to deliver the EW-Shopp platform, a solution aiming at easing the integration of multi-lingual customer and market-related data, and enriching them with weather and event data. In this way, the platform will also support a new generation of affordable yet effective business intelligence solutions (analytics and visualization), which are at the moment too expensive or require expertise in specialized sectors to be adopted by the majority of European SMEs. The impact of the project will be assessed via different business cases in the eCommerce, retail and marketing domains, where marketable data-driven services will be developed and tested.

Work performed

On the business side, the project has focused on the design and development of pilot services that harness weather and event-based analytics leveraging data consumed by third parties. The work has been conducted within three business cases: one in the domains of eCommerce, Retail and Customer Relationship Management (CRM), one in Location Intelligence, and one in the domain of Digital Marketing. Five different pilot services have been developed. 1) A weather-aware widget service for the enrichment of purchase information for customers of eCommerce platforms. 2) An integrated software ecosystem for category and marketing optimisation that uses weather and event-based sales predictions. 3) data access API and decision-making system to optimise resource allocation in CRM contact centers using weather and event-based predictions of calls/successful calls. 4) Scout+, a dashboard for location intelligence based on weather and seasonality trends. 5) A service for the weather-aware launch of digital marketing campaigns. Business and technical partners have cooperated to integrate the data, formulate analytical hypothesis, test them, and build predictive models. Finally, business partners have encapsulated the predictive models into software services.

On the technical side, time and effort was dedicated to the following activities. 1) Designing the architecture of a data enrichment and analytics platform as a set of interoperable components, accessible as a data-as-a-service solution, as well as installed on the companyâ€™s premises. In particular, the data transformation process required to support the analytics can be devised by a data expert working on a reduced sample of the original data through a responsive GUI, and then run on big data in batch mode. 2) Improving key components that are needed to support the data enrichment, testing analytics and visualization using the analytical components of the platform, and making them mutually interoperable. The components have been tested using data enrichment workflows and analytical models used to build the pilot services. 3) Making data available in such a way that data enrichment and analytics operations can be performed more easily and efficiently with the platform. In particular, formats to support interoperability for spatial temporal and product data have been specified.

Final results

At the business level, progress have been driven by the five pilot services. The first service is a widget aimed at increasing customer awareness in the purchase process and eventually converting traffic. The second service, the integrated software ecosystem for category and marketing optimisation, helps category managers to better allocate resources based on sales predictions that use weather and internal events. Results suggest that weather and internal events are important to achieve reliable predictions. In the second part of the project, focus will be given to event-based analyses and to scale the services to a larger number of categories. The third service use predictions of the number of calls/successful calls in order to optimize the number of agents and maximize the efficiency. A preliminary evaluation has found that weather-based analytics can help predict calls from/to contact centres. In the second period, will focus more on event-based analyses using general and marketing events. In the fourth service, a dashboard is introduced, which is offered to Point of Sales managers to inspect the results of weather and event-fed predictions of visitors. Clients of the service provided positive feedback. In the second period we will include in the model also events and extend the service to help businesses allocate their advertising budget based on historical analysis of customer flows vs seasonality and weather trends. The fifth service establishes the best moment to launch a digital marketing campaign predicting its performance as a function of the weather. In the second period we will improve the accuracy and scalability of the predictive models by identifying keywords that are sensitive to weather and events.

EW-Shopp has integrated Grafterizer and ASIA to support data cleaning and data enrichment workflows designed with a user interface and replicated at large scale. Data enrichment is implemented by using semantic technologies to reconcile corporate data with shared systems of identifiers, and then fetch data from sources that use the same identifiers. ASIA has been extended with data reconciliation and extension functionalities, supported by different 5 reconciliation services and 3 extension services. Grafterizer has been extended with the capability of deploying data processing workflows as a self-contained computing services over a managed cluster of resources and with a hosting backend based on a multi-model database. In the second period of the project, EW-Shopp will focus on the refining and performance optimization of both reconciliation and extension functionatilites when dealing with genuine big data in batch mode.

APIs have been created to ease the access of complex data like weather forecasts produced by the ECMWF performing aggregation on time or location. To publish the GfK product catalogs in RDF as linked data. Moreover, to improve the management of product data for SMEs, we have created two tools based on machine learning: a tool that classifies products into a hierarchical taxonomy based on their descriptions; a deduplication tool that identifies different descriptions of the same product and supports their reconciliation. We have also created RDF version of Google GeoTargets and started to reconcile them with GeoNames. This reconciliation process will continue in the second period of the project.

Finally, we have built a set of tools on top of the QMiner platform for streamlined the development of predictive models on business, weather and event data. In the second period of the project, we will t fine-tune these tools, ensure their scalability and improve their integration with the platform.