The Impact of Big Data on Firm Performance: An Empirical Investigation

Abstract

1. Impact of Data Size on Quality of Forecasting in a Theoretical Model Retail companies use proprietary forecasting models, which are often the result of a complicated engineering effort, tailored to the particularities of the business. Theoretical properties of such complicated forecasting models may never be understood completely. Hence we give a theoretical benchmark based on a well-known, state-of-the-art model used in academia to model panel time series and forecasting. In particular, we utilize the Augmented Factor model; see, for example, Bai and Ng (2002), Bernanke et al (2004), and Bai (2009). Within this fairly general, yet tractable model, we ask the question of how forecast errors would be determined and what role do the data size, the number of products and number of time periods in the available history, have.The quantity , sold by a retailer of a product at time , obeys the following equation:where , ≥ 1 is a known velocity variable, which describes the stochastic, possibly nonstationary level of the series, and , ≥ 1 is a reference demand level, where , stationary for each . Note that , = 0 if and only if , = 1 and , = 1. The quantity , (plus 1) is determined by the base demand times the multiplier , that captures the ``size of the firm" in product sales as well as "velocity" of the product. Velocity , reflects the notional popularity of the product at time , and represents the product-specific size of the retailer, as specific to the product. A simple example of , is given by the lagged sales , = ( , + 1). We can normalize the velocity at , =1 for = 1. The vector of velocities , characterizes the overall ``size of the retailer" over time. In the model, the base quantity index , is determined by latent factor components plus observed components plus stochastic shocks: (i) Time-varying factors are latent time-varying common factors, such as macro-economic factors, seasonality, and fashion factors. (ii) The product varying factors are the product-specific loadings on the latent factors , to which product demand responds differently. They include the conventional fixed effects model, when = 1. (iii) The observed component is determined by adimensional vector , of observed time-varying product features, such as prices of the product

A Holistic Overview of Anticipatory Learning for the Internet of Moving Things: Research Challenges and Opportunities

“…They also have the capability of computing by using a wide range of communication interfaces, such as Wi-Fi, Bluetooth, or NFC. The ability to sense themselves and their surrounding environments is key to generating "small data streams" over space and time in such a way that they share many characteristics of big data, including the five V's: variety, velocity, volume, veracity, and value [37][38][39][40][41].…”

Artificial Intelligence in Society (Summary in German)

“…9 As a result of these concerns, many OECD countries report legislative barriers to the use of personal health data. Health Data Governance is an important step towards a more coherent approach in health data management and use (OECD, 2016 [71]). It aims primarily to promote the establishment and implementation of a national health data governance framework.…”

“…[39] OECD (2016), Recommendation of the Council on Health Data Governance, OECD, Paris, https://legalinstruments.oecd.org/en/instruments/OECD-LEGAL-0433. [71] OECD (2015), Data-Driven Innovation: Big Data for Growth and Well-Being, OECD Publishing, Paris, http://dx.doi.org/10.1787/9789264229358-en. [62] OECD (2014), "Skills and Jobs in the Internet Economy", OECD Digital Economy Papers, No.…”

“…First, some activities previously carried out by people will be automated. Second, through machine autonomy, systems will operate and adapt to circumstances with reduced or no human control (OECD, 2017 [68]; Autor and Salomons, 2018 [71]). Research on 12 developed economies estimated that AI could increase labour productivity by up to 40% by 2035 compared to expected baseline levels (Purdy and Daugherty, 2016 [72]).…”

The Effect of Big Data on Recommendation Quality. The Example of Internet Search

“…Observing more products in a specific product category, however, has almost no effect on forecast accuracy. Similarly to our approach, Bajari et al (2018) account for non-data related improvements in forecast accuracy such as improved technology by controlling for elapsed time. As opposed to our data set, the data used by the authors do not allow to study the accumulation of different amount of data over the same time span, which is a crucial step in our identification strategy.…”