Field Experiments on the Barriers Firms Face in Realizing Gains From Data
Access Status
Full text of the requested work is not available in DASH at this time ("dark deposit"). For more information on dark deposits, see our FAQ.Author
Kim, Hyunjin
Metadata
Show full item recordCitation
Kim, Hyunjin. 2020. Field Experiments on the Barriers Firms Face in Realizing Gains From Data. Doctoral dissertation, Harvard Business School.Abstract
We are living in an age of unprecedented amounts of data. As information becomes more available and accessible, it provides new opportunities to evaluate and extend theories on how firms should – and how they do – use data to inform their strategy. It also raises a central question: as firms compete in an increasingly data-driven landscape, what enables (or hinders) them from realizing potential gains from data?This dissertation explores different ways in which firms use information and provides insights on three barriers they may face in realizing gains from data: managerial inattention that impedes awareness of even easily accessible competitor data, strong priors combined with weak incentives that lead to non-compliance with algorithmic recommendations even when they improve decisions, and multiple goals that hamper how employees process and learn from information.
Chapter 1 studies how firms use easily accessible competitor information, through a field experiment run in collaboration with Yelp across 3,218 personal care businesses. I find that nearly half of the firms lack knowledge of their competitors’ pricing, a key strategic lever in this industry, even though this data is readily accessible and enables them to improve their own decisions once they are made aware. I find evidence consistent with the interpretation that this lack of awareness is driven by managerial inattention fueled by reliance on outdated knowledge. As competitor data becomes increasingly accessible in the digital age, these findings highlight the role that competitor awareness may play in how firms make decisions, and suggest that overcoming attentional barriers may be a key factor that enables firms to realize gains from data.
Chapter 2 evaluates whether algorithms improve decision-making by partnering with Boston’s inspectional services department to compare different methods to prioritize restaurants to inspect: inspector discretion versus two algorithmic methods with different levels of sophistication. While gains from using algorithms are substantial, the greatest gains stem from using data to supplement inspectors’ priors rather than algorithmic sophistication. Yet despite these gains, inspectors are only half as likely to inspect restaurants based on algorithmic recommendations compared to those based on their own judgment. These findings suggest that incentives to ensure compliance may be more important than algorithmic sophistication, and that if algorithms are to be effective, mechanisms must be put in place to ensure implementers trust them more than their private information.
Chapter 3 explores how the communication of multiple goals impacts employee performance and learning from information on best practices. I design and implement a field experiment across linemen in a large multinational energy company, varying whether employees are communicated a single goal (safety alone) or multiple goals (safety and efficiency) to pursue as they are evaluated on a core operational procedure. I find that although the two goals are unlikely to have a natural positive complementarity, employees show little evidence of tradeoffs, improving efficiency without reducing safety. I find suggestive evidence that this may be explained by employees being inside the productivity frontier. However, despite this absence of direct tradeoffs, communicating multiple goals appears to impede how employees process and learn from information on best practices, resulting in worse outcomes across both safety and efficiency. These findings suggest that there may be cases when communicating multiple goals leads to larger performance gains than focusing, and that when choosing to pursue multiple goals, organizations may need to consider whether employees are inside the frontier and whether the task involves learning beyond effort allocation.
Taken together, these chapters suggest that despite the potential of data and algorithms to improve decision-making and inform strategy, their returns may be limited if organizations are not redesigned to make use of them.
Terms of Use
This article is made available under the terms and conditions applicable to Other Posted Material, as set forth at http://nrs.harvard.edu/urn-3:HUL.InstRepos:dash.current.terms-of-use#LAACitable link to this page
https://nrs.harvard.edu/URN-3:HUL.INSTREPOS:37364450
Collections
Contact administrator regarding this item (to report mistakes or request changes)