Can Technology-Enabled Advanced Monitoring Systems Influence Individual Performance and Team Dynamics?

Abstract

Recent advancements in technology have resulted in more sophisticated employee monitoring systems. We study the impact of a technology-enabled advanced monitoring system deployed by the National Basketball Association and explore whether it serves a decision-facilitating role and/or a decision-influencing role at the individual-level. We find that the system impacts players’ playing strategy (i.e. a decision-facilitating role), but not their effort (i.e. a decision-influencing role). The impact on playing strategy increases average individual performance across the entire sample. We also find that the effect is somewhat greater for lower ability individuals in the first year of implementation. At the team-level, we find that system adoption is associated with a smaller dispersion of effort and a greater degree of specialization among individuals in a team. Our study has implications for the use of technology-enabled advanced monitoring systems in a range of industries.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 This research benefitted significantly from feedback from seminar participants at Universidad Autonoma de Madrid, Gavin Cassar, Patrick Ferguson, Matthias Mahlendorf (editor of the special issue) and two anonymous reviewers. This project was supported by the Kinsman Grant from the Faculty of Business and Economics at the University of Melbourne. Research assistance provided by Antonius Jaeger, Yile (Anson) Jiang and Laure Pavaday is much appreciated.

2 Unstructured data is data without a readily identifiable format or structure (Syed et al., 2013).

3 The report by MicroStrategy is based on a survey of 500 enterprise analytics and business intelligence professionals from around the world about the state of their organizations’ analytics initiatives.

4 However, at low levels of monitoring, Dickinson and Villeval (2008) find that monitoring actually increases effort due to its disciplining effects.

5 The advanced monitoring system evaluates the quality of assists by calculating the shooting percentage (or accuracy of shots attempted) of the player’s teammates following a pass from that player. High shooting accuracy results in relatively easier shots for the receiving teammate, indicating that the player’s passes are of high quality (McCann, 2012).

6 It is important to note that the traditional basketball statistics (i.e., those available before advanced monitoring was in place), which are a major source of the data for this study, would continue to be recorded by officials while watching the game at the stadium. Traditional basketball statistics include points scored, offensive and defensive rebounds, assists, blocks, steals, free throws, points made, and 3-point field goals made.

7 Other new statistics available include catch and shoot points (scoring a jump shot outside of ten feet where a player possessed the ball for two seconds or less and took no dribbles), pull up shot points (i.e., scoring a jump shot outside of ten feet where a player took one or more dribbles before shooting), and secondary assists (i.e., passing to a player who then records an assist within one second and without dribbling).

8 See www.stats.nba.com. Under ‘Players>Tracking’, there are 14 different categories of player tracking data including categories such as ‘speed and distance’, ‘shooting efficiency’ and ‘defensive impact’.

9 This timeframe captures four full seasons, namely, the 2011-12, 2012-13, 2013-14, and 2014–15 seasons.

10 To conduct our statistical tests, we need changes in strategy (and changes in effort) to be captured by traditional statistics that are available both pre- and post-advanced monitoring.

11 The restricted area is the part of the basketball court that is within a four feet radius from the centre of the basket. It is called the restricted area because it marks the minimum distance from the basket that a defender must be positioned in order to draw a foul from the offensive player who is trying to get to the basket to score.

12 We have provided a rather simplified example of how is calculated here. The actual measure distinguishes the areas of the court from which the shot is taken in a more granular manner by segmenting the offensive court into seven areas: left corner 3-pointer, right-corner 3-pointer, above the break 3-pointer, backcourt, mid-range, in the paint, and in the restricted area.

13 Note that both our measures of change in strategy and change in effort are based on traditional statistics that were available both before and after the adoption of advanced monitoring. However, we expect that these measures will change (relative to the base year) to a greater extent once advanced monitoring is in place. For example, following the availability of advanced monitoring statistics, by covering a greater distance on the court at greater speeds, a player should be able to make more offensive rebounds, blocks, and steals (i.e., our measure of effort). Similarly, by having more information on offence, such as the types of plays that would increase the likelihood of scoring, players can adjust the distinct areas on the basketball court from which they take shots (i.e., our measure of strategy - ΔDF).

14 3-point shots are those that are made from at least 22–23.75 feet from the basket.

15 Because there are 60 draft picks each year, players who were undrafted were assigned a value of 61.

16 Our results are qualitatively unchanged if we dropped the one item (percentage change in BLOCKS) that failed to load onto ΔEFFORT.

17 To investigate whether any remaining time trend in the data impacts our main findings, we conduct 15 rolling OLS regressions of the models in Table 3 starting from season 1998–99 until 2000–2001 (baseline: 1997-98), right up to the period used in the study, i.e., season 2012–13 until 2014–15 (baseline: 2011-12). We do not control for CONTRACT (i.e., average years to expiry of players’ contract) as we were not able to collect these data for previous time periods. We also run pooled regressions where we combine all the data from the individual rolling regressions. The results indicate that while changes in a player’s shot selection from the constant baseline season generally increase over time, the presence of advanced monitoring is associated with an incremental change in shot selection (relative to the constant base year) albeit at lower significance levels (p<0.05) compared to the AM coefficients reported in Table 3 (Model 1). We also find that shot selection only has a significant positive association with performance in the regression for the baseline season is 2011-12, i.e., when advanced monitoring is in place for two of the three seasons analyzed. The coefficient for shot selection is not significant in any of the other 14 rolling regressions. Two separate pooled regression analyses further confirm this finding and provide compelling evidence that is consistent with the argument that a change in shot selection brings about a positive change in a player’s performance (i.e., points scored per unit of time) only when it is aided/triggered by the advanced monitoring system.

18 As an alternative to structural equation modeling, we also ran OLS regressions of the models in Table 3 (with clustered standard errors by team). The results of this analysis are qualitatively similar to what we find in our main analysis using structural equation modeling.

19 The trend of shooting more 3-pointers relative to 2-pointers could increase the total points scored because 3-point shots offer 50% more points than 2-point shots while the chances of scoring on a 3-point shot is on average only between 15–20 percentage points lower than a 2-point shot (Mather, 2016).

20 This is because there is no corresponding decision-facilitating measure available at the team-level. The decision-facilitating (DF) measure at the player-level takes into account the change in the percentage of shots taken in seven distinct areas of the court. However, as far as we know, these data are not available at the team-level.

21 We estimate Models 4 and 5 without average team effort and average number of shots attempted and find that the inferences from these models (untabulated) are unchanged.

22 The model fit is good: root mean square error of approximation (RMSEA) is 0.02; comparative fit index is 0.97; and standardized root mean squared residual is 0.02.