Advice Utilization From Predictive Analytics Tools: The Trend is Your Friend

ABSTRACT

Management decision-making is increasingly supported by new data types and advanced predictive analytics tools. Prior research suggests that the inclusion of new data types – such as social media data – in forecasting models can improve forecasting. We explore whether managers’ operational decisions differ depending on the data type used by a predictive analytics tool and the consistency of the trend with prior developments. Experimental results show that the extent to which managers use predictions from analytics tools is a joint function of the data type utilized and trend consistency. If a trend predicted by an analytics tool reveals a downward break from prior positive developments (i.e., an unexpected negative trend), managers utilize predictions less if they are mainly based on social media data rather than on traditional accounting data. If a trend predicted by an analytics tool continues a prior positive trend, we do not find such a difference. In supplemental analyses, we explore managers’ comfort level and related attitude concerning the data types and find that only in the trend-breaking condition mediation effects are observed. Together, our findings have important implications for the management accounting function that needs to embed knowledge about managers’ information utilization to facilitate decision-making.

Keywords:

• Data type
• trend
• advice-taking
• management accounting
• analytics tool
• social media

Acknowledgements

The authors are grateful for thoughtful comments from participants at the Monash Management Accounting Brown Bag series, University of Queensland research seminar series, University of St. Gallen research seminar series and the Annual Conference for Management Accounting Research (ACMAR) 2021. Specifically, we would like to thank Annemarie Conrath-Hargreaves, Sukari Farrington, Ralph Kober, Matt Peters, Sergeja Slapnicar, Paul Thambar, Xinning Xiao, Lu Yang, Leona Wiegmann, Frank Verbeeten and Nina Schwaiger for their thoughts on earlier stages of this project. We further thank two anonymous reviewers and the guest editors of the special issue for their reviews.

Disclosure statement

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

Notes

1 McKinsey (2011, p. 1) defines big data as ‘datasets whose size is beyond the ability of typical database software tools to capture, store, manage, and analyze’. We adopt this definition, consistent with prior research (e.g., Brown-Liburd et al., 2015; Franks, 2012; Moffitt & Vasarhelyi, 2013; Warren et al., 2015).

2 Predictive analytics tools are frequently embedded within modern Enterprise Resource Planning (ERP) systems which can collect, process, and store a myriad of new data types (Appelbaum et al., 2017; Elbashir et al., 2011). Predictive analytics, which seek to answer the question ‘what will happen?’, are a subcategory of business analytics, which has been defined as ‘the use of data, information technology, statistical analysis, quantitative methods, and mathematical or computer-based models to help managers gain improved insight about their business operations, and make better, fact-based decisions’ (Davenport & Harris, 2017, p. 7).

3 For readability purposes, we occasionally refer to a ‘downward break from prior positive developments’ simply as ‘breaking trend’ or ‘trend-breaking condition’, particularly when discussing our results.

4 We thank an anonymous reviewer for pointing us to the literature on robo-advice and the connection of this emerging literature stream to research on advice and persuasion.

5 Examples include additional steps for cleaning, classifying, and analysing unstructured data. Appelbaum et al. (2017) also point to challenges related to data fluctuations, data duplications, and data security weaknesses.

6 Managers are likely cognisant of the data types predictive analytics tools use and are well-aware that potentially low-quality data likely translates into low quality predictions. This relationship has been described in the slogan ‘garbage in, garbage out’.

7 Original case study: Retrieved January 19, 2021 from https://www.anaplan.com/customers/kathmandu/.

8 Our main results are qualitatively similar, and our inferences do not change, if we analyse each one of the measures of our composite main dependent variable at a time.

9 While we did not hypothesize a main effect of trend, it is interesting to note that the ANOVA results show a significant main effect of trend such that managers discount advice more if the trend is breaking. This result is consistent with expectancy violation theory.

10 For further information, see https://sraf.nd.edu/textual-analysis/resources/#Master%20Dictionary

11 The LIWC program is a research and learning tool. It reads a given text and counts the percentage of words that reflect different emotions, thinking styles, social concerns, and even parts of speech. The LIWC2015 master dictionary is composed of almost 6,400 words, word stems, and selected emoticons. For further information, refer to http://liwc.wpengine.com/

12 We perform a one-tailed t-test as we are interested in examining the use of uncertainty words larger than zero.