How to Improve Forecasting

Shoemaker and Tetlock (2016) offer suggestions on how to better a company’s forecasting. They argue that companies and individuals always get it wrong when predicting the occurrence of future events. Shoemaker and Tetlock (2016) conducted a quantitative research that involved 25,000 forecasters and a million forecasts and came up with three practices, which if well applied, can catapult organizational predictive capabilities. First, company employees are trained for good judgment on the basics of statistics and biases. Employees are trained in basic statistics, taught to avoid cognitive biases, and on psychological factors that lead to biases. They are also given confidence quizzes to determine if they are aware of their deficiencies in certain areas. The authors advise companies to tailor training programs to their circumstances because there is no one blueprint that can fit all companies. The second practice that can be applied to improve forecasting is building the right kinds of teams which must be well-managed and possess exceptional skills to succeed. These teams should comprise people who esteem data, have the ability to forecast without a struggle, are alert to bias and good in sound reasoning. These teams should have both the experts and non-experts and learn to trust one another. The third practice concerns performance and feedback. It’s important that companies get to know how they fare in predictions, as well as track the process (Shoemaker & Tetlock, 2016).

Hernandez (2017) observes that predicting the future is part of the job yet companies’   approaches seem stuck in the past. He is of the opinion that company forecasting fails to include measurable unrecorded outcomes making it very difficult to gauge one’s position. Hernandez (2017) shares insights on how their company learned to improve on forecasting. Quoting Tetlock, he borrows the insights of a ‘growth mindset’ or a ‘willingness to learn from past mistakes’ (Hernandez, 2017) where he expounds on how they applied this principle to Twitch, a subsidiary of Amazon. The gist of it is that it starts with an individual. Once an individual gains prediction precision, then the company does too (Hernandez, 2017). Thus, Twitch designed a program that encouraged all the staff to become better forecasters irrespective of their background in quantitative research, role in the company or expertise. Hernandez (2017) used a product called host mode to better his prediction capabilities. First, he put together supporting data to make the prediction believable. Next, a team was assigned the task of predicting and each of the members made predictions. Instead of supplying accurate figures, employees were encouraged to give their 80% confidence interval. Immediate feedback was then provided which assisted employees to calibrate their assessments after which they were asked to make predictions concerning their work. Hernandez (2017) also identified challenges with forecasting such as doubt that forecasting really works and that predictions could turn out to be inaccurate; staff’ fear that they did not have what it took to forecast, or that they could have their predictions misinterpreted by colleagues, which would impact negatively on them. In addition, they believed that there was no evidence of forecasting (Hernandez, 2017).

The study by Fildes and Goodwin (2007) focused on management judgment. They argued that companies still lagged behind in forecasting despite the release of these eleven principles: prefer quantitative to the qualitative; have little-prejudiced alterations of quantitative predictions; adjust for future events; insist that experts write down their forecasts; use prearranged procedures to incorporate judgmental and quantitative methods; put together forecasts from different approaches; begin with equal weights when combining forecasts; compare the previous performance of different forecasting techniques, and look for feedback about forecasts. The authors concluded that companies relied heavily on amorphous judgment and unsatisfactorily on statistical methods leading to vague forecasting.

Hofer, Eisl, and Mayr (2015) conducted a study that compared forecasting behavior between SME’s and large firms in an explosive business environment in Austria. Their findings show that SMEs, due to limited resources, bear the brunt of this explosive business environment compared to large firms, leading to less accurate predictions. On the issue of methodology, they note that large firms tend to utilize qualitative methods more than the small firms, but that both groups need not utilize intricate quantitative methods such as linear regression. Hofer, Eisl, and Mayr (2015) are of the opinion that quantitative and qualitative methods should be combined in order to ensure accuracy in forecasting because there will be less bias. Further, they conclude that Austrian companies should step up the building of technical know-how and adopt a holistic, flexible process which will effectively adapt to the ever-changing volatile conditions (Hofer, Eisl, & Mayr, 2015).

Relationship to the Course

This course is about quantitative methods in business research in Human Resource. It should be noted that forecasting is a function of Human Resource in some companies, the department tasked with the responsibility to position the company globally in terms of product output. In some companies, however, this function is housed in the operations department. Nevertheless, it should be noted that a company’s growth is meticulously planned and one of the tools used is forecasting, which, however, cannot occur without research. The most used approach to forecasting is quantitative research.

Shoemaker and Tetlock (2016) argue that with straightforward predictions, there is no need for subjective judgment because the data will speak for itself where concerns can be predicted with great accuracy using econometric and operations-research tools. But then there are also issues that are multi-faceted, managed poorly and hard to quantify, therefore, the authors call for use of data, logic, and analysis alongside seasoned judgment and careful questioning. Quantitative research in the Good Judgment Project was used which sought a million predictions from 25, 000 respondents in a bid to determine whether some people were naturally good at prediction than others (Shoemaker & Tetlock, 2016).

On his part, Hernandez (2017), is of the view that businesses fail because they fail to include unrecorded quantifiable outcomes, making it hard to track their progress. The author’s arguments also underscore the value of quantitative research in forecasting. For instance, he says that numerical predictions provide an array of benefits for large organizations. He finds these predictions accurate, succinct, and simple to communicate across work teams. In addition, if predictions would be based on mathematical probabilities, then employees would be forced to measure their own reservations about forecasting. Various forecasts can be summed up and averaged thus helping administrators to know what employees are thinking. The use of data-based predictions elucidates decisions, better communicates concerns and aids in employee motivation. In addition, Hernandez (2017) posits that the training mode involves workers predicting using confidence intervals rather than numbers, another statistical function.

Fildes and Goodwin (2007) dwell on management judgment but recommend the quantitative approach to make predictions. They sought to find out if managers used judgment based on the principles of forecasting earlier released. They found out that organizations would be better placed reducing judgmental adjustments of quantitative forecasts. Managers prefer their own judgments (which is prone to subjectivity) to judgments borne from statistical findings. To come up with these findings, they themselves used quantitative means when they surveyed 149 forecasters (Fildes & Goodwin, 2007).

An interesting perspective on methodology is offered by Hofer, Eisl, and Mayr(2015) who sought to compare the forecast behavior of small and large Austrian firms. The concluded that the larger firms were successful in forecasting because they depended on qualitative methods as opposed to the small firms which relied on uncomplicated, measurable and subjective techniques such as projecting historical data.

Application to Work Situation

I work in the Human Resource (HR) Department of a growing hotel with a 300-bed occupancy in a small town. We have never attained 100% bed occupancy yet this is our vision. The best we have ever done is 92%.  Forecasting is done but not as highly structured as would be desirable. Shoemaker and Tetlock(2016) focus on teams for successful forecasting. Indeed we work in teams, the only difference being that not all employees have been trained in statistics and data management as individuals. However, there were positive changes when management decided to introduce prediction in all departments in early 2016. Previously, the department that deals with statistics was a joint sales and accounting team who made quantitative predictions and passed them on to the HR team to debate and implement with the help of management.

The only applicable thing that our company is implementing from Hernandez’s (2017) paper is the issue of recording measurable outcomes for the purpose of feedback and improvement. Every department has a chart where daily sales and projections are recorded. For instance, who was directly responsible for finding a client? These are then analyzed monthly at a departmental meeting. Since this started, we have seen an increase in the number of guests in our hotels, at a time when bed occupancy was low. As a result of using this strategy, the marketing team was able to effectively make good use of social media to increase sales after recording their outcomes for a period of 3 months.

Previously, we relied purely on management judgment as opined by Fildes and Goodwin (2007). However, this trend shifted two years ago when management policy allowed for the inclusion of stakeholder opinion. For instance, in a meeting to discuss food, a chef, and not necessarily the head chef, must be present. It was during such meetings that the use of statistical methods began to be applied. Forecasting improved with the result that from a previous 35% bed occupancy during the low season, we have managed to climb up to 60% after intensifying our campaign for local tourism based on statistical findings (Fildes and Goodwin, 2007).

Finally, our company, according to Hofer, Eisl, and Mayr (2015), could be considered an SME. As stated above, ours was a qualitative approach before we incorporated statistics into our forecasting. We now are implementing the researchers’ recommendations by using both qualitative and quantitative approaches.

References

Fildes, R., & Goodwin, P. (2007). Against better judgment? How organizations can improve      their use of management judging in forecasting. Interfaces, 37(6), 570-576. Retrieved from http://www.jstor.org/stable/20141547

Hernandez, D. (2017). Decision: how our company learned to make better predictions about everything. Harvard Business School.

Hofer, P., Eisl, C., & Mayr, A. (2015). Forecasting in Austrian companies. Journal of Applied Accounting Research, 16(3), 359-382. Retrieved from https://search.proquest.com/docview/1732326177?accountid=45049

Shoemaker, P. J. H., & Tetlock, P. E. (2016). Superforecasting: how to upgrade your company’s judgment. Harvard Business Review.