Forecasting intermittent demand for inventory management by retailers: A new approach

The forecasting of intermittent demand is a complex task owing to demand fluctuations and interval uncertainty. Intermittent demand is essentially random demand with a high percentage of zero values. In the retail industry, there are many products which face intermittent demand and this poses a problem of inventory management. This study proposes a Markov-combined method (MCM) for forecasting intermittent demand, which takes into account the inventory status and historical sales of products. We divide the prediction process into two stages. In the first stage, the transition probabilities of the four basic states of demand and inventory are calculated. In the second stage, the corresponding and appropriate prediction method is selected according to the predicted state. Further, using two large datasets from the two biggest e-commerce companies in China, we verify our results and show that the MCM forecasts more accurately than the Single Exponential Smoothing (SES), Syntetos-Boylan Approximation (SBA), and Croston (CR) methods. The MCM can be as an alternative method for forecasting intermittent demand because it is easy to compute and typically more accurate than the classical forecasting methods.     

Joint In-Season and Out-of-Season Promotion Demand Forecasting in a Retail Environment

Inaccurate forecasts of demand during promotions diminish the already meager profit margins of retailers. No forecasting method described in the literature can accurately account for the combination of seasonal sales variations and promotion-induced sales peaks over forecasting horizons of several weeks or months. We address this research gap by developing a forecasting method for seasonal, frequently promoted products that generates accurate predictions, can handle a large number of sales series, and requires minimal training data. In our method's first stage, we forecast the seasonal sales cycle by fitting a harmonic regression model to a decomposed training set, which excludes promotional and holiday sales, and then extrapolate that model to a testing set. In the second stage, we integrate the resulting seasonal forecast into a multiplicative demand function that accounts for consumer stockpiling and captures promotional and holiday sales uplifts. The final model is then fitted using ridge regression. We use sales data from a grocery retailing chain to compare the forecasting accuracy of our method with popular seasonal and promotion demand forecasting models at multiple aggregation levels for both short and long forecasting horizons. The significantly more accurate forecasts generated by our model attest to the merit of the approach developed here.     

Demand and Revenue Management of Deteriorating Inventory on the Internet: An Empirical Study of Flash Sales Markets

In this study, we investigate demand and revenue management of deteriorating inventory in flash sales markets (FSM). Retailers operating FSM source excess inventories of products from secondary markets in fixed lot sizes and offer them as part of deals that get no replenishments and expire after running for predetermined periods (e.g., 24 hr) or when they sell out, whichever occurs first. We develop a demand forecasting model that incorporates the effects of sentiments conveyed by consumers in discussion forum posts associated with different deals on the deals’ empirical demand rates. We then conduct a survival analysis to find that the empirical demand rates projected from our forecasting model are significant predictors of the deals’ actual time to stockout, even after controlling for their initial inventory provisions and markdowns. We also find that the predicted effects of these demand rates on stockout times are stronger at low markdowns. Our investigation offers insights into different strategies that sellers operating FSM can pursue to improve their inventory performance. These strategies involve decisions that sellers must make both a priori, before deals start, and a posteriori, according to real‐time detection of departures from projected demand rates as deals run their course.     

Modeling and Forecasting the Sales of Technology Products

Managers in technology product markets require sales response models that provide substantive insights into the effects of marketing activities as well as reliable sales forecasts. Such markets are characterized by frequent introductions and withdrawals of multiple models by different companies. Thus, the data available on the performance of any individual model is scarce. A second characteristic is that the effects of product attributes and marketing activities could change over time as different types of consumers participate in the market at different points in time. Given sparse data, it becomes critical to specify a model that allows pooling of information across brand-models while at the same time providing brand-model specific parameters. We accomplish this via a hierarchical Bayesian model specification. Further, to capture the effects of changing consumer preferences over time, we specify a time varying parameter model. Our modeling framework therefore, integrates a hierarchical Bayesian model within a time varying parameter framework to develop a dynamic hierarchical Bayesian model. We employ data on digital cameras in the U.S. market to estimate the parameters of our proposed model. We use thirty-three months of national level data on the digital camera market with the data series beginning very close to the inception of this product category. We find that while there is little variation in reliance of benefits by early adopters, the second wave of adopters focus on Ease of Use followed by later adopters who rely on Storage and Image Quality. Looking at the elasticities of demand with respect to the various benefits, we find that at around the halfway point of our data series, the industry as a whole would have been better off investing in increasing image quality rather than storage if costs associated with the two are equal. However, at the end of the time horizon both benefits appear to have about equal impact. Further, the relative benefits of improving these attributes vary across brands and points in time. We then generate single period and multiple period ahead sales forecasts. We make different assumptions about information availability and find that the average (across brand-models and time) MAPE ranges from 7.5 to 14.5% for the model. We provide extensive comparisons of our model with 4 potential alternatives and find that our model outperforms these alternatives on the nature of substantive insights obtained as well as in forecasting out-of-sample especially when there is a very short time window of data.     

EXPLORING THE POTENTIAL EFFECTS OF FORECASTER MOTIVATIONAL ORIENTATION AND GENDER ON JUDGMENTAL ADJUSTMENTS OF STATISTICAL FORECASTS

The growing adoption of demand collaboration initiatives such as Collaborative Planning, Forecasting, and Replenishment (CPFR) has made judgmental adjustments of forecasts, an already widespread forecasting practice, an increasingly routine part of many logistics managers' responsibilities. This article investigates how logistics managers might improve forecast accuracy by judgmentally adjusting statistical forecasts and potential factors that may influence the effectiveness of such adjustments. In particular, our goal is to expand current knowledge in this area by focusing on individual differences, specifically motivation and gender, which have been thus far neglected in the extant literature. Our findings indicate that motivation has a significant effect on accuracy improvement and this relationship is moderated by gender. Managerial implications of these findings and future research opportunities are also presented.     

Using Box-Jenkins techniques in sales forecasting

This study analyzes the form, stability, and accuracy of Box-Jenkins forecasting models developed for 27 sales series. The order of autoregressive, differencing, and moving average factors is shown for each complete model along with “goodness of fit” criteria. Forecasting models are then presented for a reduced data set and accuracy is compared with seasonally adjusted linear regressions. The results suggest that Box-Jenkins models are often unstable, “goodness of fit” criteria are a poor guide to the best forecasting models, log transforms do not improve accuracy, and Box-Jenkins forecasts are usually (but not always) better than projections made with linear regression techniques.     

Market Reaction to Information Shocks—Does the Bloomberg and Briefing.com Survey Matter?

Bloomberg and Briefing.com provide competing forecasts for prescheduled macroeconomic announcements. This study examines the accuracy of these forecasts and market reactions to announcement surprises. Our results show that the Bloomberg survey is slightly more accurate than the Briefing.com survey. More importantly, although announcement surprises based on both surveys have a significant effect on the trading activities and returns of S&P 500 futures contracts, the Bloomberg survey subsumes the explanatory power of the Briefing.com survey. The findings suggest that on average Bloomberg forecasts are more consistent with the market consensus view. In addition, we provide evidence of asymmetric market reactions to positive versus negative announcement surprises. In particular, the market reacts strongly to inflation news in the Consumer Price Index (CPI) and Producer Price Index (PPI) announcements and negative shocks in housing price, personal spending, and retail sales.     

Product sales forecasting using online reviews and historical sales data: A method combining the Bass model and sentiment analysis

Online reviews provide consumers with rich information that may reduce their uncertainty regarding purchases. As such, these reviews have a significant influence on product sales. In this paper, a novel method that combines the Bass/Norton model and sentiment analysis while using historical sales data and online review data is developed for product sales forecasting. A sentiment analysis method, the Naive Bayes algorithm, is used to extract the sentiment index from the content of each online review and integrate it into the imitation coefficient of the Bass/Norton model to improve the forecasting accuracy. We collected real-world automotive industry data and related online reviews. The computational results indicate that the combination of the Bass/Norton model and sentiment analysis has higher forecasting accuracy than the standard Bass/Norton model and some other sales forecasting models.     

Subjective Probability Forecasts for Recessions

Probabilistic forecasts are often more useful in business than point forecasts. In this paper, the joint subjective probabilities for negative GDP growth during the next two quarters obtained from the Survey of Professional Forecasters (SPF) are evaluated using various decompositions of the Quadratic Probability Score (QPS). Using the odds ratio and other forecasting accuracy scores appropriate for rare event forecasting, we find that the forecasts have statistically significant accuracy. However, compared to their discriminatory power, these forecasts have excess variability that is caused by relatively low assigned probabilities to forthcoming recessions. We suggest simple guidelines for the use of probability forecasts in practice. JEL Classification E32,E37     

The impact of daily weather on retail sales: An empirical study in brick-and-mortar stores

In this study, we examine the influence of weather on daily sales in brick-and-mortar retailing using empirical data for 673 stores. We develop a random coefficient model that considers non-linear effects and seasonal differences using different weather parameters. In the ex-post analysis using historic weather data, we quantify the explanatory power of weather information on daily sales, identify store-specific effects and analyze the influence of specific sales themes. We find that the weather has generally a complex effect on daily sales while the magnitude and the direction of the weather effect depend on the store location and the sales theme. The effect on daily sales can be as high as 23.1% based on the store location and as high as 40.7% based on the sales theme. We also find that the impact of extreme bad and good weather occurrences can be misestimated by traditional models that do not consider non-linear effects. In the ex-ante analysis, we analyze if weather forecasts can be used to improve the daily sales forecast. We show that including weather forecast information improves sales forecast accuracy up to seven days ahead. However, the improvement of the forecast accuracy diminishes with a higher forecast horizon.     

Consumer inventory sensitivity and the postpromotion dip

We summarize and critique seven theories that might explain the lack of a postpromotion dip in sales in the weeks following a promotion. We then propose and provide empirical support for a new explanation. We argue that in markets where the consumer category purchase decision is not strongly influenced by inventory levels, the displacement effect of accelerated sales will tend to be distributed fairly uniformly into the future such that clearly defined dips are not observed. We utilize a simulation based on real data to investigate this explanation. The simulation shows that given the degree to which inventory influences the purchase decision, we would not expect to see postpromotion dips, even though promotion influences the purchase decision. However, the simulation shows that if inventory had a greater influence on the purchase decision, we would expect to see postpromotion dips. We conclude with implications for both researchers and managers.     

REPLACEMENT PARTS MANAGEMENT: THE VALUE OF INFORMATION

Exponential smoothing (ES) and weighted moving average (WMA) are the predominant methods used to predict future demand for replacement parts. They require simple calculations and make use of information readily available. By gathering more information and doing additional calculations, more accurate forecasts can be developed. However, the cost of collecting the additional data could exceed the inventory cost savings from the better demand forecast. This paper presents a straightforward method for determining when the benefits of a more complex forecasting method outweigh the total costs required to use the method.     

Sales expectations and inventory changes in retail trade

In this paper a single equation inventory investment model is estimated for the United States retail sector. Monthly data for the 1970s are utilized. In estimating the model it is alternatively assumed that expectations were formed according to a seasonal model, to perfect foresight, and to a narrowly rational expectations model. We find that a model in which expected sales and the expected rate of inflation are narrowly rational can explain most of the variation in retail inventory investment during the time period studied. The results of the estimation imply that retailers have a relatively short forecast horizon, that they can react quickly to either unexpected sales or to a deviation of actual from desired inventory stocks, and that an increase in the real of interest has a statistically negative impact on retail inventory investment.     

Forecasting Components of Consumption with Components of Consumer Sentiment

We present new evidence that existing, but long-ignored, measures of consumer sentiment can reduce errors in forecasting total consumption expenditures and its components. The component questions of the aggregate Index of Consumer Sentiment improve forecasts, not only of consumer expenditures on durables but also on non-durables and services. Empirical studies have historically focused on whether consumer sentiment improves one-quarterahead forecasts of consumer expenditures. In fact, we document that measures of consumer sentiment are especially predictive at the longer, four-quarter-ahead horizon. In addition, they typically contribute at least as much to one-quarter-ahead and four-quarter-ahead forecasts of consumption as do income and wealth variables. Out-ofsample forecasts for the 2000-2005 period further substantiate that measures of consumer sentiment can reduce consumption forecasting errors appreciably. JEL Classification C53,E21     

Why is my sales force automation system failing?

Sales force automation (SFA) is the use of software to automate sales tasks, including sales activities, order processing, customer management, sales forecasting and analysis, sales force management, and information sharing. An SFA system is often part of an enterprise-wide information system that connects and integrates sales activities with the organization's other operations. Therefore, SFA software is not only a tool critical to the success of today's sales force, but is also vital to the entire organization. SFA has the potential to empower companies to more efficiently manage their sales force and sales processes, to automate and standardize sales activities, and to connect the sales force with the rest of the organization. The value of these potential benefits in terms of lower costs or increased revenues has encouraged businesses to adopt SFA. Once adopted, however, SFA systems often fail to deliver anticipated benefits. The leading cause of SFA failures has been revealed as low user acceptance, which can be attributed to such factors as the disruption of established sales routines, sales force perception of the system as a micromanagement tool, differences in sales force and managerial expectations for the system, and lack of managerial support for the system as perceived by the sales force. Given these circumstances, managers who are aware of the major issues surrounding user acceptance of SFA will be more successful in implementing such systems. This article explores the utilization of SFA, the benefits derived from these systems, and user acceptance issues. Herein, we offer suggestions that will help organizations succeed in adopting SFA systems.     

Comparative evaluation of empirical response functions in sales management decision making

Although the potential value of an empirical response function as a sales management decision-making tool is evident, the question of whether functions with different properties will provide comparable decision guidelines has not been investigated. One important property of these functions is their ability to explain variation in sales response relationships. The present study develops and comparatively evaluates two response functions that differ in explanatory power as expressed by their R² values. Each function is used to make sales forecasts and determine optimal call levels to the same set of randomly selected retail accounts, and the two functions provide different decision guidelines. This article discusses the implications of these results and future research directions.     

Supply Chain Networks With Multiple Retailers: A Test of the Emerging Theory on Inventories,Stockouts, and Bullwhips

While various approaches to mitigating the bullwhip effect have been proposed, the composition of the underlying supply chain is often taken for granted. This article develops a set of simulation models to investigate changes to the supply chain itself and their impact on the bullwhip effect, on‐hand inventory, and stockouts. It is shown that particular supply chain networks have an impact on the bullwhip effect. Furthermore, the impact of supply chain networks on the bullwhip effect is moderated by the demand forecasting technique used. Finally, supply chain networks, forecasting techniques, and their interactions are found to influence on‐hand inventory levels and stockout rates for firms within the supply chain. Results also suggest that no one particular type of supply chain network dominates in terms of dampening the bullwhip effect, lowering on‐hand inventory levels, or reducing stockout rates. The optimal network depends on the forecasting technique used and other supply chain factors.     

Rational expectations and market efficiency in the U.S. live cattle futures market: The role of proprietary information

The role of proprietary information in forecasting and market efficiency in the U.S. live cattle futures market is investigated. Using a unique proprietary data source collected by a private firm, we test whether the initial estimates in the USDA Cattle on Feed Report and the Knight‐Ridder pre‐release forecasts are unbiased and efficient forecasts of final revised USDA Cattle on Feed Report numbers. We then use these results to test whether futures price movements are predictable based on information in the proprietary data. We also test whether the initial estimates from the Cattle on Feed Report have new information that moves prices once the information contained in the proprietary data source has been taken into account. Results suggest that the information contained in the proprietary data source does have statistically significant explanatory power for forecasting final revised Cattle on Feed Report numbers and for predicting short‐term price movements of futures contracts. The results are inconsistent with strong‐form market efficiency in the live cattle futures market. We also find that the initial estimates in the Cattle on Feed Report still have new information that moves prices even after accounting for the unique information in both the Knight‐Ridder pre‐release forecasts and the proprietary data. © 2004 Wiley Periodicals, Inc. Jrl Fut Mark 24:429–451, 2004     

Multichannel customer purchase behavior and long tail effects in the fashion goods market

The fast-paced growth of e-commerce is impacting the type and variety of products consumers purchase across channels. A commonly held theory, known as long tail theory, posits that online sales are less concentrated at the top of the sales distribution than offline sales, and that more variety is bought online, making the tails of the overall sales distribution denser with the growth of e-commerce. Most of the literature testing the long tail theory has focused on examining entertainment goods markets that do not require much physical examination, and has predominantly found results consistent with the theory. However, the magnitude and antecedents of the observed long tail effects might be different for product categories containing products that require more physical examination before purchase, such as fashion goods. In this study, using detailed individual and transaction level panel data from two multichannel fashion goods retail brands, we show that while the shift to the online channel results in a decrease in the concentration of overall sales for both brands, this change mostly results from consumers buying different products online rather than consumers buying a greater variety online compared to offline. We show that the flattening of the overall sales distribution with the growth of e-commerce in our data is driven by consumers sorting their purchases into channels based on product characteristics. In contrast to the recommendations from the previous long tail literature, our results show that fashion apparel retailers do not need to offer broader assortments online compared to offline, but they may find it profitable to carry or emphasize a different product mix online compared to offline. Our results also provide guidance to fashion goods retailers in curating their online and offline assortments and setting inventory management strategies across the channels.     

A Comparison of Consensus and BVAR Macroeconomic Forecasts

This paper provides a one-month-ahead, macroeconomic, Bayesian Vector Autoregressive (BVAR) forecasting approach that offers several advantages over conventional short-term forecasting procedures. In particular, it produces more accurate forecasts than the Bloomberg consensus forecasts, on average, for 20 major macroeconomic variables. In addition to a quantitative comparison of BVAR and Bloomberg consensus forecast, the paper focuses on five important areas of macroeconomic forecasting: the role of short-term macroeconomic forecasting, the importance of a robust forecasting approach, the importance of timing of data releases, forecast evaluation criteria, and the importance of changing model specifications as conditions warrant.