Using Data Forecasting to Improve Community College Collections Development

Three techniques for improving your collection development

Posted on May 5, 2025 in Blog Posts

No matter the discipline, academic scholarship is always evolving, which makes collection development a difficult task, especially when library resources are so limited. Sam Suber, the Electronic Resources Coordinator at Moraine Valley Community College Library in Palos Hills, Illinois, has made it a little easier through his excellent chapter in The Community College Library: Collections and Technical Services (ACRL, 2025). His chapter, titled “Collection Development Through Data Forecasting” and republished below through a CC BY-NC license, provides a wonderfully accessible introduction to a technical topic—a DIY guide to using a spreadsheet and some data and coming to some actionable conclusions.

With plenty of screenshots, Suber will explain what data forecasting is and then walk you through three different techniques for data forecasting using Microsoft Excel. If you’re more interested in applications for data forecasting, skip about halfway down the page and read his excellent case study of using data forecasting for Hoopla.

📙 Learn more about The Community College Library: Collections and Technical Services

---

Introduction

Collection development is a critical process for libraries, as it determines what materials will be included in the collection and which will be removed. Librarians often rely on various criteria, such as popular reviews and usage statistics, to make these decisions. However, collection development can be approached more objectively using data forecasting techniques.

In this chapter, we will explore how quantitative methods such as moving averages, weighted moving averages, and exponential smoothing can be used to forecast collection trends and make informed decisions about what materials to add or remove. The chapter also discusses how to measure the accuracy and reliability of these forecasts using error measures. Finally, we will demonstrate how data forecasting can be applied to collection development in a community college library and how these techniques can be valuable in other areas of library management.

To illustrate the practical application of these techniques, we will examine a scenario in which a community college library must determine the viability of acquiring Hoopla, an e-content platform that offers audiobooks, movies, e-books, television shows, and e-magazines for recreational use. By analyzing the demand for Hoopla using data forecasting, the library can predict how well it will be used in the future and whether it is worth keeping in the collection.

To understand the significance of data forecasting in collection development, it is helpful to understand the concept of collection development and its key components. This includes analyzing the community’s needs, selecting materials that align with those needs, and continuously evaluating and updating the collection to ensure that it remains relevant and useful to patrons. Furthermore, we will delve into the importance of predicting collection development trends and provide an overview of various data forecasting techniques. This will allow us to see how these techniques can be applied in library management, specifically in collection development, to make informed decisions and better serve the needs of the community.

---

🔥 Stay up-to-date with LibTech Insights by signing up for our free newsletter. Just one weekly email with our new blog posts, top tech news stories, and other bonus content. Check out some posts from our archive:

• The Power of Data Storytelling in Libraries
• Trusted Guides: How Academic Libraries Can Bridge the AI Divide
• How to Interview Your Library’s Patrons

---

Definition of Collection Development

Collection development is the process of identifying, acquiring, and maintaining library materials that align with the needs, interests, and curriculum of the library’s patrons. It is the process of building and managing a library’s collection of books, periodicals, multimedia materials, and other resources. It involves analyzing the needs of the community, selecting materials that align with those needs, and continuously evaluating and updating the collection to ensure that it remains relevant and useful to patrons. Collection development can also include weeding outdated materials, budgeting for materials, and creating policies for selecting, purchasing, and withdrawing materials. The goal of collection development is to provide patrons with access to a wide range of high-quality resources that support their learning and research needs.

The Importance of Predicting Collection Development Trends

Predicting collection trends is essential for collection development because it allows librarians and other professionals to anticipate the needs and interests of their patrons and select materials accordingly. By staying ahead of the curve, libraries can ensure that their collections are relevant and engaging, which can lead to increased usage and satisfaction among patrons.

One key benefit of predicting collection trends is that it can help libraries to identify popular and emerging topics, and allocate resources more effectively. For example, if a library sees an increase in interest in a particular subject, or a particular author, it can respond by acquiring more materials on that topic or author. This can help to ensure that the library’s collection stays current and prioritize which materials to acquire and how much to spend on them. 

Another benefit of predicting collection trends is that it can help libraries to identify gaps in their collections or areas where they may need to de-access or withdraw older materials. For example, if a library sees a trend toward more diverse and inclusive materials, it can respond by acquiring more materials that represent different perspectives and cultures. This can help to ensure that the library’s collection is inclusive and reflective of its community.

Finally, predicting collection trends can help libraries to better engage with their patrons. By anticipating their needs and interests, libraries can create programming, events, and other activities that are relevant and draw community interest. This can help to build stronger connections with patrons and increase their sense of ownership and belonging with the library.

Overview of Data Forecasting Techniques

Three specific data forecasting techniques that can be applied in collection development are moving average, weighted moving average, and exponential smoothing. These techniques can provide valuable insight into predicting trends and changes in demand for library materials. A deeper explanation of each of these techniques is explored below.

Moving averages is a data forecasting technique that involves taking the average of a set of data points over a certain time.[1] For example, a library could use moving averages to forecast the number of book loans in the next month by averaging the number of book loans from the past three months. This technique is useful for collection development because it can help the library identify trends in the use of certain materials and make decisions about whether to purchase more of those materials or not. It can also be used to predict changes in demand for different types of materials and adjust the collection accordingly.

Weighted moving averages is a variation of moving averages where each data point is assigned a weight, usually based on its proximity to the current time-period.[2] This technique is useful for collection development because it can assign more value to recent data, which is more relevant for forecasting. For example, a library could use weighted moving averages to forecast the number of book loans in the next month by giving more weight to the number of book loans from the past month than to the number of book loans from the past three months. This can help the library make more accurate predictions about future demand for certain materials.

Exponential smoothing is a data forecasting technique that involves using a smoothing constant to give more weight to recent data and less weight to older data.[3] This technique is useful because it can predict changes in demand for different types of materials, which allows libraries to adjust the collection accordingly. Additionally, this technique can help in predicting the demand for certain materials over a longer time, providing more insight into future collection needs.

Data Forecasting Techniques in Collection Development

In this following section, moving averages, weighted moving averages and exponential smoothing will be demonstrated by using sample data with Microsoft Excel. Figure 1 below illustrates the sample data we are using and Figure 2 represents the same data in a graph:

To grasp the concept, it’s important to understand the relationship between each column in Figure 1: Years, Quarters, Periods, and Demand. These terms are all ways to measure and track the passage of time. By analyzing the data associated with these measurements, we can gain insights into patterns and trends that can aid in forecasting future demand. In other words, they are the building blocks that help us make educated predictions about the future.

A year, of course, is a twelve-month period that is used to measure time in most calendars. For example, 2022 is a year. A quarter is a period of three months that is used to measure time in some calendars. For example, Q1 of 2022 is the first quarter of the year 2022, which would include the months January, February, and March. Periods are employed to measure time in a sequential and easier-to-read fashion. Essentially, they provide a means of representing time intervals in a more plain manner. For instance, when examining Figure 1, periods can be represented by numbers such as one, two, or three, corresponding to durations like a quarter, month, or week, respectively. This numbering system aids in the clear and organized description of each time interval in column 2. Demand can refer to how many books are borrowed from the library by patrons. For example, if a library has ten copies of a book and fifteen patrons want to borrow it, the demand is fifteen. It is important to note that libraries can establish their own criteria for defining and assessing demand, tailored to their specific circumstances.

Moving Averages

To perform moving averages in Excel, we can create a new column labeled “4 Period MA,” with four referencing the four quarters and “MA” as an acronym for “moving average,” as shown in Figure 4. By using the formula =AVERAGE(D2:D5) in cell E6 and copying it down to cell E14, we can calculate the moving average. This moving average is a calculation that takes the average of the last four periods of data. As we continue to copy the formula down the cells, the averages will change as it considers different periods of demand. It is worth noting that this method of moving average only predicts one quarter into the future.

Weighted Moving Averages

To perform weighted moving averages in Excel, we can create a new column labeled “4 Period WMA,” with “four” referencing the four quarters, and “WMA” as an acronym for “weighted moving average,” as shown in Figure 4. We use the formula =SUMPRODUCT($M$2:$M$5,D2:D5) in cell E6 and then copy it down to cell E14.

The SUMPRODUCT function in Excel is a powerful tool that allows you to perform calculations on multiple arrays of data. It multiplies the corresponding items in the arrays by the set weights and then sums the resulting products. The weights in cells M2 to M5 play a crucial role in determining the weighted average. It’s important to note that all weights must add up to a total of one. For instance, a larger weight like t-1 concentrates the weighted average on more recent data, giving it greater significance. In contrast, the other weights have less impact, meaning they contribute less to the overall average. Libraries have the flexibility to adjust these weights based on their specific needs and considerations, such as the desired emphasis on recent trends and the relative importance assigned to different time periods.

It is important to note that this method of weighted moving average only predicts one quarter into the future. For this specific example, we are using the demand data from the past four quarters and multiplying it by set weights to make a prediction for the next quarter.

Exponential Smoothing

Performing exponential smoothing in Excel is a simple process, demonstrated in Figure 5. First, we create a column labeled “Forecast” and in cell E2, we use the formula =AVERAGE(D2:D13) to calculate the average demand. Next, in cell E3, we apply the formula =D2*$M$2+E2*(1-$M$2) and then copy it to cell E14. This formula takes the average demand and multiplies it by the smoothing constant, also known as Alpha (a or cell M2), which we have chosen to be 0.1 in this example. The purpose of the smoothing constant is to attain optimal data weighting and minimize the forecast’s sensitivity to data fluctuations. This constant is crucial as it ensures that the data is evenly weighted, resulting in a more accurate and reliable forecast. It is important to note that the smoothing constant must fall within the range of zero to one to maintain its effectiveness.[4] It controls the amount of weight given to recent data versus historical data, allowing the forecaster to adjust the forecast to their desired level of sensitivity. If a higher smoothing constant such as 0.9 is chosen, the forecast will place more weight on recent data and will be more sensitive to changes in the data overall, while a lower constant such as 0.1 would weigh the historical data more heavily and make the forecast less sensitive.

It is important to keep in mind that exponential smoothing is limited in its ability to predict future values. Specifically, it only allows for the prediction of a single period in the future, as seen in cell E14. This means that it generates a single forecast value for the next period, and that forecast value is based on the historical data up until the current period. However, it’s worth noting that this limitation does not render the method useless. It’s still an effective approach for short-term forecasting and is widely used in practice due to its simplicity.

Error Measures in Data Forecasting

When making collection development decisions in libraries, it is important to understand the accuracy and reliability of the forecasts being used. To achieve this, librarians use various error measures to evaluate the performance of their forecasting models. In this context, error, absolute error, mean absolute deviation (MAD), and tracking signal are four key measures that are used to assess accuracy.

Error refers to the deviation between the forecasted value and the actual value, and it can be positive or negative.[5] This difference is expressed in the same units as the data being forecasted. Absolute Error, on the other hand, takes the absolute value of the error, which is always positive, and measures the magnitude of the deviation regardless of its direction. MAD is the average of the absolute errors, providing a measure of the average magnitude of deviation from the actual values.[6]

A low value of these error measures indicates high accuracy in the forecast, and they are important for evaluating the accuracy of different forecasting methods and for improving forecasting models by identifying where errors are being made.[7] Additionally, Tracking Signal is a measure of forecast accuracy over time, calculated as the average of the absolute differences between the forecast and the actual data divided by the average of the absolute actual data.[8]

By combining these error measures, librarians can gain a comprehensive understanding of the accuracy of their forecasts and make informed collection development decisions.

Error and Absolute Error

To calculate the error in Excel, in Figure 6, we use the formula =E6-D6 in cell F6 to subtract the actual demand (D6) from the forecasted value (E6). We then copy this formula down to F13 to obtain the error for each period.

To calculate Absolute Error in Excel, in Figure 7, we use the formula =ABS(F6) in G6 and copy it down to G13.

MAD and Tracking Signal

To check the accuracy and reliability of our forecast, we will use two important error measures: the mean absolute deviation (MAD) and the tracking signal. A lower MAD score means our forecast is more accurate.[9] If the tracking signal is outside of +6 or -6, it means the forecast might be biased. Bias means that the predictions made by the model or method used aren’t accurate and are consistently off from the real results. This situation can occur if the model has incorrect ideas built into it or if there isn’t enough information to make accurate predictions. These incorrect ideas can include things like incorrect data, flawed assumptions, or biases. It’s important to identify and address these issues because people use these predictions to make decisions and plans.

To calculate the MAD in Excel, we use the formula =AVERAGE($G$6:G6) in cell H6 and copy it down to cell H13, as shown in Figure 8.

Figure 9 shows the calculation of the tracking signal in Excel, where we use the formula =SUM($F$6:F6)/H6, in cell I6 and copy the formula down to cell I13.

Based on the error measures, we can conclude that this forecast displays a high level of accuracy. The MAD value of 982 at H13 is notably low, and there is no evidence of bias as the calculations between cells I6 and I13 fall within the range of ±6.

Applying Data Forecasting to Collection Development: A Case Study of Hoopla

In this section, we will use Hoopla as a case study to demonstrate how data forecasting techniques can be applied to collection development. By analyzing the demand for Hoopla using various forecasting methods, we will determine which method is the most effective for predicting the demand for this resource. Additionally, we will compare the accuracy and reliability of these different methods using error measures such as error, absolute error, MAD, and tracking signal. This case study will illustrate how data forecasting can be used to make informed decisions about the viability of adding specific materials to the library’s collection.

Methods

This case study will utilize a sample of current circulation data for Hoopla materials to accurately predict the circulation patterns for the first quarter of 2023. By experimenting with different forecasting methods, we can determine which is the most accurate. Figure 10 shows the sample data utilized.

Figure 11 illustrates the same data in a graph.

Results

Moving averages (Figure 12), weighted moving averages (Figure 13), and exponential smoothing (Figure 14) were applied to the sample data with error measures to gauge the accuracy and reliability of each forecast.

Moving Average

Weighted Moving Average

Exponential Smoothing

Summary

Based on the data shown in Figure 15, it is evident that exponential smoothing generates the most accurate forecast with the lowest MAD value of 79, compared to moving average and weighted moving average with values of 90 and 99 respectively. Moreover, the tracking signal analysis for all three forecasts shows no bias, as they are all within the range of 6 and -6, which can be found in the TScolumn in Figures 12, 13, and 14. Therefore, it can be concluded that exponential smoothing is the optimal method for predicting Hoopla circulations during the first quarter of 2023, among the three approaches evaluated.

Conclusion

Data forecasting techniques provide significant benefits for collection development in libraries. Accurate forecasting of future demand patterns can help libraries optimize their resource management, enhance the user experience, and make informed decisions about the acquisition of new materials. By utilizing these techniques, libraries can ensure that they are providing the resources that their users need while also optimizing their operations and budget.

Moving forward, it is important for libraries to consider additional forecasting methods and technologies beyond those discussed in this chapter. Optimization methods can help libraries identify the most efficient allocation of resources and acquisition strategies to maximize the library’s resources. Libraries can also explore the use of other forecasting methods, such as Holt’s method and quadratic trend analysis, as well as regression analysis, to make informed decisions about collection development and resource allocation.[10]

Additionally, it may be beneficial for libraries to consider how they can integrate these forecasting techniques and technologies with other data sources, such as user feedback and social media analytics. This will enable libraries to gain a more comprehensive understanding of user needs and preferences and make more informed decisions about resource allocation and acquisition strategies.

As technology continues to evolve, libraries must also consider how they can stay up to date with new forecasting techniques and technologies. This may involve investing in training and professional development opportunities for staff, as well as exploring partnerships with external organizations that specialize in data analytics and forecasting.

Overall, by embracing a range of forecasting methods and technologies and integrating them with other data sources, libraries can continue to provide effective services that meet the evolving needs of their users.

Additional Resources

Banton, Caroline. “Moving Average, Weighted Moving Average, and Exponential Moving Average.” Investopedia. Last modified May 17, 2023. https://www.investopedia.com/ask/answers/071414/whats-difference-between-moving-average-and-weighted-moving-average.asp.

Shumway, Robert H., and David S. Stoffer. Time Series Analysis and Its Applications: With R

Examples. 4th ed. Springer, 2017.

Notes

[1] Phan Tran, “Moving Averages and Exponential Smoothing in Time Series Analysis,” HKT Consultant, August 31, 2021, https://phantran.net/moving-averages-and-exponential-smoothing-in-time-series-analysis/.

[2] David R. Harper, “Exploring the Exponentially Weighted Moving Average,” Investopedia, updated June 2, 2022, https://www.investopedia.com/articles/07/ewma.asp.

[3] Tran, “Moving Averages.”

[4] Tran, “Moving Averages.”

[5] ImvivRyan, “Basics of Forecast Accuracy,” Medium, July 12, 2020. https://medium.com/analytics-vidhya/basics-of-forecast-accuracy-db704b0b001b.

[6] ImvivRyan, “Basics of Forecast Accuracy.”

[7] ImvivRyan, “Basics of Forecast Accuracy.”

[8] George E.P. Box et al., Time Series Analysis: Forecasting and Control, (Wiley, 2016).

[9] ImvivRyan, “Basics of Forecast Accuracy.”

[10] Bruce Bowerman et al., Forecasting, Time Series, and Regression (Cengage Learning, 2004).

Bibliography

Bowerman, Bruce L., Richard O’Connell, and Anne Kohler. Forecasting, Time Series, and Regression. 4th ed. Cengage Learning, 2004.

Box, George E.P., Gwilym M. Jenkins, and Gregory C. Reinsel. Time Series Analysis:

Forecasting and Control. 5th ed. Wiley, 2016.

Harper, David R. “Exploring the Exponentially Weighted Moving Average.” Investopedia.

Last modified June 2, 2022. https://www.investopedia.com/articles/07/ewma.asp.

ImvivRyan. “Basics of Forecast Accuracy.” Medium, July 12, 2020. https://medium.com/analytics-vidhya/basics-of-forecast-accuracy-db704b0b001b.

Tran, Phan. “Moving Averages and Exponential Smoothing in Time Series Analysis.” HKT

Consultant, August 31, 2021. https://phantran.net/moving-averages-and-exponential-smoothing-in-time-series-analysis/.

Related Posts