What if there was a new formula for forecasting a product’s sales prior to a launch that is cost effective, accessible and accurate? There would be no need for costly consumer surveys – that typically only big firms can afford – or expert judgment, which is often biased. Oliver Schaer, PhD, an assistant professor in Drexel University’s LeBow College of Business, together with Nikolaos Kourentzes of Skövde University, Sweden, and Robert Fildes of Lancaster University, UK, led new research recently published in the European Journal of Operational Research that introduces a new, data-driven model for pre-launch predictions of a product’s sales.
Schaer’s new methodology introduces a cost effective strategy by using publicly available data from Google Trends to augment product life cycle models. Schaer spoke with the Drexel News Blog about the research leading to the introduction of this new model and how specific industries can utilize it.
What is your new model for predicting a product’s sales prior to launching it?
Our study introduces a data-driven method for predicting new product sales prior to launch by combining historical analogies with Pre-Release Online Search Traffic (PROST) gathered from Google Trends.
Traditionally, pre-launch forecasting methods rely on biased expert opinions or expensive consumer surveys. This research seeks to overcome these limitations by linking pre-release consumer search volume directly to a product’s overall market potential, enabling firms to generate accurate, long-term forecasts across the entire product life cycle.
We empirically validated the framework on sequential video game sales across franchises and found that PROST contains predictive information up to 18 weeks before release and can increase life cycle sales forecast accuracy by up to 21%. Our approach can be implemented with minimal data requirements, making it a versatile and accessible tool for firms. Moreover, Google Trends data is free and publicly available, allowing smaller studios and firms to gain crucial insights into their operational planning without incurring significant marketing intelligence costs.
Is there anything that surprised you about your findings?
We were a bit surprised that pre-release online search traffic predicted total life cycle sales (End-of-Life) more accurately than it predicted opening-week sales within 9 weeks prior to release. It also turns out that search volume is superior to ad spend information, which had very little predictive value.
Consumer-driven search traffic, however, seems to reflect the true underlying interest far better than corporate marketing budgets. This challenges the traditional corporate assumption that more ad spend will lead to more sales and suggests that tracking and managing consumer intent provides a significantly clearer strategic lens.
How would this new method change how companies make product and marketing decisions?
The approach provides actionable intelligence across five distinct operational decision points:
- Price Determination: Companies can optimize launch prices, plan future discount schedules and structure product bundling throughout the life cycle based on the predicted adoption curve.
- Investment and ROI Management: Reliable lifetime sales forecasts prior to release enable executives to make objective decisions about project continuation, divestment or resource allocation much earlier in the cycle.
- Sourcing and Infrastructure Strategy: For physical goods, long-lead forecasts allow firms to use cheaper, slow-lead manufacturers rather than expensive, short-lead-time replenishment. For digital products like video games, it enables IT infrastructure teams to plan and manage hybrid solutions that include cost-efficient bare-metal servers and virtual machines in the cloud to handle demand spikes more effectively.
- Release Timing: Firms can monitor their own search traffic relative to competitors to optimize launch dates based on expected market success.
- Ad Spending Allocation: Instead of heavily funding post-release campaigns, managers can focus on specific pre-release activities designed to spark and manage initial consumer buzz.
What are the benefits of utilizing this new model in place of current models?
One of the biggest advantages of our model over other quantitative methods in the literature is that it is exceptionally data efficient. Our best-performing model requires data only from a single franchise that has observed one full product generation. We also provide guidance on forecasting “new to the world” products that lack a direct predecessor. By pooling and averaging historical pre-release search and sales from other established franchises within the same publisher, the model can still outperform naïve baselines for entirely new product introductions.
In addition, the model generates a forecast for the entire product life cycle. Therefore, covering both short-term and long-term planning needs. This is important because any short-term forecasting model requires several weeks of training data to produce a reliable forecast. In practice, these short-term forecasts might only become available when the new product has already reached its peak demand. Our single-estimation framework eliminates that operational friction. Lastly, our method relies entirely on firm-owned historical sales data and publicly available Google Trends data, eliminating the need to purchase expensive market intelligence datasets.
Is this specific to video games or can this be implemented for other products?
While the empirical validation focused on the video game sector due to its high online community engagement and heavy reliance on sequels, the methodology is designed to transfer directly to other industries that exhibit multi-generation or sequential product categories with high consumer search involvement. These can include consumer electronics (digital cameras, smartphones and computer hardware components), automotive products and highly anticipated fashion products. We believe that the method is less suited for low-involvement everyday items or niche products, where it will be difficult to measure pre-release buzz.
Is there anything else we should know about this new model?
Our model consistently tends to underestimate consumer demand. This sets itself up well for a potential combination with human expert forecasts, which tend to over-forecast due to overconfidence. Therefore, combining our data-driven method with expert judgment provides the opportunity to mitigate opposing biases and maximize overall organizational forecast accuracy.
Reporters interested in speaking with Schaer should contact Annie Korp, associate director, News & Media Relations, at 215-571-4244 or amk522@drexel.edu.
