As for the issue of new product success and failure, an abrupt sales change at early diffusion stages such as the event of takeoff is probably one of the most critical information, because such a change determines the optimal allocation of managerial resources to various decision areas including production, R&D, and marketing mix strategies. Despite high odds of new product failure and its enormous impact on a company`s survival, the diffusion literature has only a few studies examining the event of takeoff, not the issue of structural change accounting for epidemic new product failure as well. The event of sales takeoff implicitly assumes the eventual success of new products and recognizes the first abnormal sales increase in product life cycle (PLC). However, most of the new products ever introduced flop in the market and even the successful ones are likely to go through volatile sales patterns. In addition, as the time to takeoff varies across products, the industry takeoff does not necessarily coincide with a specific firm`s takeoff. Understanding structural change can help a company decide when to pull the plug on a new product that does not perform well or when to step on the accelerator to shorten the time to takeoff of a successful innovation. In view of the importance of the issue of structural change in diffusion process, it is fruitful to carefully study the event of structural change with respect to its definition, identification, and key determinants. First, we explicitly address the nature of new product sales phenomena by defining structural change composed of new product failure as well as success in the marketplace. In the paper we define the event of structural change as the point of systematic shift in sales patterns at the early stages of product life cycle, which encompasses not only a significant sales increase but also a sales decrease in the diffusion process. Therefore, the definition of structural change subsumes the event of takeoff as a special case in which takeoff is identified as the first positive turning point from introduction to growth stage of PLC. Second, we propose decision tools to identify structural change from both the firm and industry perspectives. At the firm level sequential analysis is employed to determine structural change. By incorporating a management perspective in the form of sales forecast, the sequentially updating algorithm helps a company classify structural change and make an opportune decision at its discretion. By assessing negative as well as positive departure of sales from the prediction, it can capture either a market failure or a success concurrently. From the application viewpoint, critical is the immediate identification of structural change from the firm`s perspective, especially when market environments require immediate managerial actions. By fitting the sequential analysis to the datasets of paired actual sales and forecasts for 13 product categories, the paper illustrates how one firm can determine structural change at the firm level. If the company has a sales forecast and pre-specifies a significant mean change, then it can test sequentially whether structural change takes place according to its own criterion. Since this sequential decision process may detect one point as structural change due to inaccurate prediction, not because of significant change in sales, however, a three-stage hierarchical Bayes (HB) model is formulated on the basis of analyzing actual sales variation only. The proposed Bayesian model, which examines past aggregate sales data and captures any systematic shift in sales trends in conjunction with the Bass model, can provide a valid measure of identifying takeoff at the industry level. When sales are assumed to follow well-established diffusion models such as the Bass model, the occurrence of structural change is tested and identified by inferring any significant change in the model parameters from a given sales trajectory. Especially, estimation of takeoff using the posterior means of the Bass model parameters provides rigorous statistical validation that the existing approaches to takeoff identification require. The lack of statistical measures in takeoff identification can lead to substantial bias in examining the role of key factors on takeoff. The use of the Bass model also has the advantage of understanding the event of takeoff on the vast knowledge of diffusion process. By analyzing number of wireless subscribers across 64 countries, the HB model identifies the event of takeoff in the worldwide wireless service industry. The time to takeoff is relatively homogeneous across countries: its mean is 10.4 years and standard deviation is 2.7 years. Third, we examine the impact of key factors on the probability of takeoff occurrence by calibrating various specifications of proportional hazard regression, which use the posterior mean of takeoff as the dependent variable in the analysis. By assessing a single innovation across countries over moderately uniform time periods, we draw a clear-cut picture of takeoff characteristics. Among all the variables we investigated, quality improvement has the largest influence on takeoff. The introduction of digital technology, a technological breakthrough at a certain point in times, boosts the hazard rate of takeoff more than 10 times. Therefore, functionality change can lead to enduring substantial sales growth without sacrificing profits. Contrary to the conventional wisdom, however, price turns out to have no significant impact on takeoff in the wireless industry, which possibly results from the industry characteristics: the wireless market had been regulated by governments and operators had adopted the price policy of keeping price level constant with additional services added. Introduction year is not related to takeoff in general conditions. Because introduction year accounts for continuous change in market environments over PLC but takeoff is a specific turning point in PLC, it does not explain discrete takeoff event. Penetration may not be a proper metric of indicating takeoff. Innovation type and characteristics primarily determine both diffusion rate and penetration level, and takeoff by definition signals a transition in diffusion process. Thus, penetration level at takeoff does vary across products as diffusion speed does. In the paper, we study the issue of structural change in diffusion process in terms of the definition, identification, and nature of structural change. The findings reported here can shed new light on the mysteries of new product failure and success for both researchers and practitioners.