VLSIresearch's methodology for compiling information relies heavily on market simulation and consists of several approaches. The basic theme common to all is data triangulation. Wherever possible, we try to box-in an issue from at least three different directions, sometimes more.
Surveys of raw market data are made across a broad front. Basic data is obtained from both suppliers and users. Publicly available information is used in addition to the information gained by questionnaires. Annual reports and other press reports are used wherever possible to augment and validate survey results.
Technical issues and trends are obtained from surveys, technical symposia, technical journals, and trade journals. Technical proceedings of relevant symposia compared with data obtained via private industry sources. As a result, the material developed contains a wide variety of original data that is then further cross-checked and compared with published sources.
Analysis occurs on three fronts. First, a Data Analytics tool is built to closely resemble the real-world market being examined. Second, survey data is assembled to ascertain the supply side of the industry. Third, publicly available information is analyzed to cross-check survey results. Next, buyers' data is assembled to determine the demand side of the industry. Data is expanded, as VLSIresearch's proprietary models decompress the missing pieces to gain intimate detail. All of these sources are analyzed together to triangulate specific industry issues. The data is then confirmed via a data-beta site.
Market forecasting is performed via a combination of economic tools, technological analysis, and judgment. All of VLSIresearch's forecasting models use causal relationships as primary driving forces. Two types of models are used for forecasting the market. One is used for short-term forecasts. The other is for long-term forecasting.
Econometric models are used for short-term forecasting. They are based on inertial driving forces such as economic or general electronic trends. Depending on the need, these may include neural network analytic approaches. The focus is mostly on business aspects such as inventories, backlogs, etc.
Technological market models are used for long-term forecasting. These are based on a merger of technological issues, economic frameworks, and business principles.
Judgmental and non-linear error analysis is added to these as time unfolds with each data set to enhance forecasting accuracy.