# Bass model

Hi. Is it feasible that SpiderFinancial develops an Add-In for the Bass model? This model is a diffusion model to forecast sales of a new product for which there is no time series data since the product has not been launched yet. The model was developed by Prof. Frank Bass almost fifty years ago. Initially, it was intended for the forecast of durable products, and gradually, it was used for the forecast of different kinds of products and services. From pharmaceuticals, vehicles, agricultural equipment, photovoltaic systems, mobile and internet use, and so on.

The model is a diffusion model because it relies on the idea that the first adopters of the innovation spread the word of the application used by "word of mouth," and others imitate behavior and adopt use later.
The issue is that to forecast sales; there is a need to use two coefficients: p (known as the innovation coefficient) and q (the imitatiton coefficient). Also, there is also a need to use the market potential, m. The model is represented by a Riccati type differential equation. Its solution form is used for the forecast. There are three forms to obtain such parameters: linear regression, nonlinear regression, and similar products.

Nonlinear regression is the best alternative because linear regression delivers biased parameters. This is so because the model is discretized, and the original model is a continuous function. Analogous imply using coefficients of similar products. For instance: internet use is similar to tweeter use.

Anyway, most excel templates rely on linear regression and excel solver to estimate parameters. Unfortunately, the solver algorithm does not guarantee that the solution is a global optimal. It could be only a local optimal. The optimization criteria is the MSE. This is another reason why nonlinear regression is more reliable. Unfortunately, I have not seen many Add-Ins using the nonlinear method, and this means relying on R or Matlab. But this means scripts must be used. Add to this that initial values are required to run a nonlinear regression. It is almost an ordeal to get the initial values using the NLS package from R. It would be convenient to have an Add-In that would simply allow to plug in some input values and, with just a few clicks, deliver the coefficient values.