During my time at Keysight as an intern, I developed solutions for two types of forecasting jobs:  qualitative and quantitative. I also developed dashboards within their application of choice: TIBCO Spofire. Spofire shares a lot of similarities to Tableau. In addition, it let’s you add IronPython, R, and C# scripts to produce virtually any desired output/visualization of your company data. It even let’s you use JavaScript to program friendly UI’s for your dashboard.

Quantitative Forecasting

1. Quantitative forecasting focuses on predicting values for a time-series representation of data. I developed an ARIMA forecast for single dimensional time series data of company order values subject to filters within their data analysis platform TIBCO Spotfire, which only supports Holt-Winter forecasts.

The dynamic ARIMA forecast would generate in Spotfire within 2 seconds using a series of IronPython scripts for data pre-processing and R for the statistical modeling.  It is dynamic in the sense that the forecast works with Spotfire built-in filters. In essence, a user can make insights by activating the monthly, quarterly, or yearly forecast for Product line X, sold to Customers Y in region Z.

Since the solution and the data it presents is proprietary to Keysight, I will not have any code demos or screenshots of the final product. However, I have included a presentation that visualizes the performance of the ARIMA forecast surpassing that of Holt Winters on my team’s historical order value data.

Qualitative Forecasting

2. A supply chain forecast would be an example of a qualitative forecast. The input data is no longer a time series, but a conglomerate of data coming from Oracle baseline forecasts, big deal insider information, new product info (NPI), historic trends, and Salesforce funnel data.  It was up to demand planners to take all of these into consideration when planning their demand forecasts two months ahead of time. It was up to me to develop a centralized dashboard (and move away from localized Excel sheets) to visualize their performance by merging their forecast data with actual order data, as well as explore other opportunities to improve the forecasts.