Fine Tune Your Forecasts with Advanced Controls
The blog explains how Pyramid’s BI platform enhances forecasting by allowing users like Lauren, a BI analyst, to fine-tune pre-configured models such as Holt Winters through advanced controls like outlier sensitivity and smoothing weight adjustments, enabling more accurate and customizable sales forecasts within existing reports and dashboards compared to typical “black box” methods.
In a previous blog, Pyramid’s unique forecasting capabilities were discussed, highlighting the ability to choose from among eight algorithms to produce forecasts within existing reports and dashboards—even on large datasets. This blog focuses on how users can further enhance their forecasts by applying advanced adjustments to the pre-configured models, allowing for fine-tuning and increased accuracy.
The problem
Third-party BI tool vendors often provide proprietary “black box” forecasting methods. Users typically have no understanding of the algorithm being used and lack the ability to tweak the algorithm to better fit their specific analyses. There is also a need to adapt existing algorithms by introducing additional inputs or logic.
Pyramid’s solution
Pyramid’s fully integrated forecasting solution uses a simple point-and-click methodology to provide sophisticated forecasting algorithms within existing reports and dashboards. The advanced forecasting option allows users to tweak the pre-configured models with different choices based on each model selected. Users can also select a custom number of periods to forecast into the future using the forecast dialog.
Business case
Lauren, a BI Analyst for M&N Manufacturers, uses SAP BW to analyze their SAP ERP and Pyramid for BI development and analytics. Lauren wants to perform forecasting tasks on her sales dataset and fine-tune the forecast to generate the most accurate results possible.
Sales forecast
Lauren selects the Holt Winters algorithm to predict future sales. Since she has at least 12 months of sales data, she uses the “Calculate History” option in the Advanced Forecast dialog box to see the forecasts that would have been made. She observes reasonable accuracy in the forecast.
Using the Advanced Forecast dialog again, Lauren tweaks the outlier sensitivity and smoothing weight options further. This results in a more accurate forecast, providing a more realistic means of incentivizing the sales staff.
Sales by Manufacturer Forecast
In a second analysis, Lauren wants to forecast sales per manufacturer to assist in preparing logistics around distribution. She selects the Arima method to forecast sales for the Princess Manufacturer. Lauren then fine-tunes the Arima algorithm to improve her forecast. The resulting forecast is further improved.
Summary
Third-party BI tool vendors often provide no details about the forecasting method used and offer only limited ability to tweak the algorithm. Pyramid’s forecasting solution provides sophisticated forecasting algorithms within existing reports and dashboards with a single click. Recognizing that data analysts often need to apply their own adjustments for more accurate forecasting, Pyramid gives users options for adjusting the algorithms without having to write them from scratch in Python or R. In a future blog, there will be a demonstration of how data scientists can develop their own Python or R forecast scripts to generate their own appropriate solution.