The most frequent question I have noticed executives ask after a forecast is revealed is “what’s next?”, and as data scientists, it is part of our job to answer that and other forthcoming questions, as well as usher C-suite and board members towards the richness our insights can bring about for the organization.
The approach we have developed at Wisdom Analytics caters a forecast towards the efforts of exceeding the company’s goals. The forecast, with its accuracy and error margins, serves as a proxy that cascades a series of actions from the commanding stage and runs in several directions. We have identified three main actions as pivotal events when a forecast is published:
- Gauge: A forecast of a specific KPI will set the pace of concurrent actions by the operations team as communicated by the leadership team. It is a gauge, a valve per se, that helps the organization assess itself as to where they are, and hence, speed of execution becomes essential in a coordinated effort that will require to have everyone on board. This is especially true if the forecast shows declining trends where the opposite is expected.
- Trigger: Depending on the forecasted horizon, it helps operation teams adjust accordingly as they approach the horizon of the first and future data points. For example, we helped a call center forecast its workforce allocation to account for the slow and high demands in the year, so a meticulous analysis of trend, seasonality and cyclicality factors was key in arriving at high accuracy rates. Hiring decisions were made adjusting to the forecasted figures.
- Optional agent input in dynamic optimization: When targets have been set up, especially for the new year, the forecasts may be injected in the optimization equation that tries to resolve that “x, y by z” if its value falls below the target. To illustrate better, we supported a technology company’s multi-million dollar savings campaign by forecasting how much they would save if expense-related metrics changed along the way. Sometimes, if an optimal solution is not found, we may consider re-engineering the equation and plug in the forecast value while dynamically evaluating the metrics’ performance.
Forecasting is not easy, especially if it’s done with basic tools such as Excel, or with built-in forecasting modules in reporting tools like Tableau. Forecasting is as specific and unique as the challenges your organization is trying to solve and it should be handled with heavy-lifting tools languages like R or Python and rolled out with carefully-crafted approaches data scientists are able to devise.
