LinkedIn
Twitter
I just enabled the Forecasting feature in Looker and let me say that the charts that can be produced look absolutely amazing. The only problem with the feature is enforcing a certain model to generate the forecast and I don’t understand the reasoning behind that decision. Using a generic model can be useful for some applications, but that’s usually not the case. Many companies have their own proprietary models that they use to generate forecasts. So instead of enforcing a certain model, why just let the user use their own forecasted data?
When it comes to forecasting, transparency is very important, and one way this can be done is to show historical forecast and compare it with actual data. I posted a chart below that demonstrate one way to visualize forecasted data that we generated by our own models with the actual data. If we use the forecast feature, it looks really good but it doesn’t work very well because we’re using different models here. Alternatively, if the user can separate their own data to show historical forecast with a solid line and new forecast with dashed line as well as a highlighted region showing confidence interval it’s gonna tell a much better story.
The long story short is just let the user control their data. Looker is great because it enables users to visualize and control their data. Forecasting feature is cool because of the visualization it produces but the model itself is not. Using a “Looker” algorithm can be a nice option but enforcing it as the only option is bad.
