Let’s say you have 8 teams of 4 people making widgets. Each “batch” of widgets is produced every two weeks like clockwork. Each widget is unique enough so the team estimates with a number. Higher the number the greater the complexity.
So the most obvious measurement over time is the average correct? 80 widgets produced over 2 months is 4 batches , so 20.
Managers notice that this is not a predictable way to forecast though. Each batch produced might be as low as 2 or a high 30. So accuracy forecasting any given batch is not valuable.
The historical data goes back years and years. To the naked eye you can also see production dips in Nov and Dec, but generally increases during march and May.
Other data points available.
Cycle time, the time the team starts to complete.
Estimate number: how complex the widget is
Number of people on each team
My question is this, which statistical methods could used to explore the data, any possible correlations etc given its not normally distributed, though I’ve muddled with functions, removing outliers etc to normalize it.
This ultimate goal would be a better forecasting model .
Note: this is a personal venture and not homework :)