Recommendation of Dispense Amount Based on Historical Averages

Question

I am working on creating a function to provide recommendations for the amount to dispense. I have historical data on the average daily dispense amount for each item, but I lack information on the number of dispenses made in one day. Given this limitation, how can I make an accurate recommendation? For my school project, I can assume there is an average of 4 dispenses a day and then track them to provide a recommendation based on the average divided by 4. Are there alternative approaches to address this issue effectively?

Answer 1

This kind of forecasting problem is not pure mathematics, To make a useful prediction for how much to stock one of the first things you need is the cost of overestimating and the cost of underestimating. If overstocking is free, just get lots and lots.

Then you need some information about the variability in the daily numbers.If they are pretty steady then using the average as a predictor will be nearly right often. If the daily numbers swing wildly then many days will be far from the average. (That's when the costs of errors either way matter most.)

In any case, a useful report in a real life situation would contain a range of predictions, with an estimate of the probabilities and costs for each one.

This is much more than you can do in a school project but your project will be better if you acknowledge these issues. You could at least provide a range of predictions by calculating with $3$, $3.5$ and $4$ dispenses per day.