Rather than my usual news-item commentary, I am going to write about something I’ve learned from data modelling and visualization.
The biggest expense for many employers is salaries & wages. This can be a tricky expense to model and visualize, because there is a lot of seasonality in the way payroll expenses are paid. I will give one example of a visualization that looked great until June or so — then started failing.
My graphs were focused on showing the trend of different pay types (overtime, maternity leave, etc.) month-over-month. What special kind of payroll would peak in summer? Vacation!
A full year of data would have showed that vacation time dwarfs most other pay types. The comparisons were much harder to see with this huge gap drowning out the scale. I needed to consider a better visualization to account for the abnormal build in vacation time versus other pay types.
Moral of the story? Think about where your data is going. Think about seasonality and what might come next before you get too confident in your visualization. Do a lot of backtesting, or fitting old data into your model or visualization.