Hi Andrew, this is an excellent first attempt at using data science to shape business strategy. In these types of scenarios with limited data, I look for business value by first eliminating what I’d expect to see from the data, and then discussing things I wouldn’t normally expect.
For instance, I’d expect demand to be high
- on workdays
- among loyal customers (read ‘registered users’), and
- when weather conditions are favorable.
I’d also expect to see reduced demand with
- high humid days
- holidays, and
- days with bad weather conditions.
When there are spikes in demand, I’d expect that things return to normal immediately. (Hint: it doesn’t always work that way)
For this challenge (with its detailed weather information), I think the business strategy comes from the question:
“When should supplies be dynamically adjusted?”
In this context, “supplies” refers to both bikes and types of customers that ride.
In answering this question, I love the data science work that you’ve begun. To find insights I first looked at the Correlation Matrix for Bike Rental Demand. I focused on any unexpected behaviors in the last 3 rows as they represent things the business either can change or reasonably apply effort to change: count, registered, and casual. Here are some of my observations
- when temperatures are favorable, less loyal customers (read ‘casual riders’) show better demand [Customer Acquisition Impact]
- ‘Casual riders’ prefer holidays and avoid workdays, while ‘registered riders’ prefer workdays and avoid riding on holidays [Marketing & Promotions Impact]
- ‘Casual riders’ give more thought to how humid it is, when they decide to use the service [Marketing & Promotions Impact]
In examining the graph of rental fluctuations around holidays I saw some interesting anomalies:
- Before the dip in demand on Independence day, there’s a noticeable surge in demand in both 2011 and 2012 [Marketing & Promotions Impact]
- There are two dips in the first weeks of 2012. Could one of these dips have been caused by severe weather conditions? [Inventory Impact]
- There’s a noticeable difference between rest-focused holidays and more commercialized holidays [Marketing & Promotions Impact]
So as I said at the start, this is a good first attempt at discovering business strategy from data science.