Cold weather and bike reservations

I was inspired by Tom at Seattle Bike Blog's analysis of the fremont bridge bike frequency dataset to look at my own web traffic data for SMBR. Also inspired by the fact that it snowed yesterday, and that cold weather made a very clear showing in website traffic. Above I've displayed the (normalized) daily site visits in red and the daily temperature high (in degrees F) in blue for the last month 2/22 to 3/22.

The results were slightly surprising -- I had thought that some of those weird spikes, like the drop on 3/12, were due to cold weather. But that one, for example, was not due to a cold day. But the decline in traffic 3/21 and 3/22 is clearly due to declining temperature (and rain, which for some reason almanac.com was always placing at precip = 0 even though my memory tells me that that wasn't true. Need a better historical weather data source). The most clear finding looks like temperature swings are significant and my traffic model is like this -- there's some baseline level of traffic, but a much warmer day bumps it up and a much colder day bumps it down. Smaller temperature changes have no effect.

I hypothesized that website views are sensitive to change in temperature, not just temperature -- if it was 50 one day and then 56, folks suddenly start thinking about biking. So I made that plot (the first day doesn't have data because the n-1 point doesn't exist in this dataset). Here, at least in the later March data, I'm seeing a small day-shift in web traffic. A down-trend in weather forecasts a decrease in web traffic the next day, but only if that temperature decrease is sizeable. I'll be able to prepare more data like this when traffic picks up even more. Of course weather variability will decrease in the summer, so the temperature delta data won't be super interesting.