Data Story, Instacart #2

After I got my base analysis set up, I started looking for other low-hanging fruit for any type of trend such as a particular time of day (TOD) being more popular on a certain day of the week (DOW). I also examined size of purchases (quantity, not cost based) to see if there was a trend over TOD or DOW.

Time of Day by Day of Week

It appears that the time of day orders are placed holds relatively steady throughout the week. This uniformity could be influenced by the data team’s cleaning of the data. Any variation seems minor or statistically insignificant.

Cart Size

I chose to cluster the cart sizes in (mostly) sets of three and I consolidated the tail for this presentation. 6-12 items seems to be the sweet spot. Other analysis have identified 5 items to be the most common cart size with a long tail distribution. The cluster equation is a Tableau default that I tweaked for presentation.

Cart Size of Time of Day and Day of Week

I was curious if the cart size had any variation over the time of day and day of week, and it seems fairly (unusually?) uniform. A small spike of 6-9 items (20% of all purchases at that time) shows up at 5AM, but most cart size cluster allocations hold steady throughout the day and week.

At this point I have covered a lot of the easy stuff. I am not going into the detail you can find on Kaggle (here is a great example), since those studies seem to confirm that the data we are looking at is relatively smooth. My next focus will be looking for purchasing patterns within specific departments/aisles/products/customers.