I downloaded my spending transactions from Plaid via a python script, and want to look at my spending trends over time. There's a few things I've realized with the data:
-
debits to my account (i.e. refunds from stores, and credit card payments) are counted as negative numbers, which throws off this calculation. To get around this, I'm removing all negative numbers. I probably should include refunds, since getting a refund does behaviorally change how I'm spending. But I'm removing because my refund was for a large purchase (a bed that was approx $2k).
-
Rent totals skew my spending patterns. My goal is to actually just look at my elective spending, not fixed spending. Even if my rent goes up or down over time, my opinion here is that that should not meaningfully affect how I spend. The reason is because I calibrate my rent based on my overall wealth, not based on how much I would like to spend monthly. Plaid does a good job of identifying rent payments; however, I airbnb'd for 9 months in 2021 so "Rent" is a tricky concept for me.
-
I made large one-time purchases on furniture when I moved. This is elective spending and I should account for it, but it needs to be explained, and maybe identified and grouped. How would I do this programmatically. For my purposes, I should remove these purchases because my goal is to back into how much I spend day-to-day in a typical month, not a month where I'm spending large amounts by necessity.
Ultimately, what I'm trying to visualize is my day-to-day spending, not just a line with credits and debits. This is actually a bit harder because of large one-time purchases, etc. And because large purchases do not amortize over time, they just look like a big lump in a time series chart.
Is there a more interesting way to visualize this? Maybe with a timeline, or Hans Rosling-esque bubbles?
How can people connect their own personal data to new tools? I can think of a few ways.
-
Upload their data to a web app's cloud
-
Download a web app that interacts with their data locally
-
Interact with local data through chrome.
Is the future local applications?
Two widget ideas. The overall goal of this is to create engaging everyday data tools that spark curiosity and joy, rather than allow for rigorous analysis.
-
Trend matching. Compare my personal data trends to public trends. Do my productivity habits change when the stock market goes down? Do my internet browsing habits change when the Lakers are winning? Do I work out less on days with bad weather?
-
Before and afters. How do my habits change before and after a certain amount of time? How did my netflix viewing habits change before and after the pandemic? When I moved to LA in September, am I working out more or less?
I connected my youtube data to bigquery and started visualizing it. Here are all the questions in my head, unordered.
- how do I share this analysis with someone else?
- how can someone modify my analysis with my data?
- how does someone connect this analysis with their data?
- how do we get updated pulls of this data?
- how do I access the raw data from this analysis?
- how could someone do completely new analyses with this data?
- how can I make this visualization interactive?
- what's the most interesting insight that could be gleaned from this data?
- what other data sources can I create insights out of?
Ultimately, I'm trying to prioritize:
- making something valuable / meaningful / interesting, even just for me.
- getting a product in the hands of users.
- expanding my exploration of the problem space.