Ian Johnson

Love this exploration, and it's a great example of how a lot of people could engage w/ data. And, there is motivation (saving money! helping the planet) for a homeowner to want to care about this type of data. Plus everybody understands the context / concepts (summer, energy usage, money, price, days, etc) so you can jump immediately into thinking about the data itself & feel connected to it.

what would be the canonical way to do this?

Promise.all(
  months.map(month => readxlsx(fa(`kwh-${month}.xlsx`))
                    .then(d => d.json())
                    .then(d => d.get("Sheet1")))
).then(d => d.flat())

Might be nice to show a chart w/ weather in Georgia to illustrate just how hot? I might try to throw something simple together w/ a threshold to specifically show "hot as hell" :).

Annoyingly, I needed to download 8 excel files: one per month for kwh, and one per month for dollars.

Since y-axis isn't labeled, maybe state in words what you changed & what you observed that led to the realization the price is rising? 

"To test this, let's divide by money by kilowatt-hour so that the y-axis shows "dollar per kilowatt-hour". If they correlate exactly, we'd expect to see a horizontal line.

Oh - interesting! The step change show us that the price is rising. This makes sense because I'm on a "default" plat where the summer gets more expensive."

(note - I liked that you connected what you saw in the chart (price is rising) with something in real world (summer gets more expensive). 

Side note: might want to multiply d.money / d.kwh by 100 so it's cents per kwh instead of dollar).

same as Zan, I thought summer was "more expensive" because consumption increased; maybe say "summer rates are more expensive"? (you have increase in consumption TIMES increase in price…)

What about using two shades of grey, one for weekday / one for weekend? Will give some hint at what's going on even with time of day missing from dataset. Likely you have different behavior on weekdays vs weekend. 

Also, are there any days that you know that you weren't at home (out of town)? Perhaps could highlight those manually, to see if they look different than other days?

Could also color the bars here based on the max temperature of the day. Or, show a second graph below this one with the same x axis but where the data is number of degrees above/below the standard "comfortable" temperature (72). 

You could also add a vertical grid line for the date in which you bought the dehumidifier. 

Do you charge your car daily or only on days after you use it a lot? 

On x-axis, perhaps only show a tick mark once a week so it's both easier to read and reinforces that sense of "week" which is inherent in the data. 

Could add Mike Freeman's tooltip: https://observablehq.com/@mkfreeman/plot-tooltip
import {addTooltips} from "@mkfreeman/plot-tooltip"
and then just add title: d => ` ${d.metric}` in the plot mark object. 

Um.... I feel like I'm just saying here is a bunch of work you should do ;). I'm brainstorming, so it's open-ended. Feel free to ignore it all. And, I'm also happy to pair on this or for me to fork it, to try out some of these out.

If you added enough solar to power the AC then switching to the day night plan would pay off.  Tesla power wall would also be great if you had the funds.

Maybe a costing/energy production notebook for solar and powerwall.

Brett Cooper

todo: drop the grid, and add a line for the median?

cool that you can see hints of diagonals in here, suggesting a non-weekly pattern that emerges for periods of time

3ish / 6ish day imperfect cycle that lasts for a little while and then restarts

note: left 3 / right 2 bars are shown because moving average is centered & doesn't hit the edges.

Can you assign this cell to a constant so I can import into the other notebook?

for now I copied the code, but would rather 'import with' so it stays synced

Purpose-built for displays of data