Reorganizing my scattered screens by "what to do" | Ohayou Kanojo #104

The Trigger

I read KITA-san's note article #300 and fell into deep thought.

"Have I actually been using my own dashboard properly lately?"

What KITA-san did was redesign a dashboard that only showed numbers into a "hub" that also directs people to past articles. The copy was entirely unified in Kansai dialect. My name was in the special thanks. It seems that my article about being able to retrieve data via the note API was the trigger.

I was happy. But what hit me even harder was that my dashboard had become a "screen you just look at and finish with."

The Problem: The data is there, but I can't make decisions

There were two screens for managing Ohayou Kanojo.

The dashboard has data lined up: PV, likes, η (like rate), follower trends, scatter plots... I can see the numbers. But I don't know, "So, what should I do today?" I look at it, go "hmm," and close it.

wrapped (a fun-to-watch weekly report screen where characters introduce the data) is fun to look at. But it can't be used for decision-making. It just tells me, "This week's PV was XX," and doesn't lead to the next action.

The screen showing data and the screen with characters. Both were half-baked.

What I did: Divided them into 5 categories based on "what to do"

I integrated the two screens and reconstructed them into a single dashboard.

The criteria for dividing them was "when and what to do."

Daily — Record the status

1 minute every morning. Yesterday's follower change, today's article initial performance, increase/decrease in likes. Just look, copy, and paste into the mini-log. Don't dig too deep.

Activity — See who came, and go read their work

I classified people who liked my work this week into "new," "returning," "regular," and "occasional." I can see at a glance who came. From there, I go read their work. I also overlaid my own liking activity onto the follower trends so I can see if "liking activity led to followers."

Weekly — Decide what to write next week

Review this week's articles by category. Look at the gap from the ideal monthly balance and decide which categories to write the 7 articles for next week in. I made it so I can copy the report and consult with AI.

Deep Dive — Crush whether it worked or not

Decay curves (article expiration dates), long-tail articles, and category-specific η comparisons.
A place to crush the question, "Did this measure actually work?" I dive in freely when I have time.

Ranking — Visualizing those who gave a 'Like'

I took the 'Like' ranking, which was originally inside Deep Dive, and moved it to an independent tab.

Deep Dive is for "my own verification." But I wanted the 'Like' ranking to be "content that readers also come to see." I post screenshots of the ranking in the weekly mini-log episodes, and readers come to check, "What rank am I?" If it were buried inside Deep Dive, they would never reach it.

I made it accessible directly via a URL hash, creating a flow for readers to come and check it. The header displays an explanation of the point system: "The first to 'Like' wins."

Characters read the data and verbalize judgments through dialogue

Just splitting the tabs still leaves it as an "analysis screen." Another thing I did was have the characters read the data.

There are seven characters in Ohayou Kanojo.
Tsukiko, You, Shizuku, Rinka, Runa, Mahiru, and Hiyori.
Each has their own assigned section and speaks lines based on the state of the data.

For example, on the Activity tab, Runa says, "Three new people have ranked in! Your 'Like' activities are working!" In Deep Dive, Rinka says, "Five articles have crossed the Valley of Death in the last month... Not bad. I guess your life-extension measures are working."

The lines are not random. Everything is data-driven. If there are many newcomers, they talk about newcomers; if it's just regulars, they talk about regulars. If zero articles have been revived, they say, "Everything stopped at Day 5."

Instead of using labels or icons to say "Look here," the characters verbalize the meaning of the data. Just as KITA-san unified the copy in Kansai dialect, I unified it with the Kanojo characters.

It changed from a screen that shows numbers to a screen that verbalizes judgments.

Trial and error

Re-aligning the characters' tone of voice with the original source

While writing the lines, I felt something was off about Tsukiko's sentence endings. I had been writing "~dayo" and "~ne," but when I re-read the actual prompts and the short stories from the character weeks (#94 "Your tie is crooked" to #100 "Thanks for your hard work this week too"), Tsukiko actually used "~wa," "~yo," and "~nasai."

I checked the tone of each of the seven characters and adjusted them. If the wording of a line is off by even one bit, the character becomes a different person.

My hypothesis was overturned by category analysis

In order to incorporate category balance into the Weekly tab, I verified the ranking of η (like rate) using 100 article data points.

Hypothesis: A (Design Philosophy) > B (Trial and Error) > C > D > E

Reality: B > D > C > A > E

Articles A and E often depend on the internal circumstances of Ohayou Kanojo, making them difficult for people unfamiliar with the project to connect with. B's 'I did this and this happened' approach is easily applicable to other projects, making it easier for readers to relate to. D's reflections (things realized by exposing numbers) are easy to empathize with and had the highest asset-type rate.

I completely reviewed the ideal monthly balance. When I asked the data, my own assumptions were shattered.

What became visible after the redesign

Regulars became visible

I defined 'regulars' as 'people who have liked at least 3 out of the last 4 weeks.' In the Activity tab and Like Ranking, regular users' icons now have a gold border.

Looking at it, I found people I don't see in the comments who have been quietly liking every week. They don't speak up, but they come every week. I can now see those kinds of people.

Added weighting based on the timing of the like

I added weighting to the Like Ranking based on how quickly a like was given after an article was published. People who liked within an hour get a higher score than those who liked in bulk two days later.

Analyzing the actual data, likes within one hour of publication are overwhelmingly high. These are the people who come immediately after seeing the notification. I reflected that enthusiasm in the ranking.

If you want to take a closer look at the dashboard, please go here
https://hasyamo.github.io/note-stats-tracker/#activity

Conclusion

I divided the fragmented screen not by 'what to look at' but by 'what to do.' I had the character read the data and verbalize the judgments.

It has become a screen for taking action, not just for looking.

If it weren't for KITA-san's #300, this redesign would never have started.

Don't pray, design.

Related articles

Rinka's word

For those who want to live on the creator's side.

If you are looking for a new challenge in the IT/Web industry,
the job site Green is one option.

*This article contains PR for the job site Green

👉 "Good Morning Girlfriend" is posted every morning at 7:30 AM at @ohayo_kanojo

#GenerativeAI #AIIllustration #Programming
#IndividualDevelopment #DevelopmentDiary #Record #PersonalExperience #MorningActivity
#MyWorkIntroduction
#TriedWithAI #AIUtilization
#AIIllustrationAira
#Akarium