Lab · Part 1 of 2 · Research · Nov 2026
208,274 unread
Before I changed a single email, I studied my inbox the way I’d study a user: read-only, by the numbers, without judgement. The research killed my original goal, inbox zero, and replaced it with a brief that set the direction for everything I built next.
The 30-second version
7 min read
Role
Researcher, with myself as the user
Sample
227,002 conversations, plus a 7-day inflow count
Method
Read-only Gmail search, label counts, receipts
Timeline
2–3 Nov 2026, follow-up 8 Nov
I compared what I say I want (subscriptions) with what I actually do (opens, replies, purchases). The gap was the finding: the mail that arrives most is the mail I open least, and only about 1 in 100 emails needs me. That turned “get to zero” into “decide what the inbox is for.”
92%
of all mail never opened (208,274 conversations)
1.0%
of a week’s mail came from people or about jobs
1 brief
that replaced “inbox zero” and set the design direction
The question
“How do I empty this?” was the wrong research question
Most inbox advice starts with tactics: filters, unsubscribing, zero. I didn’t trust any of them, because I didn’t know what problem I had. An inbox with 227,002 conversations isn’t a sorting problem. It’s a record of thousands of small yeses, and I wanted to understand them before undoing them.
So I reframed the question the way I would on a product team: what is this inbox for, according to my own behaviour? Behaviour, not intention, because intention is what filled the inbox in the first place.
Method
Four lenses, all read-only
Nothing was moved, labelled or deleted during the research. Every number below is a Gmail search or a label count.
Lens | Data | The question it answers |
|---|---|---|
Volume | Conversations, unread, 30-day and 7-day inflow | How much arrives, and does it ever leave? |
Source | Senders grouped by type and purpose | Who am I saying yes to? |
Revealed preference | Opened vs unopened, per category | What do I actually choose to read? |
Action | Sent mail, drafts, purchases after marketing | What does the inbox make me do? |
Finding 1
Almost all of my email life was input
On 2 Nov 2026 | Count | What it means |
|---|---|---|
Conversations in the inbox | 227,002 | Years of saying yes, never undone |
Unread, all mail | 208,274 | 92% collected, never opened |
Received in 30 days | ~4,700 | About 155 a day |
Sent in 30 days | 24 | Under one a day |
Received for every email sent | ~196 | The soldier’s ratio |
A week of exact counts later confirmed the shape. Of 1,006 threads that arrived from 1 to 7 November, 10 needed me: 7 about jobs, 3 from people. Add receipts and bookings and it’s 44, or 4.4%. Everything else was publishing, teaching or selling.
Finding 2
Six people I keep saying yes to
Grouping senders by purpose, not by Gmail tab, described six versions of me. None of them is wrong. Together they buried the one message that mattered during a job search: a reply from a recruiter.
Who | What they subscribe to | Lands in |
|---|---|---|
The builder | AI and engineering news, product launches | Newsletters, The Information |
The founder-curious | Startup newsletters, investor updates, pitch events | Newsletters, Events |
The lifelong student | Course platforms, bootcamps, webinars | Courses |
The designer and maker | Tools, type foundries, maker communities | Newsletters |
The city regular | Luma, Partiful, venues, meetups | Events |
The shopper with taste | Fashion and home brands | Unwanted, Saved (receipts) |
Qualitative grouping from sender names and subjects. The folder names on the right are the ones the system later used.
Finding 3
What I open is the inverse of what I receive
Subscriptions are what I say I want. Read rates are what I actually choose. Gmail’s own unread counts per category gave me a revealed preference I couldn’t argue with.
I open 6 in 10 job emails and 1 in 50 newsletters. The publication I pay for is the least-read thing in the account: 16 of 1,338 issues. It sends 8 emails a day.
Finding 4
The aspiration gap
Multiply volume by read rate and a third pattern appears: the distance between who I want to become and what I have time for.
Category | Per week | Opened | What the gap says |
|---|---|---|---|
Courses | 180 | 4% | 26 invitations a day to become someone new |
Newsletters | 324 | 2% | The builder in me subscribes faster than I read |
The Information | 56 | 1% | Paying to feel informed, not to be informed |
Jobs + People | 10 | 46–62% | The only mail that changes my week |
The subscriptions describe who I want to become. The inbox had turned them into a pile of things I hadn’t done yet.
Finding 5
Persuasion works on me, and the receipts prove it
Every marketing email in the hero image uses a named tactic: scarcity, a deadline, an abandoned cart, flattery, desire. I wanted to know whether they worked, so I matched marketing emails against order confirmations.
Brand | Marketing | Orders | Pattern |
|---|---|---|---|
Quince | 63 in 90 days | 2 | Pressure: one order nine days after “Complete your order before we sell out!” |
Aritzia | 4 in 90 days | 18 in a year | Habit: 7 orders in 23 days over the 2025 holiday sales |
One brand runs on pressure, the other on habit. Both run through the inbox, which means the inbox isn’t neutral storage. It’s a channel other people use to set my priorities.
Finding 6
The maker stalls under volume
Across the account there are 1,943 sent conversations and 302 drafts that were never sent: about 1 in 7 emails I started. In the 30 days to 8 November I sent 21. I don’t lack things to say. I lose them under everything arriving.
The decision
What the research killed, and what it set
Killed
Inbox zero as the goal. An empty inbox that refills to 155 by morning just makes me a faster soldier.
“Unsubscribe from everything.” Six real interests sat behind the volume. Deleting them would delete parts of me.
AI triage first. With 1% signal, one wrong guess on a recruiter email costs more than a thousand right ones save.
Set
One brief: my inbox is for people, opportunities and records. Everything else has to earn its place.
One metric: signal share, the % of daily mail that needs me. 1% today; 25% is the target.
Two constraints: nothing irreversible, and the inbox has to make sense on my phone.
My inbox is for people, opportunities and records. Everything else has to earn its place.
The design brief, 3 Nov 2026
Part 2 is what I built from that brief: Untangling 227,002 emails →
Try it
Six splits to run on your own inbox
None of these need code. Each is a Gmail search and a decision, and together they move you from soldier to captain.
Soldier | Captain |
|---|---|
Sorts by sender or tab | Split by intent: act, record, consume or being sold to. Only “act” stays in front of you. |
Keeps every subscription “just in case” | Read-rate audit: anything opened less than 5% of the time in 90 days gets unsubscribed. |
Measures how empty the inbox is | Signal share: the % of daily mail from people or about decisions. Aim for 25%. |
Lets volume decide the day | Inflow budget: under 30 new emails a day. Every new sign-up replaces one. |
Signs up for the next course | Close the aspiration gap: one course and one paid publication at a time, read in one weekly block. |
Answers whatever is newest | Out versus in: move toward 1 sent for every 20 received. Finish or delete the drafts. |
Limits
What this study can’t tell you
n = 1. This is one person’s inbox. The method transfers; the numbers don’t.
“Opened” is a proxy. Gmail marks mail read when I preview it, and I sometimes mark mail read without reading it. Real read rates are probably lower.
Mixed windows. The 30-day figures are Gmail estimates, rounded. The weekly inflow is exact but was counted after sorting began, so categories follow the system’s folders.
Reflection
The data was kinder to me than I was
Going in, I felt like a hoarder, paying for storage to keep mail I never read. The data described someone else: curious, entrepreneurial, saying yes to new ideas, courses, events and cities. Shame had kept me avoiding the inbox. Understanding it got me to act.
The principle I’m taking into product work: research earns its keep when it kills a goal, not when it confirms one. Inbox zero was a fine goal. It just wasn’t mine.