“Check email two or three times a day, not constantly” is close to universal productivity advice. It is also advice most people quietly ignore, because the fear underneath it is real: what if the one email that actually mattered sat unread for four hours.
The advice is not wrong. The fear is not irrational either. The fix is not choosing one over the other, it is making batching safe.
Why batching works
Every time you check email outside a dedicated block, you pay a cost that has nothing to do with the email itself: the task-switching cost of returning to whatever you were doing before. Research on interruption cost puts the recovery time after a context switch at well over ten minutes for many kinds of focused work. Check email 15 times a day and you have paid that cost 15 times, regardless of how many of those checks contained anything worth acting on.
Batching to two or three fixed windows (morning, early afternoon, end of day, for example) cuts that cost to two or three payments instead of fifteen. The inbox does not get less full. Your day gets far less fragmented.
Why people abandon it
The problem is not batching itself. It is that most inboxes give you no way to tell, from outside the client, whether anything urgent has arrived. So the batcher either:
- Checks anyway, out of anxiety, defeating the purpose, or
- Sticks to the schedule and occasionally misses something that actually needed a same-hour reply.
Both outcomes are bad enough that a lot of people give up on batching entirely and go back to constant checking, which is worse on average but at least removes the anxiety.
What urgency scoring actually fixes
The tension exists because “batch less often” and “never miss something urgent” look like opposing goals. They stop being opposed the moment something can reliably tell you, without opening the client, whether the batch you are about to skip contains anything time-sensitive.
That is what urgency scoring is for. Every incoming thread gets evaluated for signals that correlate with real time pressure: is the sender someone you have a fast reply history with, does the language suggest a deadline or a blocker, is this a reply in an active back-and-forth versus a cold first message, has the sender already followed up once. STAMP runs this scoring on-device as mail arrives and separates the result into an urgent queue distinct from the general inbox.
With that in place, batching stops being a bet. You can check the urgent queue on a much tighter loop, even a glance every 30 to 60 minutes, without doing full triage, because that queue is short by design: usually a handful of threads, not fifty. The general inbox, everything that is not time-sensitive, stays batched to two or three real sessions a day.
The practical split
- Urgent queue: check briefly, often. A 15-second glance is enough. If it is empty, close the client and get back to work. This is not triage, it is a status check.
- General inbox: batch to two or three windows. Full triage happens here, following whatever structured routine you use (sort, quick replies, real work, snooze or archive the rest).
- Notifications: off, except for the urgent queue. If everything pings you, batching is impossible by definition. Let the urgent/not-urgent split do the filtering instead of your attention doing it in real time.
What this is not
Urgency scoring is not a promise that nothing will ever be missed. Occasionally something genuinely urgent will not read as urgent to a model, the same way it would not always read as urgent to a person skimming subject lines. The goal is reducing the odds from “whatever your anxiety decides” to a real system, not eliminating risk to zero. Nothing does that.
What to try this week
Turn off email notifications entirely except for one channel: a single daily or twice-daily check of anything flagged urgent. Batch everything else into two fixed windows. See if the anxiety that made you check constantly actually shows up, or if it turns out the urgent queue was doing its job the whole time.
For the wider argument on structure over willpower, see STAMP and what we changed on the changelog.
Batch with confidence, not anxiety. hello@stamp.email