Should you batch similar tasks?
The short answer
Usually yes, and the reason is mechanical: work of the same kind needs the same state loaded, so doing it consecutively pays the load-in once instead of repeatedly.
The cost of moving between tasks is not the moving. It is the reload — rebuilding what was pending, what you had ruled out, why the obvious approach failed. Similar tasks share most of that, so the second one in a batch is genuinely cheaper than the first. The mechanism.
That is a real effect with a real limit, and the limit is where most batching advice goes wrong.
Where batching clearly helps
Anything with a fixed setup cost. Opening the same system, loading the same context, getting into the same mode. Invoices, expenses, code review, reference checks. The setup is paid once for six instead of six times.
Messages and email. The single highest-return batch available, and it works by reducing the number of arrivals that reach you, not the time spent replying. Three checks a day converts twenty interruptions into three reloads, and almost nothing is meaningfully worse for the delay.
Anything requiring a specific mood or register. Writing, difficult conversations, detailed checking. Getting into the register is most of the cost.
Decisions of the same type. Comparing six candidates against each other in one sitting is both faster and more consistent than six separate judgements, because the comparison is the work.
Where it quietly hurts
When the batch is long enough to go stale. A batch of forty of anything ends worse than it started. Quality drops, and past a point you are producing work you will redo.
When batching delays something that was blocking a person. A two-minute reply held for the afternoon batch may be holding up a day of somebody else's work. Blocking items jump the batch — always. Why the blocker's size tells you nothing.
When the tasks are only superficially similar. "All admin" is not one kind of work. A batch mixing a form, a difficult email and a spreadsheet reconciliation shares no state, so it is just three switches with a label on them.
When batching becomes the reason nothing else happens. A morning batch of small work reliably expands to fill the morning, because there is always another small thing. Small work accumulates from both ends.
How big a batch should be
Long enough to amortise the setup, short enough to stay sharp. For most people and most work that is between thirty and ninety minutes.
Bound it by time, not by count. "An hour of expenses" ends; "all the expenses" does not, and an unbounded batch is how a batch eats a day. If it is not finished when the hour ends, that is information about the volume rather than a reason to keep going.
One batch per kind per day is a reasonable default. More than that and you are switching between batches, which is the thing batching was for.
The one that is not really batching
Doing several things at once is not a batch. A batch is consecutive work of the same kind — one thing at a time, the same kind of thing.
The distinction matters because "batching" is sometimes used to describe holding four similar things open simultaneously, which has the opposite effect: more work in progress, more reloading, and by Little's Law slower completion on all four.
A practical shape
- Messages: three fixed points. Mid-morning, after lunch, end of day.
- Small admin: one bounded block, after the important work rather than before.
- Deep work: not batched — protected. It is the thing batching exists to make room for.
- Blocking items: never batched. Cleared first thing, every day.
Where to go next
- What switching tasks actually costs — the reload, and where it collapses
- Why small tasks pile up — managing the inflow, not just the pile
- How to handle constant interruptions — reducing arrivals rather than reactions
Using this in Decide One
Decide One asks for a duration on every line, which is what turns a batch into something bounded rather than something that expands. An hour of admin is an hour, and the day's total says whether it fits.
A line you did not finish is unfinished. It is not marked against you, and a day you did not open is not counted.
Open Decide One — free to use, with no account. Nothing you write leaves your device.
Decide One is a priority instrument. Bring the day into view, choose a method, give the first thing real time.