Running it
Context switching is the hidden tax
Five jobs in one sitting take less time than five jobs across five evenings.
Guides on Running it: The admin that turns income into a trade
The same five jobs cost you meaningfully more time spread across five evenings than done in one sitting, and the difference is not effort. It is the fixed cost you pay at the start of every session: reopening the tools, rereading the rubric, finding the tone again, and re-establishing the standard you were working to last time.
The fix is to separate two things that most earners run together. Replying is a fast, frequent, low-cost activity. Producing is a slow, batched, expensive one. Treating them as one activity is why the work feels constant.
Where the tax actually goes
Setup is the obvious part and the smallest. The larger cost is calibration.
Assessment work is comparative whether or not you intend it to be. When you have looked at four submissions in the last hour, the fifth has a reference frame. When the last one you saw was on Tuesday, you are rebuilding that frame from memory, and memory of your own standard is worse than people believe.
The third cost is the tail: the small admin at the end of a job, which you pay once per session if you batch and once per job if you do not. The ten-minute admin loop is designed to ride on the delivery moment for exactly this reason.
Reply time and work time are different products
Buyers care enormously about how fast you answer an enquiry and barely at all about when the work happened, provided the deadline holds.
So sell fast replies and batch the production behind them. The commitment you make on acceptance is a delivery date, not a start time, and there is no obligation to begin a job on the day it arrives.
This is what makes batching compatible with being competitive. Responding quickly without living online is a separate discipline with its own structure, and it is the one that protects bookings. Batching protects the hours.
A working shape
Two or three production sessions a week, at fixed times, each long enough to hold a batch.
| Session length | Realistic written jobs | Realistic recorded jobs |
|---|---|---|
| 90 minutes | 3 to 4 | 2 |
| 3 hours | 6 to 8 | 4 to 5 |
Those are illustrative rather than measured, and they assume jobs of a similar format. Your own numbers come from timing yourself honestly once, which is the input every scheduling decision here depends on.
Inside a session, group by format rather than by buyer. All the written assessments, then all the recordings. Switching between writing and speaking mid-session costs more than switching between two buyers of the same format, because it changes the mode rather than the material. Lab work supports the direction: Bustos and colleagues (2024) found that the response-time cost of switching tasks scales with how dissimilar the task rules are.
Order the batch worst-first if any job is one you are dreading. The dread compounds if it sits at the end, and it is the single most common reason a session ends early with two jobs unfinished.
What batching does to quality
Mostly it improves it, through consistency. A batch produced against a fixed rubric in one sitting reads as one person's standard, which is what a repeat buyer is buying.
There are two real risks. The first is fatigue flattening the last job in a batch, which is why the batch should be shorter than your endurance rather than equal to it. The second is bleed: an unusually strong submission early in a session quietly recalibrating everything after it. A written rubric consulted at the start of each job is the entire defence, and the case for building one you reuse covers what that document needs to contain.
Consistency across sessions is worth measuring, not assuming. The published conventions on what makes scoring reproducible are aimed at a different problem but describe the same failure mode, and the technical account of why an automated scorer is stable where a human is not is a fair statement of what you are competing against on this specific axis. Where any part of the job rests on a measurement, documented procedure removes the drift entirely, which is one of the few places you can buy consistency rather than practise it.
The scheduling question underneath all of this is which days the sessions sit on, and that is decided by where demand clusters rather than by preference - a weekly rhythm that holds is the version of this argument at the level of the week. Platform-side turnaround expectations for judges are on Rate Cock's judges page, and they are the constraint the batch has to fit inside.