An article may require two days of work without being deliverable two days from now. Waiting for sources, review and corrections also affects the calendar. That is the first distinction I make when estimating.
Break the assignment into meaningful stages
Research, drafting, editing, verification and delivery involve different demands. Rather than assign minutes to every sentence, I identify what each stage needs. Unfamiliar work should carry an explicit uncertainty note.
A previous assignment provides a reference, not an automatic estimate. Source readiness, language count and the number of decision-makers can change the conditions substantially.
Give waiting time an owner
Awaiting information or approval is not active production, but it affects delivery. I record the required input and the latest useful date. The consequence of a late response should be discussed before it occurs.
Imagine a page due on Friday whose verified copy arrives on Thursday evening. The remaining design work cannot be assumed to happen instantly. Empty calendar space is not the same as available production capacity.
Use conditions rather than false precision
Where uncertainty is substantial, a conditional range may be more honest than one exact day. The range should expose known variables, not disguise an unconsidered guess.
An intermediate review can also help reveal a mistaken assumption early. Giving revision rounds a purpose makes those checkpoints more useful.
Learn from the difference
After completion, I compare the estimate with actual work. Was the difference caused by additional production, waiting or changed scope? Recording only that it took longer teaches little.
A useful estimate is not a perfect prediction. It makes clear which assumptions support the commitment and what needs revisiting when those assumptions change.