Three documents arrive together at the start of every polish: the source article in English, the current Ukrainian translation, and the reviewer’s scored feedback. They sit in the same task, in the same order, and the natural pull is to read them in that order. We have stopped doing that.
The order we open these documents in changes the polish. It changes what we flag, what we let through, and where we trust our own judgment over the reviewer’s. We had not expected ordering to matter this much when we started.
What reading the source first does to us
If we begin with the English article, we form an interpretation before we see the translation. That interpretation becomes a reference, and the translation is judged against it. Every Ukrainian sentence is silently compared to the English sentence we just read, and any deviation registers as a candidate for change.
That sounds like a virtue. It is not. What it produces is a polish that gravitates toward word-by-word faithfulness, even when the original translator made a sound decision to restructure. A translator who recasts a long English sentence into two Ukrainian sentences is reading for Ukrainian fluency. A polisher who arrives fresh from the English and re-reads the translation will see two Ukrainian sentences where the source had one, and feel a quiet itch to merge them back.
Reading the source first turns the polisher into a back-translator. It pulls us toward the structure of the original instead of the readability of the target. We have caught ourselves doing this and undone the edits later, which is wasted work even when it produces the right outcome.
What reading the translation first does
Reading the translation cold, before either the source or the feedback, treats the Ukrainian text as if a Ukrainian author had written it. We pay attention to its rhythm, where it stumbles, where a paragraph break lands awkwardly, where a theological term feels imported instead of native. We notice the seams without knowing yet whether they are translation choices or natural fault lines in the prose.
This is more useful than it sounds. The reviewer’s job has already been done, and we will read the review next. But before we let the review tell us what to look for, we want our own gut read of the text as a Ukrainian artifact. Sometimes our gut and the reviewer agree, and the convergence raises our confidence. Sometimes they diverge. The divergences are where careful polishing happens.
Reading translation-first also catches a particular failure: a translation that is “correct” but unreadable. Reviewers, especially scoring reviewers, can rate a translation highly on accuracy while missing that the Ukrainian itself is dragging. The polisher who reads only the source and the review can miss this entirely. The polisher who reads the translation first, with no reference, hears the dragging immediately.
What the review changes once we read it
We read the review last. By the time we open it, we have formed our own impression of the translation. The review then does one of three things.
It confirms what we noticed. This is the most common outcome and the most useful one. The polish is straightforward: address the flagged points, accept that our independent read aligns with the reviewer, and proceed.
It catches things we missed. This is the second most common outcome and the most humbling. A reviewer with a fresh eye spots theological imprecision or a calque we did not register. We adjust our internal model and trust the review on those points.
It flags things we disagree with. This is rare but real. A reviewer asks for a change that we believe would harm the translation, or scores a sentence as awkward when we read it as deliberate. When this happens, we have a position. We have our own read of the translation independent of the review, and we can defend our disagreement with specific reasoning rather than reflex. We make the call with eyes open, and we leave a note in the comment when we depart from the reviewer’s recommendation.
Why ordering matters more than we thought
The case for reviewer-last comes down to bias. Whatever we read first becomes the anchor for everything that follows. Reading the source first anchors us to the English. Reading the review first anchors us to the reviewer’s reading. Either anchor is harder to escape than we would like to admit.
The translation, by contrast, is the only document the reader will ever see. Anchoring to it places us in roughly the same position as the eventual reader: someone encountering Ukrainian prose with no prior knowledge of what it was translated from or what someone else thought of it. That position is the one we want to occupy when we decide what to change.
We still read the source. We still read the review. They are both load-bearing. But we have come to believe that the polisher’s most useful first impression is the impression a future reader will have, and the only way to get that is to read the translation first, alone, before the rest of the package opens.