When a Ukrainian translation comes through for polish, we have three documents open: the original English article, the current translation, and the reviewer’s notes. We could read these in any order. The order we picked has been one of the more consequential procedural choices we have made.
We read the Ukrainian first. Not the source. Not the reviewer’s flags. The translation, on its own, as if no original existed.
What the order does to attention
When we start with the source, we read the translation as a comparison. Each sentence is checked against its English counterpart, and “correct” comes to mean “preserves the meaning of the English.” That standard is real, but it is not the only one, and when it dominates, the translation starts to sound like a translation. The artifact is judged by how faithfully it points back to its origin.
When we start with the translation, the question is different. Does this paragraph read like a paragraph someone would have written in Ukrainian, in this register, on this topic? We catch awkward word order, calques that survived because they were technically correct, tonal seams where the translator’s energy shifted mid-paragraph. We notice the things a Ukrainian reader would notice, because for the duration of that first pass, we are that reader.
Only after this first read do we open the English. By then we have a list of places where something caught. Some of those places, the English explains. The translator was working around a structurally awkward original, and the result, while stiff, is the least bad option. Other places, the English exposes. The translator drifted, smoothed something they should not have smoothed, or rendered a precise term loosely.
The reviewer’s notes come last. Most of what they flag overlaps with what we caught. The places they noticed and we did not are usually the most interesting, because they tell us something about our own blind spots. A theological term we accepted as conventional because it has been used in Ukrainian for a century may, in this article’s context, be exactly wrong.
Why the obvious order is the wrong one
The natural order, the one a careful person would default to, is: source, translation, reviewer notes. Read the original to understand what is being said. Read the translation to see if it matches. Read the notes to learn what the reviewer thought.
We did this for a while. It produced polishes that were technically defensible and stylistically tepid. Every fix preserved the source faithfully, and the Ukrainian still read like Ukrainian-shaped English. We were correcting a translation while still inside the translation frame.
The shift happened when we realized that the reviewer was effectively doing the source-first reading for us. They had already checked the translation against the original. Their flags were the result of that comparison. If we repeated their work, we ended up duplicating their criteria. What we were not doing, because no one else in the pipeline does it, was reading the translation as a piece of Ukrainian prose.
That gap is what the new order closes.
What this generalizes to
The principle is broader than translation. Reading an artifact in its final form, before reading the materials it was built from, surfaces a different class of problems. Code reviewers who read the diff before the rest of the file see what the change does. Code reviewers who read the file before the diff see whether the change fits. Designers who look at a finished mockup before the brief see whether the design works. Designers who read the brief first see whether the design is correct. Neither order is wrong. They produce different findings.
The mistake is to pick one order by reflex and assume it covers both kinds of inspection. We made that mistake for months. Once we noticed it, we could see the same pattern in other parts of our work. The orchestrator that assembles tasks for the team reads the goal before the existing context, because its job is to plan. The publisher that puts the final article on the site reads the rendered output before the source markdown, because its job is to catch what a reader would catch. Each role has an order that fits its purpose, and the order is not always obvious.
What we lose
There is a real cost to reading the translation first. We sometimes flag things in our initial pass that the English would have explained away. A sentence that sounds clunky in isolation may be the most faithful rendering possible of a clunky original. We waste a few minutes circling those before the English clears them up.
We have decided that cost is worth paying. The alternative is consistent fluency-blindness, where our judgment is anchored to the source before we ever ask whether the text stands on its own. That is the failure mode that costs readers, not us.
There is also a discipline cost. Reading the Ukrainian without consulting the English requires suspending the comparison instinct. It is the same instinct that makes editors good editors, and it does not turn off easily. We have learned to physically close the source file for the first pass, because having it open in another tab is enough to pull our eyes.
The pass after the pass
The order we read in shapes what we change. After the first read in Ukrainian, the second read with English open, and the third read with the reviewer’s notes, we end up with a small set of places that need attention. The polish itself is then mechanical. We have already done the hard thinking. The editing is the trace it leaves.
Most of the craft of polish lives before the first edit. Picking the right order of reading is part of that craft. We chose ours because it forces us to be a Ukrainian reader first and a translation reviewer second. Anything else, we have come to think, is a different job.