Is this post-editing or a rewrite? Quote the work before you accept the discount.

The client asks for a discount before anyone has inspected the draft

The enquiry sounds simple: the translation is already done, so your team only needs to check it. The draft reads smoothly. Then a linguist compares it with the source and finds inconsistent terms, missing qualifiers and passages that need rewriting. The deadline and discount have already been agreed.

Now the PM must choose between absorbing the extra effort, renegotiating during production or asking the linguist to work faster. None of those decisions can repair an estimate based on the assumption that machine output is always cheap to finish.

In a discussion about accepting AI post-editing assignments, translators describe sharply different conditions under which the work makes sense. Another post about client price pressure describes demonstrating the required work on sample pages. These accounts motivate a practical intake test, not a universal rule against post-editing.

The question for your LSP is whether this particular workflow reaches this client's acceptance criteria more efficiently than a fresh translation. Establish that before treating a draft as a saving.

Define the output quality before measuring speed

Ask what the content must do, who will read it and how the client will approve it. A support article, a brand campaign and a safety instruction can require different expertise and review. Agree the reference terminology, style expectations, context and final deliverable.

Avoid using light and full post-editing as self-explanatory labels. Write down which defects must be corrected and whether independent revision, layout checks, terminology research and explanatory feedback belong in the quote. The same label can hide different expectations.

Confirm the provenance of translation-memory matches too. A high match score describes similarity to stored material; it does not independently prove that the stored target was reviewed or remains appropriate for this context. Ask what evidence supports any reduced review scope.

Compare post-editing and retranslation without making the test misleading

Choose a mixed sample covering the actual subject matter, long and short segments, terminology density, tables and awkward source passages. Separate materially different language pairs and content types. Record why each part of the sample belongs.

Have appropriately qualified linguists work to the same acceptance criteria with equivalent context and resources. Avoid having someone post-edit a passage and then immediately translate that identical passage from scratch. Their earlier reading will influence the second timing.

Use matched portions or counterbalance the order across participants where practical. Record the limitations if a small team cannot control every variable. A pilot estimates your workflow; it does not need to pretend to be a universal benchmark.

• Measure source reading and terminology research.
• Measure target correction or fresh drafting.
• Include the planned independent review and final checks.
• Record file preparation, queries and required reporting.
• Note interruptions, uncertainty and passages that exceed the proposed scope.

A pilot can reveal that one method works for routine sections while another suits a difficult section. If you split the workflow, name the boundary and include the coordination cost. Do not assume that the linguist can switch methods without extra handling.

Show the cost of each route, including the hidden work

For each route, multiply the observed role-specific hours by your actual agreed costs. Add the actual platform, preparation and handling costs. Apply your business's chosen pricing policy explicitly; a desired margin and a markup on cost are different calculations.

Do not extrapolate the sample mechanically across a heterogeneous book or multilingual programme. Estimate separately where languages, genres, draft provenance or requirements differ. Give the client a range when the remaining input is uncertain.

If a fixed per-word quote is commercially useful, derive it from that effort estimate and define its assumptions. If the client cannot provide representative input, consider a paid assessment or an hourly cap before a fixed commitment.

A small edit can require a large amount of checking

A reviewer might spend several minutes checking a single number or a subtle qualifier, then make a one-character change. Another segment may need many mechanical changes that are straightforward. Edit counts describe the result of editing, not the full process.

The original research paper The Trials and Tribulations of Predicting Machine Translation Post Editing Productivity explicitly distinguishes edit distance from cognitive effort and time. Its study reports useful relationships in its data, but that does not make a universal tariff or predict your next language pair.

Use edit distance alongside time, observed defects and final review outcomes. If someone changes very little, inspect whether the input was good or whether a difficult issue went unnoticed. If they change a lot, inspect whether those edits were necessary under the brief.

This checkpoint protects both sides. The client can decide whether the additional quality requirement is needed, and the LSP can describe the work before it becomes an invoice dispute. Record the agreed threshold in the quote rather than inventing it halfway through.

Give account managers a repeatable intake checklist

Store the pilot result with the quote so production can see how the estimate was made. Keep language, domain, method and required output attached to the timing. A reusable rate card should retain those boundaries instead of averaging away the difficult work.

If your enquiry process cannot capture these inputs, use a discovery call to map one quote from intake to delivery. A small assessment form or integration may remove repeated handoffs. The aim is a quote that production can execute and the client can understand.