Phrase Launches Managed Outcomes: Is Localization Really Moving Beyond Per-Word Pricing?

On September 29, language-technology platform Phrase and language-services provider Vistatec launched Phrase Managed Outcomes. Instead of charging by word volume, the service begins with the enterprise defining what “done” means for each language and content type. Phrase then builds and operates the localization pipeline, while Vistatec provides linguistic calibration and ongoing optimization. Fees are set by language and content type rather than the number of words processed.

The process starts with Discovery, where the partners run the customer's own content through a live pipeline and make a go-or-no-go decision before committing to an outcome. Build, evaluation, and live operation follow. The system combines deterministic controls, AI-based evaluation, and human linguistic judgment. When a recurring error appears, linguists do more than correct individual sentences: they trace it back to terminology, rules, data, or routing and adjust the pipeline that produced it.

Why per-word pricing explains less than it used to

Per-word pricing was a practical proxy for human effort: longer content usually required more translation and review time. Once machine translation and large language models handle drafting, terminology retrieval, format checks, and parts of quality evaluation, equal word counts can represent very different work. Low-risk interface strings may be highly automated, while a legal notice, brand line, or safety instruction may require domain specialists, additional validation, and a named accountable reviewer.

Managed Outcomes changes the purchasing unit from throughput to the ability to maintain an agreed standard. Phrase says early customer work has reduced global content-production costs by as much as 80% and manual review hours by 70%. Its product page also describes results involving software-interface adaptation, enterprise knowledge-base content, and fewer manual tasks. These are vendor-reported figures, not independent audit results, and they should not be generalized across languages, content types, or organizations.

The difficult part is defining “done”

If quality remains vague, outcome-based pricing merely moves the dispute from word count to the outcome itself. Which errors are critical? How will terminology, tone, formatting, and turnaround be measured? How will sampling avoid selecting only easy content? Who decides whether a sentence works in the local market? These definitions need to become acceptance rules before the automated pipeline goes live.

Phrase's model lets enterprises set a different risk posture for each content type. Support material, interface strings, and internal documentation may run with lighter human intervention; brand content and higher-risk material need deeper assurance. Critical and safety errors should sit outside an average score and trigger immediate action. Human expertise therefore does not disappear. It moves from editing every segment toward calibrating standards, handling exceptions, detecting drift, and owning judgment.

Three procurement lists to rewrite first

• Acceptance: Define accuracy, terminology, tone, formatting, timing, and critical-error thresholds separately by content type and language, using the organization's own historical content as a reproducible baseline.
• Risk and human review: State which content may publish automatically, which is sampled, and which requires line-by-line approval by a qualified linguist or domain expert. Define escalation for uncertainty, drift, and abnormal results.
• Contract and data: Specify sampling and audit rights, notification and remediation when the outcome is missed, and ownership and portability for translation memories, termbases, style guides, and evaluation data.

Per-word pricing will not disappear overnight. It remains simple and transparent for one-off projects, creative work, and tasks where human effort still tracks length closely. Outcome-based models fit high-volume, recurring, measurable enterprise content more naturally. A prudent pilot starts with one content type and one or two languages, establishes current cost, editing effort, critical-error rates, and turnaround time, and then compares results before expanding.

For global businesses, the important signal is not that localization will stop counting words. It is that procurement is beginning to buy verifiable quality, risk control, and accountability instead of processing volume alone. A pricing sheet can change quickly; the value of the new model depends on whether the buyer can define the outcome clearly enough to test and enforce it.

Sources: Phrase company announcement (September 29, 2026); Phrase Managed Outcomes product overview; VistatecMAP overview. Cost and manual-review reductions are vendor-reported and will vary by language, content type, existing language assets, and the agreed quality standard.