The Goal
Improve what already works.
Any gains prove smart management.
About
Game Localization
Workflow Upgrade
Test-First Approach to MTPE Implementation
- 37% cost savings
- 95%+ TQI quality guaranteed
- 12 languages
Why Upgrade?
Behind this story is one of Europe's fastest-growing mobile game companies*. They already had a localization process that worked. But with AI reshaping the industry, they asked: could we modernize without compromising quality?
[ 01 ]
Technology adoption
AI is reshaping localization workflows industry-wide
[ 02 ]
Process modernization
Opportunity to improve an existing workflow
[ 03 ]
Measurable results
Demonstrate responsible innovation with clear metrics
[ 04 ]
Budget optimization
Positive side effect if achievable without quality loss
* Details have been anonymized because of NDA restrictions.
What We Did
We tested 10 AI engines, validated the best performers with a 1,500-word pilot, then transitioned all languages to MTPE. The same translation team stayed on, now post-editing AI output instead of translating from scratch.
Transit approach
100% traditional → MTPE with mandatory human post-editing
Safety measure
100% linguist retention: same team, new efficiency
Scope
All 12 languages
Timeline
Pilot (February 2025) → Full production (July 2025)
Results
* TQI — Translation Quality Index. Each linguist in our pool carries a TQI score — Translation Quality Index, an industry-standard metric — based on their reviewed project work. This is how we verify translation quality by the numbers, objectively.
The Game Title Hidden
- 20M+installs
- 1M+reviews
- Globally scaledtitle with a deeply engaged player community
Languages
European Big Four DE · ES · FR · IT
Asian Classic JA · KO · ZH-CN
Asian Rising Stars ID · MS · TH
Also NL · PL
Localization Scope
Data Pack UI, inventory & location names
Narrative Quests
Tech Crowdin + API + CLI
This Project's SLA
| Task type | Confirmation | Delivery |
|---|---|---|
| Regular | 3 hours | 1 day |
| Urgent up to 200 words | 1 hour | 12–24 hours |
Why test first
With this scale and complexity, we needed to test carefully before introducing AI into the workflow.
Test First, Scale Second
We designed a three-phase approach to make the transition safe and measurable.
Evaluation: Technical Testing
- Tested 10 AI engines across all languages
- Measured quality with Alconost Quality Index
- Selected best performers per language
Pilot: Real-World Validation
- Ran 1,500-word pilot with actual game content
- Tracked linguist effort and MT performance
- Confirmed quality achievable at 95%+ TQI
Rollout: Full Implementation
- Gradual, by batches (4 languages per month)
- Retained existing translators
- Maintained TQI 95%+ quality guarantee
Principle
We validated every step before committing.
Phase 1: Engine Evaluation
We tested different AI engines to find the best match for this game's content.
10 engines tested
Gemini, ChatGPT-4o, Grok, Mistral, Microsoft, and 5 more. All 12 languages were tested across each engine.
AQI as a quality metric
AQI (Alconost Quality Index) components: CometQE, nTER, BLEU, CHrF, COMET, BERTScore + Linguist Evaluation.
| Quality | AQI | Lang. | Description & recommendation |
|---|---|---|---|
| Good | > 85 | 6 of 12 | Near-human quality; fluency and adequacy are strong. Accept as-is or very light post-editing. |
| Acceptable | > 70 | 11 of 12 | Good fluency, minor errors; meaning mostly preserved. Light to moderate post-editing. |
| Borderline | 69 | 1 (MS) | Understandable but noticeable issues. Moderate post-editing needed for fluency. |
Phase 1 takeaway
All 12 languages ready for MTPE with human supervision. More editing effort for Malay; light touch for others.
Verdict:
AI Can Work Here
Recommended workflow:
Supervised MTPE
AI-generated translations achieved acceptable to good quality. Human post-editing brings output to 95%+ TQI.
Next step:
Real-world pilot
Validate with 1,500 words of actual game content.
Phase 2: MTPE Pilot
With engines selected, we ran a real-world test using actual game content.
Pilot Setup
Sample volume1,500 English words
Content typesUI strings and quest lines (mix)
EnginesBest performers from Phase 1 evaluation
AssetsTranslation Memory, Glossary, Style Guide
What we tracked
Linguist effort and MT performance quality
Edit distance is one parameter showing how well we configured MT (prompting, settings). Lower distance means less linguist effort and better MT setup. We also tracked overall quality using a 5-step MTPE process to confirm 95%+ TQI is achievable.
Pilot Results: AI Output Quality
~40% No edits needed
Pre-translation delivered at target quality. Linguists approved as-is.
~50% Light edits
Edit distance up to 30%. Quick fixes for fluency or terminology.
~10% Significant work
Edit distance 61–99%. Meaning issues or adaptation required.
Phase 2 result: Green light for full rollout
Quality achievable, time savings confirmed, linguists comfortable with the workflow.
Phase 3: Implementation
We began by switching the top European languages to MTPE.
After a month of careful performance monitoring, Dutch and the top Asian languages followed.
A month later, the remaining languages were successfully transitioned as well.
Results
The upgrade worked: careful testing + clear communication + team continuity = sustained results.
Production Updates
4 major updates delivered on schedule during MTPE period
Player Experience
No increase in localization issues reported
Cost Reduction
37% savings after MTPE implementation
Delivery Speed
SLA maintained: 3-hour confirmation, 1-day delivery
Quality Maintained
95%+ TQI guaranteed
Team Continuity
100% linguist retention
Ready to optimize your
game localization
workflow?
We test AI solutions systematically, guarantee quality at 95%+ TQI, and help you make evidence-based decisions.
- Systematic evaluation — Test engines on your content first
- Quality guaranteed — 95%+ TQI in SLA
- Team continuity — Same linguists, new efficiency
- Measurable results — Clear metrics from day one
- 1,000 games localized since 2004 — See more cases
Get Started
Request a Quote
Tell us about your game. We’ll share a tailored plan and pricing within one business day.
Services
Translation & Localization
360° Localization Full-stack localization Software & Product Localization Marketing & Content Translation SEO, ads, SMM Game Localization Website Localization 24/7 On-Demand Human TranslationMedia & DTP
Audio & Video Localization Transcribing, subtitling Desktop PublishingQuality & Process
Service Levels & Pricing Languages, project planning, workflows Machine Translation Post-Editing (MTPE) Human and AI Localization Testing & QA Functional, linguistic, cultural Localization QA Proofreading & Editing Cultural Analysis Vendor Management Outstaffing, full-stack teams, mixed Asset Management Glossaries, TM, style guidesOur Work
By Platform
Mobile iOS & Android Web Apps Desktop Software PC Gaming Steam, Epic, GOG Game Consoles PlayStation, Xbox, NintendoBy Content Type
UI/Software Strings Game Content Dialogue, quests, items Video & Audio Subtitles, voiceover E-commerce Product listings, catalogs Technical Documentation Manuals, API docs Marketing & Web Ads, SEO, websites Support Help center, KB, FAQ Training E-learning, explainer videos Legal Terms of service, privacy Finance Reports, banking HR Internal communicationsBy Industry
Gaming Technology & SaaS Mobile Apps E-Learning & Education Media & Entertainment Automotive Customer Support Healthcare Travel & Hospitality Sports & Fitness Logistics Finance & Banking Telecommunications E-Commerce Food & Beverage IoT & Hardware Marketing & Advertising Real Estate Energy & Sustainability Legal ManufacturingTechnology