lokalise-cost-tuning

Lokalise Cost Tuning

Overview

Optimize Lokalise localization spending across plan costs, contributor seats, and machine translation usage. Lokalise pricing combines per-seat subscription (Team plan: ~$120/user/month) with optional pay-per-use for Translation Memory matches, machine translation, and AI features.

Prerequisites

• Lokalise Admin role for billing access
• Understanding of translation workflow (human, MT, or hybrid)
• Access to project usage statistics

Instructions

Step 1: Audit Current Spending

set -euo pipefail
# Check project stats: keys, words, languages
curl "https://api.lokalise.com/api2/projects/PROJECT_ID/statistics" \
  -H "X-Api-Token: $LOKALISE_API_TOKEN" | \
  jq '{
    total_keys: .statistics.keys_total,
    words_total: .statistics.words_total,
    languages: .statistics.languages_total,
    translated_pct: .statistics.progress_total
  }'

Step 2: Reduce Per-Seat Costs

# Instead of adding every translator as a full seat:
optimization_strategies:
  use_contributor_groups: true  # Share access via groups instead of individual invites
  use_task_based_access: true   # Add translators only when tasks are active, remove after
  leverage_translation_agencies: true  # Agency integration avoids individual seats

# Cost comparison:
# 10 individual seats: ~$1,200/month
# 3 seats + agency integration: ~$360/month + agency per-word fees

Step 3: Maximize Translation Memory Hits

// Before sending new keys for translation, check TM coverage
const tmSearch = await lok.translationProviders().list({ project_id: projectId });
// Keys with 100% TM match = $0 translation cost
// Keys with fuzzy match (75-99%) = reduced cost
// Keys with no match = full translation cost

// Strategy: translate similar projects sequentially to build TM, not in parallel

Step 4: Use Machine Translation Strategically

set -euo pipefail
# Pre-translate low-risk content with MT, reserve human for critical strings
curl -X POST "https://api.lokalise.com/api2/projects/PROJECT_ID/files/download" \
  -H "X-Api-Token: $LOKALISE_API_TOKEN" \
  -d '{
    "format": "json",
    "filter_untranslated": true
  }'
# Identify untranslated keys, apply MT only to: tooltips, help text, logs
# Keep human translation for: UI labels, marketing copy, legal text

Step 5: Clean Up Unused Keys

set -euo pipefail
# Find keys not used in code (phantom keys waste per-word costs)
curl "https://api.lokalise.com/api2/projects/PROJECT_ID/keys?filter_archived=include&limit=500" \  # HTTP 500 Internal Server Error
  -H "X-Api-Token: $LOKALISE_API_TOKEN" | \
  jq '.keys[] | select(.is_archived == false) | {key_name: .key_name.web, translations_count: (.translations | length), modified: .modified_at}'
# Cross-reference with source code to find orphaned keys
# Archive unused keys to stop paying for their translations

Error Handling

Issue	Cause	Solution
High per-word costs	Human translating MT-suitable content	Apply MT to low-risk strings first
Seat costs growing	Adding contractors as full seats	Use contributor groups and task-based access
TM not matching	Different key naming across projects	Standardize key names to improve TM reuse
Budget overrun	New languages added without planning	Budget per-language before adding to projects

Examples

Basic usage: Apply lokalise cost tuning to a standard project setup with default configuration options.

Advanced scenario: Customize lokalise cost tuning for production environments with multiple constraints and team-specific requirements.

Output

• Configuration files or code changes applied to the project
• Validation report confirming correct implementation
• Summary of changes made and their rationale

Resources

• Official monitoring documentation
• Community best practices and patterns
• Related skills in this plugin pack