How we solved multilingual advertising and built Ads Engine
Concept 1: The Multilingual Advertising Problem
What is it?
When a company sells products in multiple countries, they need ads in multiple languages. This sounds simple, but it creates serious operational challenges.
The Core Tension
- Business goal: Reach customers in their native language
- Reality: Most marketing teams are small and speak only one language
Specific Pain Points Introduced
| Problem | Why It Hurts |
|---|---|
| Manual keyword translation | Hundreds of rows, one language at a time |
| Character limits (30 for headlines, 90 for descriptions) | Translated text almost always exceeds limits |
| Every violation needs manual fixing | Multiplied across every language × every ad group |
Key Insight
Every day spent fixing character limits = a day your campaign is absent from that market = lost revenue to competitors
Concept 2: Translation vs. True Localization
What is the Difference?
Translation = Converting words from one language to another Localization = Adapting the entire experience to fit a culture
Why Does This Matter in Advertising?
Localization goes beyond words. It includes:
- Colors → carry different emotional meanings across cultures
- Dialect → Latin American Spanish ≠ Spain Spanish
- Visual context → a black-and-white design feels modern in the US, but can signal mourning in Japan
- Brand naming → Diet Coke becomes Coke Light in Latin America because "diet" has different cultural weight
The Video Problem Illustrated
What they achieved: Audio ✅ (dubbed voice, tone, pacing preserved)
What was missing: Visuals ❌ (text overlays, UI screenshots still in English)
Result: Videos sounded local but LOOKED American
The Scaling Problem
When they did invest in full localization (visuals + dubbing + copy together):
- Quality was high ✅
- Took weeks per campaign ❌
- Required freelancers + dedicated production cycles ❌
- Could not scale to every campaign in every language ❌
Concept 3: The Production Bottleneck in Static Ads
The Cycle That Kept Repeating
High-performing English creative identified
↓
Translate headlines in spreadsheet
↓
Brief the design team to rebuild image with new text
↓
Wait 3–5 business days (sometimes longer)
↓
Manually upload new assets to each ad platform
↓
[Repeat for every language × every new creative]
Why This Is a Business Problem, Not Just an Inconvenience
- While waiting for the design queue → campaign runs in English only
- Competitors are actively showing ads to those non-English users
- Every new winning creative restarts the same slow cycle
- Multiple disconnected tools involved = multiple handoff points = multiple delays
The Fragmentation Problem
The team was juggling:
- Ad platforms (Google Ads, Bing Ads)
- Spreadsheets for translation
- Creative measurement platforms
- Design request queues
Each tool handoff added days, not hours.
Concept 4: The Engineering Solution — Collapsing the Workflow
The Design Goal
If a creative performs well in one language, the time between that signal and the localized version going live in every other market should be minutes, not days.
What the Internal Tool Did
| Old Step | New Reality |
|---|---|
| Download keywords → translate in spreadsheet | Automated with character-limit compliance built in |
| Brief design team → wait 3–5 days | Automated image adaptation with localized text |
| Manual video dubbing coordination | Dubbing V2 integrated directly |
| Manual upload to each platform | Pushed back automatically |
The Result in Human Terms
Before: 4 English-speaking people → English-only campaigns
After: Same 4 people → campaigns in 7 languages
No new headcount. Same team. Dramatically expanded output.
Concept 5: Measuring the Business Impact
Why Measurement Matters Here
The article is careful to connect the operational improvement to financial outcomes. This is important — efficiency alone doesn't justify investment. Revenue does.
The Numbers
- +17.6% lift in conversions from non-English campaigns vs. English alone
- $3.78 million in incremental conversion value generated
- 7.16 ROAS (Return on Ad Spend — meaning every $1 spent returned $7.16)
How They Found the Opportunity
They used native-language search demand data — meaning they looked at what high-intent users were already searching for in their own languages, then built campaigns to meet that existing demand.
This is a key strategic concept: demand already existed. The bottleneck was purely operational, not market-based.
Concept 6: From Internal Tool to Commercial Product
The Logic of Productization
The company solved a problem for themselves → the solution worked exceptionally well → they recognized the problem is not unique to them.
Most brands face the same gap:
- They know international markets represent revenue
- They cannot close the gap because production is too slow, expensive, and fragmented
What Ads Engine Does (The Product)
Connect to ad accounts
↓
Pull existing creatives
↓
Localize: text (with platform spec compliance) + images + dubbed video
↓
Push finished ads back to platforms automatically
The Continuous Loop (Beyond One-Time Localization)
This is the more advanced concept introduced at the end:
Campaigns run
↓
Performance data flows back in
↓
Fatigue detection identifies when creatives are wearing out
↓
Fresh variants generated automatically
↓
No need to restart production from scratch
This transforms localization from a one-time production event into a continuous optimization system.
Summary: The Conceptual Arc of the Article
Problem identified
(multilingual ads are slow, manual, fragmented)
↓
Root cause diagnosed
(translation ≠ localization; production bottlenecks; tool fragmentation)
↓
Solution engineered
(unified workflow: translate → adapt → dub → push)
↓
Results validated
($3.78M incremental value, Google award)
↓
Insight generalized
(this problem exists for every brand, not just us)
↓
Product launched
(Ads Engine: the internal tool made available externally)
The article teaches a complete cycle: identify a real operational problem → solve it rigorously → measure the outcome → scale the solution to others who share the same problem.