Marketing ops · Production

UGC Outperforms Branded Stock Visuals In Facebook Ads for Air France

The problem

Air France wanted to diversify their social ad visuals and better relate to audiences, but their team was manually searching for images on Instagram, making UGC discovery slow and impractical at the scale required.

First attempt

Before Stackla, Air France's social media team manually searched for UGC images on Instagram, a process that could not scale to the volume needed across multiple destination campaigns.

Workflow diagram · grounded in source
1
A/B test design
trigger
“Six travel destination were selected for the test: New York, San Francisco, Havana, Costa Rica, The Maldives and Réunion. To guarantee the reliability of the test, the ad copy used in the paid posts were the same and just the images that…”
2
AI content aggregation
ai_action
“they aggregate content from multiple social networks by geolocation and then leverage automated and custom content tags for fast discovery, filtering and distribution”
3
Visual recognition auto-tagging
ai_action
“relying on Stackla's visual recognition tool to scan and auto-tag each image with relevant visual markers like "bridge" or "person"”
4
Manual filtering and curation
human_review
“These tags allow their team to quickly filter out images that feature people and search for only images that feature the Brooklyn Bridge. Once they've identified the images they want to use throughout each of their targeted campaigns, th…”
5
Rights management
integration
“Air France leveraged Stackla's comprehensive Rights Management workflows”
6
Facebook ad launch
output
“Air France launched the Facebook ads for all six destinations”
7
A/B test measurement
feedback_loop
“Running for 9 months, the test revealed that the ads featuring user-generated images performed better than their branded stock photo ads across all key performance metrics”
Reported outcome

Over a 9-month A/B test across six destinations, UGC ads outperformed branded stock photos on all key metrics: 4% higher CTR, 1% higher CVR, 3% lower CPC, and 9% lower CPA.
A second flash sale round showed even larger margins. The team also reported saving hours each week.

Reported metrics
average CTR (UGC vs stock)4 percent higher
average CVR (UGC vs stock)1 percent higher
average CPC (UGC vs stock)3 percent lower
average CPA (UGC vs stock)9 percent lower
Show all 8 reported metrics
average CTR (UGC vs stock)4 percent higher
average CVR (UGC vs stock)1 percent higher
average CPC (UGC vs stock)3 percent lower
average CPA (UGC vs stock)9 percent lower
flash sale average CTR (UGC PPL vs stock)11 percent higher
flash sale average CPC (UGC PPL vs stock)21 percent lower
flash sale average CPA (UGC PPL vs stock)2 percent lower
team time savingssaving hours each week
Reported stack
Stacklavisual recognition tool
Source
https://www.nosto.com/case-studies/air-france/
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Over a 9-month A/B test across six destinations, UGC ads outperformed branded stock photos on all key metrics: 4% higher CTR, 1% higher CVR, 3% lower CPC, and 9% lower CPA.

What tools did this team use?

Stackla, visual recognition tool.

What results were reported?

average CTR (UGC vs stock): 4 percent higher; average CVR (UGC vs stock): 1 percent higher; average CPC (UGC vs stock): 3 percent lower; average CPA (UGC vs stock): 9 percent lower (source-reported, not independently verified).

What failed first in this deployment?

Before Stackla, Air France's social media team manually searched for UGC images on Instagram, a process that could not scale to the volume needed across multiple destination campaigns.

How is this marketing ops AI workflow structured?

A/B test design → AI content aggregation → Visual recognition auto-tagging → Manual filtering and curation → Rights management → Facebook ad launch → A/B test measurement.

Related marketing ops cases