Meta variant testing: How Best Mobilier scaled ROAS by 2.3x

When running dynamic product ads on Meta, delivery algorithms often prioritize rapid impression distribution over fair experimental testing.

French eCommerce furniture retailer Best Mobilier faced this exact challenge during a high-stakes promotional window where they were trying to find which items and variants really perform best.

Partnering with digital performance agency NextMomentum, Best Mobilier leveraged Channable's Meta Variant Testing to run a controlled split test. The objective was clear: identify which creative variant genuinely drove the strongest performance during the sale period, rather than allowing Meta's delivery algorithm to determine the winner based on early delivery signals.

Within seven days, the team isolated the winning ad design, driving a +50% increase in purchases, achieving a 2.3× ROAS, and reducing acquisition costs by over 60%.

Key Takeaways

To solve Best Mobilier's creative distribution bottlenecks, NextMomentum deployed a controlled testing framework through Channable. This structured experiment delivered significant revenue expansion and efficiency gains:

• Controlled creative delivery: The team tested three feed-driven image variants using Channable Meta Variant Testing, reducing the risk of Meta favoring one creative prematurely.

• Expanded purchase volume and ROAS: Best Mobilier achieved a +50% increase in total purchases alongside a 2.3× ROAS during high-competition sale weeks.

• Maximized spend efficiency: The winning ad variant generated nearly one-third (~33%) of total campaign purchases while consuming less than 15% of the overall budget.

• Dramatically lowered acquisition costs: During peak promotional windows, Cost Per Acquisition fell by more than 60% compared to the rest of the ad account.

• Accelerated time-to-result: NextMomentum identified the winning creative within one week, allowing immediate budget consolidation behind proven ad assets.

• Controlled creative workflow: NextMomentum created variants, pushed them directly into Meta's native split-testing environment, and scaled the winning creative into live campaigns — connecting creative production, testing, and scaling in one workflow.

Overcoming algorithmic asset bias in Meta Advantage+ campaigns

How Meta's algorithm can distort creative tests

Meta’s Advantage+ delivery system is built for rapid, cost-efficient delivery rather than controlled strategic learning. When performance marketers add multiple creative assets to a standard campaign, Meta’s machine learning models assess early auction signals, such as initial clicks, bids, and predicted action rates, to decide which ad to serve.

This delivery method creates an early delivery feedback loop. If Creative A receives a few early interactions, Meta's algorithm quickly funnels the majority of the campaign budget toward it.

Meanwhile, Creative B and Creative C receive minimal spend, suppressing their impressions before they can demonstrate their true revenue potential.

This means standard delivery tells advertisers which creative Meta served most often, rather than which creative actually drives return on ad spend (ROAS). For a deeper look into pre-auction parsing and auction mechanics, review Meta ad quality and auction signals.

Bottlenecks of manual creative testing

To bypass algorithmic bias, performance teams often attempt to run manual split tests inside Meta Ads Manager. However, manually building ad sets, mapping individual image links across external tools, and monitoring separate campaigns can create operational bottlenecks:

• Media buyers lose hours manually copying, pasting, and checking asset mapping across disjointed tools.
• Design teams waste resources creating one-off visual variations without catalog-wide automation.
• High-spend promotional windows can pass before teams have enough performance data to confidently identify a winning creative.

Controlled creative experiments with Channable Meta Variant Testing

Step 1 – Building catalog-driven creative variants

NextMomentum began by connecting Best Mobilier's product catalog directly to Channable. Using automated feed rules and the built-in Dynamic Image Editor, the team generated three distinct visual treatments directly from the product feed.

Rather than altering copy, audiences, and visual layout simultaneously, NextMomentum isolated a single design variable across the catalog. This created a controlled comparison, giving the team a clearer view of how the visual overlay influenced purchase volume and ROAS. .

Product feed management ensures your underlying catalog data remains clean across all ad channels if you are scaling multichannel catalog feeds.

Step 2 – Automated push to Meta's native split-test API

After creating the image variants inside Channable, NextMomentum pushed them directly into Meta’s native split-testing environment via API. Channable MVT acts as the creative coordination layer between product feeds and Meta ads, eliminating manual campaign construction.

To maintain a clean testing environment, NextMomentum structured the test cells in accordance with standard Meta native split-testing requirements:

• Ad set budget controls: NextMomentum used ad set budgets instead of Advantage+ Campaign Budget Optimization (CBO), guaranteeing that each variant received equal daily spend.

• Audience and placement parity: Target audiences, placements, optimization events (purchases), and attribution settings were kept strictly identical across all ad sets.

• Isolated single variable: Each ad set contained exactly one creative variant, allowing Meta's split-test framework to split delivery evenly across comparable audience segments.

Step 3 – Evaluating performance signals and scaling the winner

Once the campaign went live, NextMomentum monitored performance data across multiple layers, tracking top-of-funnel scroll-stop signals (CTR, CPC) down to bottom-of-funnel conversion intent (add-to-carts, purchases, CPA, and GA4 revenue data).

Within seven days, a clear winner emerged. GA4 reporting confirmed that while two creative variants failed to hit break-even return thresholds, the third variant outpaced all expectations.

NextMomentum immediately deactivated the losing test variants and moved the winning creative asset directly into Best Mobilier's live scaling campaigns through Channable's Meta product feed integration.

Exceptional scale and budget efficiency

By replacing standard delivery assumptions with a controlled split test, Best Mobilier transformed its peak sales strategy. The table below outlines the commercial impact achieved by scaling the proven winning variant against standard account performance:

Key performance metric	Account baseline / standard ads	MVT winning variant	Total commercial impact
Purchase volume	Baseline	+50% Growth	Scaled order volume
Return on ad spend (ROAS)	Baseline	2.3× ROAS	2.3× revenue return
Cost per acquisition (CPA)	Account average	>60% Lower CPA	CPA roughly halved vs. launch
Budget efficiency	100% Account spend	<15% Budget share	Generated 33.3% of total purchases
Time to identify winning creative	2–3 Weeks (Manual)	7 Days (1 Week)	Quicker creative allocation

The commercial outcomes demonstrated the impact of structured creative testing. The winning variant generated close to one-third (~33%) of total campaign purchases while using less than 15% of the media budget, giving the team a clear basis for concentrating spend behind the strongest-performing creative.

Furthermore, CPA was more than 60% lower than the account average during the peak sale week, improving acquisition efficiency during a highly competitive period.

Tip: Advertisers can track all their catalog performance metrics using Channable's Insights & Analytics.

Building a repeatable creative engine

To turn creative testing from a one-off experiment into a repeatable growth engine, eCommerce marketers and performance agencies should adopt a structured testing framework:

NextMomentum established systematic creative testing principles during the Best Mobilier campaign that apply to any catalog-driven advertising strategy:

• Isolate one design variable per test: Test specific visual hypotheses, such as a promotional badge, seasonal border, or price overlay, against a clean control image. Testing multiple variables simultaneously hides why an ad won.

• Automate Dynamic Creatives for your feeds: Use catalog feed rules to apply winning visual templates across entire product categories instantly, eliminating graphic design bottlenecks.

• Separate testing from scaling campaigns: Validate creative assets in dedicated split-test ad sets before introducing them into broad Advantage+ (or any other) budgets.

• Build a library of tested creative insights: Document winners across product segments to inform future seasonal promotional strategies. Industry research into digital marketing creatives shows that structured creative iteration is the strongest lever for paid social scaling according to McKinsey.

For step-by-step guidance on structuring your campaign architecture, explore our comprehensive Meta creative testing guide and optimize your underlying Facebook product feed.

Scale winning Meta Ads with certainty

Stop letting ad platform algorithms guess which creative designs deserve your media budget.

By combining feed-driven automation with Meta’s native split-testing API, Channable Meta Variant Testing gives performance teams the workflow to run controlled creative experiments, identify what performs best, and scale what works — just like NextMomentum and Best Mobilier.