Business: a fragrance subscription company
Scale: More than 1 million subscribers.
Creative testing budget: millions of dollars per month
Main markets: US, Canada, and UK.
Recognition: Named Best Perfume Subscription Box in Vogue’s beauty subscription guide.
Results
CUSTOMERS
CPA
+11%
AD SPEND
Two-month live Meta test: Plurio’s rule vs. the existing spend-only rule on matched ad sets.

Client quote
“Plurio’s ML models found better stopping thresholds for our creative tests, and its platform applies them automatically in Meta. When weak ads are paused, Meta can spend more on the ads that are working, helping us scale acquisition faster and more efficiently. We’re rolling the rule out across all Meta campaigns and bringing the same approach to the main product.”
Oleg Popov
VP of User Acquisition at Scentbird
Challenge
Stopping an ad with no subscriptions is a difficult call: keeping it running can waste test budget, but stopping it too early can cut off a future winner. Scentbird makes that decision for 1,000+ new creatives every month while trying to grow acquisition without raising CPA. Its existing rule paused every ad with no subscriptions at the same spend threshold. Two facts explain why that spend threshold was not enough:
12% of new creatives brought in no subscriptions
A high spend threshold could let these ads keep spending test budget before the rule stopped them.
Just 5% of new creatives became winners, generating 83% of subscriptions
A lower spend threshold could stop potential winners that take longer to earn their first subscription.
Solution
Plurio found the best early signals and thresholds for stopping zero-subscription ads to reduce wasted spend on creative tests while limiting the risk of stopping future winners. The ML analysis identified a lower spend threshold for stopping zero-subscription ads, along with new CPM and CPC thresholds that allow even earlier stops. Plurio rules apply the chosen logic automatically, checking active ads and pausing them directly in the Meta account when the stopping conditions are met.
Plurio rule
Existing rule
Paused zero-subscription ads at the lower spend threshold, or earlier when CPM or CPC conditions were met.
Paused zero-subscription ads at the original, higher spend threshold only
How Plurio turned ML analysis into automated ad stops
01
Examine all potential early signals
Plurio reconstructed historical ad performance hour by hour, using spend, impressions, clicks, video engagement, and first-subscription events. The models checked which early signals (Spend, CPM, CPC, CTR, hook rate, or 15-second view rate) could show which ads would bring in subscribers and which would perform poorly. Spend, CPM, and CPC emerged as the strongest combination, so the rule uses all three.
02
Model the tradeoff
03
Test the rules on new data
04
Run the rule automatically in Meta
Two months of live creative tests
Test setup
Scentbird tested the rule across 6 weekly creative batches, launching the same creatives in matched test and control campaigns. Plurio’s rule governed the test campaign; the existing spend-only rule governed the control.
Result
Metric
Plurio rule vs. existing rule
CTR
No change
CPC
−7%
CPM
−7%
Ad spend
+11%
Customers
+21%
CPA
−9%
Average spend was 19% lower for zero-subscription ads managed by Plurio than for those managed under Scentbird’s previous spend-only rule. This freed more budget for Meta to allocate to better-performing creatives in each ad set, helping deliver 21% more customers at a 9% lower CPA.
CPA was compared with and without outlier adjustments to see whether unusually strong creatives were driving the result. The test campaign had a lower CPA in both analyses.
Who is Plurio for
Monthly spend
$200K+ in ad spend
Ad volume
1,000+ ads per month
Business setup
Multiple products, a multi-step funnel
Growth goal
Scale acquisition without scaling manual work
Why Scentbird Chose Plurio
Media buying automated from planning to optimization
From your spend targets, our agent builds campaign plans, launches campaigns with new creatives, and optimizes them: stopping underperformers, scaling winners, and moving budgets based on your playbook.
Results compound as ML models refine decisions each cycle
Our models test combinations of signals and thresholds on your data, weighing projected net impact against risk. New results guide refinements; proven improvements become part of the rules that run daily.
20+ years of expertise behind every automation
Dedicated AI marketing engineers learn your goals, funnel, and campaign setup. They combine hands-on expertise in each ad channel with our library of proven skills, then tailor the skills and rules to your business.

Client quote
“Plurio brings together automation and a team of acquisition experts. They learn our goals and campaign setup, find and test more efficient ways to reach those goals, then automate what works best within our guardrails.”
+21%
CUSTOMERS
−9%
CPA
+11%
AD SPEND



