EVAavto case study: how to scale e-commerce from $5K to $75K without losing cost per lead

E-commerce is one of the most competitive and, at the same time, most mathematically transparent niches in digital marketing. There's no room for “faith in the brand”, only a formula: cost per click × conversion rate × average order value = ROAS. In this case study we show how we improved each part of that formula and ended up growing 15x in 12 months.

Context: the starting point

In 2024 we were approached by EVAavto, a Ukrainian manufacturer and seller of car mats made of EVA (ethylene-vinyl acetate). These are dense honeycomb mats that keep out water, dirt and sand. The premium segment of the Ukrainian car market.

Starting conditions:

• A working website with a catalog of 800+ models for different cars
• Made in Ukraine, 35-40% markup
• Average order: $80-150 per set
• Starting budget: $5,000/month
• Main KPI: ROAS of at least 250%

The client's team had already tried running ads themselves. Results: a 150-180% ROAS, essentially breaking even after operating costs. The strategy needed a complete overhaul.

Strategy: 4 stages of scaling

The main task in e-commerce isn't “setting up ads” but building a system that can grow by orders of magnitude. A combination that works at $5K/month usually breaks at $20K/month, as Facebook's algorithm starts burning through audiences. To grow, you need parallel “ladders” of different audiences, creatives and placements.

Stage 1. Months 1-2: the foundation (budget $5K → $8K/month)

In the first 2 months we did 3 things:

1. Technical foundation: we set up the Facebook Pixel + Conversions API for every user action (product view, add to cart, checkout, payment). Without the Conversions API, 40% of data has been lost since iOS 14.5
2. A product catalog in Facebook: we uploaded 800+ items with stock and prices updated automatically via a feed
3. A baseline test: 6 audiences at $50/day with test creatives. We found 2 winners

Stage 2. Months 3-5: consolidation (budget $8K → $25K/month)

Once the baseline combinations are found, we gradually increase the budget (by no more than 25% a week) and test in parallel:

• Lookalike of buyers: the best-performing channel, with a 450%+ ROAS
• Dynamic retargeting: the ad shows the product the user viewed on the website
• Abandoned carts: a separate campaign with a 5% discount code for people who added a product but didn't buy

Stage 3. Months 6-9: scaling (budget $25K → $55K/month)

The most critical stage. This is where 90% of campaigns usually break: the algorithm burns through the core audience and ROAS drops. To keep growing, we opened up new “directions”:

• 2%, 3% and 5% lookalikes based on different sources (buyers, visitors, add-to-cart users)
• Interests by car model: Toyota Camry, BMW X5, Volkswagen Polo, etc. Each audience with its own creative
• Video views on Reels: a separate campaign with short videos of mats being installed

Stage 4. Months 10-12: the peak (budget $55K → $75K/month)

At the peak, we added more channels and held ROAS steady:

• Google Shopping to capture hot demand
• YouTube TrueView with video reviews of different mat types
• Expansion into Poland: the Ukrainian diaspora plus local buyers

Key findings

1. Dynamic retargeting is the main “gold mine” channel

Over 12 months, dynamic retargeting delivered 32% of all revenue on just 8% of the budget. That means this channel's ROAS was around 1,500%, 4 times the project average.

The secret is that the user has already chosen a specific mat model on the website. Facebook shows them that exact item (by SKU) in the feed and Stories. The purchase decision is often made right in Facebook, without going back to the website.

“If you run an e-commerce business without dynamic retargeting, you're leaving 30-40% of potential revenue on the table. It's the first thing to set up BEFORE you scale cold traffic.” — Natalia Maslakova, founder of MAS Project

2. Installation videos delivered +50% conversion

The best-performing format across all 12 months: short 15-second videos of mats being installed in a real car. Not a professional shoot, just a simple first-person phone video.

Why it works: buyers see the mat fitting their car model, understand it will fit, and see the material quality on video. Doubts are resolved without them having to message the store.

3. Expanding by 25% a week

The main rule of safe scaling: no more than +25% to the budget per week. It's the rule beginners break most often. You want to go faster, but Facebook can't optimize in time.

In the EVAavto case we broke this rule once, raising the budget by 80% in a week at the client's request. The cost per lead jumped from $5 to $12 in 3 days and ROAS fell from 380% to 180%. We rolled it back, and recovery took a month.

What didn't work

Over 12 months there were plenty of failed tests too. The main ones:

• Ads aimed at “all drivers”: a broad audience with no interests delivered an 80-120% ROAS. Targeting all drivers seems logical, but conversion is low
• Creatives with a “30% off” promo: we tried big discounts to grab attention. Leads were 40% cheaper, but purchase conversion halved. Buyers were “waiting for an even bigger discount”
• Influencer vlogs: we paid $2,000 for 3 integrations with car bloggers. ROAS was about 70%. The bloggers' audiences didn't match car mat buyers

The project's final economics

Of the 15,000 leads, 25% became buyers (the norm for mid-market e-commerce): 3,750 buyers. At a $115 average order, that's $431,000 in gross revenue and a 380% ROAS after ad spend and operating costs.

Universal takeaways for any e-commerce business

1. The Conversions API is a must: without it you lose 40% of data since iOS 14.5
2. A product catalog in Facebook + dynamic retargeting: the most profitable e-commerce channel
3. Installation/usage videos work twice as well as static photos
4. Scale by 20-25% a week, no more: Facebook's algorithm is sensitive to sudden jumps
5. Lookalikes of buyers, not leads: buyers share common spending patterns
6. Don't offer discounts that are too big in your creatives: it lowers purchase conversion