The PPC Framework That Delivered 5.2× ROAS for Our Client
This is the exact framework we used to take Elara Commerce's paid media from a 1.8× ROAS to 5.2× in six months. We're sharing it in full — not as a theoretical model, but as the specific structure, logic, and decisions that drove the result. If you're running paid media and your ROAS is stuck, this is where to start.
Step 1: Rebuild the Campaign Structure
Elara's original account structure was a classic mistake: broad-match campaigns with generic creative aimed at everyone. No meaningful segmentation, no funnel logic, no budget weighting by intent level.
We rebuilt around a three-tier funnel structure. Tier one (Awareness): broad interest-based targeting for new audience prospecting. Budget allocation: 25% of total spend. Goal: impressions and engagement, not conversions. Tier two (Consideration): retargeting audiences who had engaged with content, visited product pages, or watched 50%+ of a video ad. Budget: 35%. Goal: add-to-cart and initiation of checkout. Tier three (Conversion): highest-intent audiences — cart abandoners, product page viewers from the last 7 days, email list lookalikes. Budget: 40%. Goal: purchase, with ROAS as the primary metric.
The key insight is that ROAS is a lagging indicator. You can't optimise for it at the awareness stage. Each tier needs its own KPIs and its own creative logic — which leads to the second step.
Step 2: Fix the Bidding Strategy
Manual CPC bidding was eating into margin and limiting scale. The algorithm can optimise more effectively than any human for conversion-based goals — but only once it has enough signal to learn from.
We moved to Target ROAS bidding, but only after reaching the critical threshold: 50+ conversions per ad set over a 7-day rolling window. Below this threshold, automated bidding strategies actually underperform manual bidding because the algorithm is guessing. We spent the first six weeks building up to this threshold before switching.
We also implemented portfolio bid strategies across related campaigns, allowing the algorithm to allocate budget dynamically between campaigns based on real-time conversion probability. Combined with dayparting adjustments for peak purchase hours, this added approximately 15% efficiency on top of the ROAS improvement from the structure rebuild.
Step 3: Build a Real Creative Testing Framework
Creative is where most ad accounts leave the most performance on the table. Elara was running 3–4 ads per month, with no systematic testing framework, and keeping ads running long after they'd fatigued.
We introduced a structured creative testing programme: 40+ ad variants per month across Meta and Google, organised by test variable (hook, offer, format, proof element). Every ad is a hypothesis: 'We believe [this hook] will outperform [the control] for [this audience segment] because [reason].'
The testing framework operates on a two-week cycle. Week one: launch test variants with equal budget. Week two: identify winner, pause losers, brief next set of challengers. Winning creatives run for a maximum of four weeks before being rotated out — ad fatigue in e-commerce is real and faster than most advertisers account for.
The result: a continuously refreshed creative library that maintains performance rather than decaying. This alone accounts for roughly one full ROAS point of the total improvement.
Step 4: Fix the Audience Architecture
Custom audiences from Shopify purchase data, segmented by LTV cohort (top 10% of buyers vs. all buyers), provided the seed data for lookalike modelling. The critical insight: lookalikes based on your best customers dramatically outperform lookalikes based on all customers. The difference in CPL can be 30–50%.
We also built out a comprehensive exclusion architecture: recent purchasers excluded from conversion campaigns (they already converted — stop spending on them), existing email subscribers excluded from prospecting (you have a cheaper channel to reach them), and high-bounce audiences excluded entirely to improve signal quality for the algorithm.
ROAS is an outcome, not a lever. The real levers are campaign structure (does your account reflect how customers actually buy?), bidding strategy (is the algorithm getting enough signal?), creative freshness (is your creative fatigued?), and audience signal quality (are you giving the algorithm your best data?). Get those four right and the numbers will follow.
Want Personalised Growth Advice?
Stop reading — start growing. Let our team analyse your business and give you a tailored roadmap for compounding digital growth.