A staged operating system for scaling DTC brands from cold start to seven figures. Tap any card to expand the rationale, key levers, and context behind each principle.
Build the foundation. Gather data. Establish a baseline the algorithm can learn from.
The primary goal is to gather data and allow the campaign to move in a direction based on its core settings: Country, Language, and Product Feed.
This phase establishes a solid foundation for scaling the campaign efficiently. Without it, every euro spent later is built on guesswork.
PMAX is a black-box system — it needs conversion volume to learn. Premature optimization (adding tROAS, asset groups, restrictive bidding) starves the algorithm of signal and locks in bad assumptions.
Treat the first 14–28 days as data acquisition, not performance. Resist the urge to "fix" things in week one.
Move in small steps (20–30% max) and only after a stable 5–7 day window above the ROAS threshold. Aggressive jumps reset the learning phase and waste the data you've collected.
2.0 is the minimum viable signal that a market-product-feed combination has fit. Below this, the issue is rarely the campaign — it's the store, the feed, or the product mix. Diagnose first (run the checks below), then act.
Every constraint you add — asset groups, tROAS, audience signals — narrows what the algorithm can test. In Learning, you want maximum exploration, not control.
Maximize Conversion Value without tROAS lets PMAX find conversion paths you'd never predict manually. Add the constraints once you have enough data to know what to constrain.
Days 1–7: If no impressions or clicks → check feed approval, bid floors, product disapprovals in Merchant Center. The campaign isn't even in the auction.
Days 7–14: Impressions but no conversions → diagnose the store (CVR, speed, pricing) before blaming the ad.
Day 14+: Conversions flowing → now you can analyze ROAS meaningfully. Sample size matters.
Campaign performance is downstream of store performance. Before scaling Google Ads, audit each of these:
Google Ads is an amplifier, not a fixer. A great campaign on a broken store just buys you faster losses. Audit these eight levers before touching ad spend — fixing them costs nothing compared to bleeding money on traffic that can't convert.
When the campaign stalls or signals start to emerge, run the relevant check below. Each is a self-contained diagnostic — pull the lever that matches the symptom. Click any check to expand.
Optimize across three levels. Refine the foundation. Aim for the $100K/month milestone.
PMAX handles discovery and cold demand. Branded search protects your owned demand from competitor bidding (cheap, high-intent traffic). Standard shopping bestsellers isolates your proven winners with manual bid control, capturing volume PMAX leaves on the table.
Together they cover the full intent funnel — without overlap, since each is targeted at a distinct audience state.
The Labelizer is a script that tags every SKU by performance archetype — zombies (high spend, no return), turtles (slow but profitable), ghosts (impressions, no clicks), birds (top performers). It turns an opaque PMAX into a structured, manageable asset.
Budget increases follow a tiered, rules-based model. Maximum scaling cap: 25–30% per move.
| Daily Budget | Average Scale | Increment |
|---|---|---|
| €200 — €500 | +50% | 1–2× per week |
| €500 — €1,000 | +30% | 1–2× per week |
Implementation runs on exact <30-day data. Increase or lower in 10% or 20% increments — never larger.
The principle: aggressive moves break learning. Each adjustment resets a small piece of the algorithm. Smooth and frequent beats big and rare.
Bundles, navigation refinements, considered color use, social proof through reviews, strategic discounting.
Run these as controlled A/B tests with a single variable each. A 0.3% CVR lift compounds across every euro of ad spend you'll ever run — the leverage is enormous.
Three parallel diagnostics for when the campaign hits a ceiling. Each addresses a different root cause — feed health, product mix, or audience expansion. Click any check to expand.
Discover. Experiment. Find a custom fit. Scale efficiently. Maximize ROI.
Phases 1 and 2 are about execution discipline. Phase 3 flips the brief — the proven engine is humming, so the work now is portfolio management: where to place exploratory bets, when to kill them, and how to find the next compounding lever before the current one decays.
Experimentation costs money you won't recover on every test. With less than 20% margin, the cushion isn't there — one bad quarter and you're under water. The 20% floor is what makes the testing portfolio survivable.
PMAX is the cash engine that funds the experiments. If it slips below 2.7, you cut experimentation and put the resources back into stabilizing the core. Experiments are a privilege earned by performance, not a default state.
Treat the experiments as a VC portfolio — most won't return. Set a clear kill threshold (e.g. ROAS < 1.0 after 30 days at €50/day) and stick to it. The winners pay for the failures, and one breakout channel can reshape your entire revenue mix.
The levers haven't changed since Phase 2 — but the granularity has. tROAS adjustments are now data-driven to the percentage point. Asset groups are restructured around actual performance clusters, not hypotheses. Labelizer outputs feed weekly review cycles.
By this phase, revenue, budget, and ROAS are highly consistent. The work is methodical, not heroic.
Boredom. The system works. The temptation is to "improve" things that don't need improving, which usually breaks them. At this stage, discipline beats creativity. Save the creative energy for the experimentation portfolio.