§1 The account, in context
This account belonged to Coffeehit, a UK specialty coffee retailer I managed at Nikos Media. 18.85 ROAS sounds like a number worth bottling and selling — but context matters more than the headline. This was a mature brand with strong repeat-purchase behaviour, three years of first-party data, and an average order value that made the unit economics forgiving of imperfect targeting.
None of that is a knock on the result. It's the reason I'm writing this up as a dissection, not a template — the goal is to separate what was strategy from what was simply a favourable starting position.
§2 What the data actually showed
| Metric | Before | After 6 months |
|---|---|---|
| Account ROAS | 4.2 | 18.85 |
| Cost per acquisition | £18.40 | £6.10 |
| Branded search share | 22% | 51% |
| Returning-customer revenue | 34% | 58% |
The single biggest mover wasn't a bidding trick — it was the shift in branded search share. As campaigns matured, a growing share of "purchase intent" traffic was already brand-aware, which is exactly the kind of traffic that converts at high ROAS regardless of platform sophistication.
§3 Audience layering and bid strategy
We layered first-party customer-match audiences on top of standard shopping campaigns and let Google's automated bidding optimise toward a target ROAS once there was enough conversion volume to do so reliably — generally 30+ conversions per campaign per month, as a rule of thumb.
- Customer-match lists segmented by purchase recency and order value
- Separate campaigns for branded vs non-branded search, budgeted independently
- Smart Shopping / Performance Max layered carefully to avoid cannibalising branded search
- Negative keyword lists reviewed weekly, not "set and forget"
§4 Creative and landing page alignment
High-ROAS accounts share one quiet trait: the ad promise and the landing page experience match almost exactly. We rebuilt several landing pages so the product shown in the ad was the first thing visible above the fold, with consistent pricing and imagery — removing the seconds of doubt that cause high-intent clicks to bounce.
§5 What can — and can't — be replicated
An 18.85 ROAS on a mature, repeat-purchase brand with strong first-party data isn't the same achievement as an 18.85 on a new account with no purchase history. If you're benchmarking your own account against this one, benchmark the inputs — audience maturity, AOV, repeat-purchase rate — not just the output.
§6 Takeaways
The transferable parts of this account were the audience-layering discipline, the creative-to-landing-page alignment, and the weekly negative-keyword hygiene. The parts that won't transfer cleanly are the brand maturity and AOV — which is exactly why "what worked for account X" is rarely a safe instruction to follow literally.