AED 11.20 per conversion is what the Printo Google Ads account settled at: 3,750 conversions on AED 42K of total spend. This post is the full system behind that number — the attribution work that found where roughly AED 11K a month was being misallocated, the RSA testing that produced a 5.83% CTR ad over 82K+ impressions, what the autonomous agent handled versus what stayed manual, and the tactics that didn't work. No cherry-picked screenshots; the point of publishing a named case is that the numbers hold together in context.
The starting point
Printo is a printing and e-commerce business — a price-sensitive mix of B2B and B2C demand, high query volume, and lots of product variety. When I took the account on, the structure had grown organically: overlapping campaigns competing for the same queries, budget concentrated where it had historically been set rather than where returns were, and conversion signals that told an incomplete story about which spend actually produced customers. I'm deliberately not quoting a baseline CPA figure — the honest framing is directional: spend was drifting toward what was easy to buy, not what was worth buying.
The strategy: attribution before bidding
The first serious work wasn't bidding — it was measurement. Rebuilding the attribution picture across Ads and analytics showed that a meaningful slice of spend was buying conversions that looked fine in-platform but didn't hold up downstream. That modelling identified roughly AED 11K per month of spend worth reallocating — away from campaigns winning on last-click vanity and toward the queries and products where conversions became actual revenue. Only after that did bidding strategy changes make sense; bid optimisation on top of misallocated budget just gets you to the wrong place more efficiently.
Campaign structure followed the same logic: consolidating overlap so signal pooled instead of fragmenting, and giving Smart Bidding clean conversion actions to optimise toward. The bidding automation was Google's; the judgment about what it should chase was not.
The RSA work: how one ad reached 5.83% CTR
The top responsive search ad reached 5.83% CTR across 82K+ impressions, and it got there through boring rigor rather than copywriting genius. The approach: treat RSA assets as a testing portfolio, not a finished ad. Headlines were written to cover distinct angles — price, speed, product specificity, trust — rather than ten rephrasings of one idea, so asset-level reporting could actually distinguish what pulled. Weak assets got replaced on a regular cadence based on performance ratings and impression share, not gut feel. What won, consistently, was specificity: headlines that named the product and the concrete promise outperformed clever generic lines every round.
Where automation took over
ONE TACTIC A WEEK
Partway through the engagement, SARA — my autonomous Google Ads optimisation agent — took over the mechanical layer: search term harvesting and negative keyword additions, budget pacing adjustments, asset rotation flags, and anomaly alerts, all inside guardrails with human sign-off above a spend threshold. The architecture lives in its own post, so I'll keep this to the division of labour: SARA did the daily reading and routine execution; the attribution modelling, restructure decisions, and RSA angle strategy were manual. This result was system plus strategist, and I'd distrust anyone who claims a comparable number was 100% autonomous.
The numbers, in full
The complete picture: AED 42K in total spend across the engagement produced 3,750 conversions — a blended CPA of AED 11.20. Blended matters: that's the average across all campaigns, with brand and high-intent product terms cheaper and colder prospecting more expensive. Roughly AED 11K per month of spend was reallocated based on the attribution modelling. The top RSA ran at 5.83% CTR across 82K+ impressions. Read together, the numbers describe an account where measurement discipline and asset testing compounded — not a single trick.
One caveat worth stating plainly: what counts as a "conversion" shapes how any CPA reads, and the honest way to present a blended figure is alongside its spend and volume, which is why I keep repeating the AED 42K and 3,750 together. A CPA quoted without its denominator is marketing, not measurement — and that standard applies to my numbers as much as anyone's.
What didn't work
Credibility requires the misses. Broad-match expansion experiments on colder product categories burned budget on queries the negatives list couldn't keep up with, and got rolled back. An early attempt to push aggressive target-CPA values on thin-data campaigns throttled volume without improving efficiency — the targets had to be loosened and re-tightened gradually. And some ad-copy angles I was confident about — clever, brand-forward lines — repeatedly lost to plain specificity. A real account history includes abandoned tactics; a highlight reel doesn't.
Lessons for other UAE advertisers
What's replicable: attribution discipline before optimisation, structural consolidation so bidding gets clean signal, RSA testing as a portfolio with distinct angles, and automating the mechanical layer so strategist hours go to decisions only humans can make. What's not replicable: the number itself. AED 11.20 reflects Printo's vertical — printing and e-commerce with high query volume and a transactional conversion mix. A B2B account with a six-month sales cycle should expect a very different CPA, and comparing across verticals is how advertisers get sold fantasies. The methodology travels; the benchmark doesn't.