What was broken. What we did.
What happened.

B2B e-commerce · Professional equipment · Europe

Rebuilding Shopping for a €20M equipment retailer

The situation. A 30,000-SKU catalog of professional kitchen equipment, with Performance Max campaigns treating a €4,000 oven and a €3 spare part identically. Margins varied wildly across the catalog, but the bidding strategy couldn't see any of it. Impression share was collapsing on hero products while budget bled into low-value accessories.

What we did. Feed surgery first: custom labels for margin tier, stock depth and price competitiveness, applied via feed rules so they survive every catalog sync. Then we split PMax into asset groups by margin tier with separate ROAS targets, carved brand traffic out into its own campaign to expose true incrementality, and rebuilt titles on the top 2,000 SKUs from actual query data.

What happened. Within one quarter, ad-attributed revenue rose 62% on a budget increase of only 18%. High-margin categories absorbed the extra spend; accessories were demoted to a catch-all group with a strict target. The honest footnote: our first brand-exclusion pass was too aggressive and cost ten days of volume before we recalibrated.

Online education · D2C · France

Fixing the funnel, not just the ads

The situation. A wellness training platform with solid courses, healthy traffic, and a ROAS that made no sense. The campaigns looked fine. The landing pages looked fine. The numbers said otherwise.

What we did. Instead of touching bids, we instrumented the full funnel step by step. The data showed 75% of paid traffic abandoning at a single point: a mandatory account-creation step before checkout. No bid strategy on Earth fixes that. We specced and shipped guest checkout, then rebuilt the conversion architecture so Google optimized toward completed purchases — including installment-plan sales tracked through to actual cash collected, not just the first payment.

What happened. Conversion rate tripled at the step we fixed, cost per sale dropped 47%, and — for the first time — the dashboard could distinguish billed revenue from cash actually collected. The lesson we repeat to every client: sometimes the highest-ROI Google Ads work happens outside Google Ads.

The geeky part nobody sees: the fix was a WooCommerce hook, not a bid adjustment. — Aries Star delivery note

Lead generation · Services · APAC

From lead volume to lead value

The situation. A services business generating plenty of leads at a reasonable cost — and a sales team quietly drowning in junk. Google was optimizing for form fills, so Google delivered form fills. Nobody had told the algorithm what a good lead looked like.

What we did. We connected the CRM to Google Ads with offline conversion import, so every lead carried its real outcome back into the auction: qualified, proposal sent, closed-won, with actual deal values. Bidding switched from "maximize conversions" to value-based targets. We also killed two campaigns that produced 30% of lead volume and almost zero revenue — a conversation that required the CRM data to win.

What happened. Same monthly budget, 89% more sales-qualified leads, and a cost per SQL down by a third. Raw lead volume actually dropped 12% — which is the point. The sales team stopped complaining about marketing, which the client described as the real KPI.

E-commerce · France · Aries Labs tooling

Cleaning 2 million search terms, semantically

The situation. A French e-commerce account with a large catalog and years of broad match and Performance Max history had accumulated over 2.1 million unique search terms. Buried in there: thousands of queries with zero purchase intent — competitor brands, DIY tutorials, job seekers, spare parts the client doesn't sell. At this scale, no human can read the report. So nobody had.

Why it matters. Every irrelevant query does double damage. First, the obvious: it burns budget on clicks that can never convert. Second, the insidious one: it poisons the data Smart Bidding learns from. Google's algorithm optimizes toward whatever converts in your history — feed it noise, and it bids on noise. Aggressive negative keyword management isn't housekeeping; it's how you steer an automated bidding system. Cleaner queries also lift CTR and expected quality, which lowers the CPC you pay on the queries you actually want.

What we did. Reading 2.1M terms line by line is impossible, and keyword-level negatives can't keep up with how broad match rewrites queries. So Aries Labs built an in-house tool that clusters search terms semantically — grouping queries by meaning rather than exact wording. Whole intent families ("how to repair X", "X second hand", "X jobs") surface as clusters with their aggregated cost and conversion data attached. A senior strategist reviews each cluster — the judgment call stays human — and approved clusters are converted into structured negative lists, applied across Search and PMax, and re-run monthly as new terms flow in.

What happened. Roughly a quarter of non-brand spend turned out to be going to queries with no realistic path to purchase. Reallocating it did more than save money: with cleaner conversion data, Smart Bidding stabilized and target ROAS could be raised without losing volume. The monthly re-run matters as much as the initial purge — broad match never stops drifting, so neither do we.