520% average ROAS after a PrestaShop rebuild, tracking fix, and automation work

Meca Express, run by Vincent Junillon, sells catalytic converters and related automotive parts online through two French storefronts: e-catalyseur.fr and meca-express.fr. Both sit under the same company name. The commercial motion is specialised e-commerce — catalogue depth, fitment-sensitive demand, and a paid-search budget large enough that bad measurement is expensive.

At the time of the engagement the stack was PrestaShop on an outdated PHP version. That is not a cosmetic detail. An old PHP runtime and an old PrestaShop release constrain checkout reliability, page performance, and what you can safely change in the measurement layer. Google Ads was already live against that foundation, with monthly spend in the roughly €5,000–7,000 range.

The brief was not "fix a broken purchase tag and leave." It was to make the sites fit for modern commerce measurement and then run Ads against a signal that could be trusted. Fadhil Mezouar led the work for Propulse Agency as a direct engagement with the client.

What looked from the Ads UI like a performance problem was three separate problems sharing one P&L.

First, the platform itself. Running PrestaShop on outdated PHP is an infrastructure problem before it is a marketing problem. Checkout paths that feel slow or fragile suppress conversion rate on their own. Template and module constraints make clean event instrumentation harder than it should be. You cannot honestly ask Smart Bidding to recover what the storefront is losing at the moment of purchase.

Second, tracking was not correctly configured. We are not inventing a colourful list of broken triggers — the client facts do not include a defect-by-defect inventory. What we can state accurately is that the measurement layer was not set up correctly for e-commerce reporting into Google Ads. When that is true, the account is managed against a partial view of reality: some purchases under-attributed, some values unreliable, some funnel steps missing. Bidding then optimises against whatever fraction of the truth the tags happen to catch.

Third, site automation was not optimised. Again, the specific workflows are not itemised in the facts we are allowed to publish, so we will not invent them. The operational point is simpler: when catalogue, order, or reporting automation underperforms, paid search inherits the friction — delayed feeds, delayed status changes, delayed reconciliation between Ads and the order book.

Individually, any one of those can move ROAS. Together, they make the Ads dashboard a poor place to diagnose cause. A 380% baseline in that context is not a clean "campaign quality" number. It is the combined output of an ageing storefront, incomplete measurement, and under-tuned automation, with media spend riding on top.

The diagnostic order mattered. We did not open Google Ads first and invent a bidding story.

We started with the commercial reality of both storefronts: how orders were actually completed, what the PrestaShop/PHP versions would and would not allow, and where the checkout path was fragile enough to suppress conversion independent of media. Platform risk had to be named before any media recommendation, because a tracking-only brief would have left the conversion-rate ceiling untouched.

Only then did we inspect the measurement layer — the same class of work we document on our GA4 audit practice: what events reached the ad platforms, whether purchase values looked reconcilable against orders, and whether the funnel was visible enough to manage. The finding was not "Ads is broken." The finding was "Ads is flying partially blind on a storefront that also needs a platform upgrade."

From there we scoped site automation: what was supposed to keep the catalogue and order-side processes current, and where that was underperforming relative to a paid-search operation spending several thousand euros a month. The engagement plan that followed treated platform, measurement, and automation as one combined rebuild window — not three unrelated tickets.

Lead with the representative figure, not the spike. After the combined rebuild went live, Google Ads ROAS moved from a 380% baseline to a 580% peak in the first full month, then settled at a 520% sustained average across the engagement that followed. Monthly spend stayed in the roughly €5,000–7,000 band. The measurement window is the first full month after the website upgrade and corrected tracking went live, versus the period before; the platform upgrade year is 2022.

How to read this number. This result reflects three things changed together — an outdated PrestaShop/PHP platform upgraded, tracking corrected, and site automation optimised — with Google Ads managed on top of that cleaner foundation. It is not a tracking-only before/after. We are presenting it that way because that is how the engagement actually ran: a site on old PHP with incorrectly configured tracking and unoptimised automation is a common pattern, and fixing those layers together is usually what real ROAS recovery looks like in practice.

We do not have a clean split between recovered attribution and genuine performance gain. The same window changed the storefront conversion environment, the conversion signal Ads could see, and the automation feeding the operation. Any attempt to assign "X points to tracking, Y points to platform" would be invented precision. The revenue behind the ROAS was real either way; the honest claim is that the account stopped being managed blind to a broken foundation.

We also do not claim server-side tagging, Meta CAPI, or Enhanced Conversions as part of this engagement — those were not confirmed in the facts for this page. If you are scoping a similar rebuild today and need first-party collection, that is a separate decision under server-side tracking. Consent Mode requirements have moved since 2022; a 2026 rebuild should include Consent Mode v2 in the measurement design from day one.

Most paid-search accounts we inherit are not failing because someone forgot a negative keyword. They are failing because the storefront, the measurement layer, and the operational automation underneath Ads are each contributing a quiet tax — and the Ads UI can only report the combined residue.

If you are running PrestaShop (or any ageing commerce stack) at a few thousand euros a month in Google Ads, the useful question is not "can we squeeze another bidding strategy?" It is whether the platform, the dataLayer, and the automation would survive an honest audit. That is the engagement pattern this case study documents.

Book a free technical audit. We will say whether your ceiling is media, measurement, or the site itself.