Five real brands. Real engagements.

Founder-led Shopify fashion brand running a mixed paid stack across Meta and Google Shopping. The dashboards reported double-digit ROAS on parts of the spend that, on Shopify, were not showing up as new orders. Standard founder dilemma: the platforms looked great, the bank account did not.

First causal read on the GA4 export disagreed with both Meta and Google. Two audience segments inside Meta were absorbing roughly 2K EUR per month while the causal model attributed essentially no incremental revenue to them. They were re-buying conversions the brand was already getting through organic and direct.

We paused the two audience segments and reallocated to Google Shopping at a corrected target ROAS. The webstore was mid-migration to a new platform at the time, which gives the 2025 lift number an additional headwind context. Q1 2026 reflects the brand acting more aggressively on subsequent causal recommendations.

Founder considering pulling Pinterest entirely. Last-click in GA4 showed Pinterest taking credit for a small share of orders; the dashboards were arguing for cutting it. The intuition was: Pinterest is browser traffic that converts later, and the click does not get attributed.

Causal analysis on the brand's GA4 + Shopify data resolved the debate cleanly: Pinterest was driving incremental sales the last-click report missed. The conversion path was running through other touchpoints by the time the order landed, but the causal model isolated Pinterest as the channel actually moving the needle on a chunk of weekly orders that would not have happened without it.

The brand kept Pinterest live and rebalanced the rest of the mix. The 2024 revenue lift below reflects the year's cumulative effect of rejecting the obvious dashboard read and acting on the causal one instead.

Brand was scaling paid acquisition on Meta and watching last-click ROAS hold steady while gross profit dropped. The platform report read like growth; the P&L read like a leak.

The causal read surfaced what the last-click view structurally cannot: customers acquired through Meta were returning product at a materially higher rate than other channels. The platform credit-allocated revenue was real at point-of-sale and gone again by the time the returns processed two weeks later. Net-of-returns ROAS on Meta-acquired customers was substantially below the headline.

We rebuilt the bidding around net contribution rather than gross revenue and capped the Meta-acquired customer cohort at a defensible lifetime value. The 2024 revenue lift is the cumulative effect of stopping the hemorrhage and putting the spend behind cohorts the causal model identified as genuinely profitable.

Influencer-driven launch cadence on a Shopify storefront. The founder's question was structural: when a launch goes well, how much credit belongs to the influencer push versus everything else running concurrently? Without a causal answer, the brand was either over-paying influencer programs or pulling them too early.

Honest finding: at the brand's then-current paid + organic volume, the causal model could not isolate the influencer-specific incremental contribution with confidence intervals tight enough to defend a decision against. The data was not yet deep enough for the model to separate signal from noise on that specific question.

Our recommendation was to skip the continuous tier and revisit when paid spend and order volume scaled. This engagement is one of the reasons we publish a fit floor on every money page: causal attribution earns its keep at roughly EUR 5K+/month paid spend with a meaningful GA4 history. Below that, the brand's existing analytics already do most of the job.

Customer brought us in to audit channel-level attribution under the working hypothesis that meaningful revenue was being misattributed across the paid + organic split. Real engagement, GA4 + shop export, full causal read.

Honest finding: at the brand's then-current spend level, the channel-level gap the causal model surfaced sat within the confidence interval. The signal was there, but it was not yet significant enough to defend a meaningful reallocation decision against. Their existing analytics stack was already telling them most of what a EUR 99 causal read could add at that scale.

Our recommendation was the same as Vivid Merve: skip the continuous tier, revisit when paid spend scales. This engagement is one of the reasons we publish EUR 5K+/month paid spend as the floor of our fit qualifier rather than pretending the product fits every Ecommerce brand.