Explains why backend, GA4 and Google Ads revenue differ with a bridge of causes — tax, shipping, refunds, consent, attribution, time zones — from supplied numbers only.

Reads out an A/B test honestly: SRM check, sample size against the plan, significance and intervals, guardrail metrics and novelty effect before any "winner" call.

Validates Product, Offer, AggregateRating, shipping and return-policy JSON-LD against Google rich result requirements and the visible page, and reports every mismatch.

Codes return reasons from free text into root causes (size/fit, damaged, not as described) and links each cluster to a concrete product page or packing fix.

Compares your prices with competitor data you supply, matched by EAN, and recommends moves that respect your margin floor. Uses only provided or public data.

Reviews a product detail page element by element — images, price, variants, delivery, returns, reviews — and writes testable fixes, never invented product claims.

Translates thousands of SKUs in resumable batches with a glossary and do-not-translate list, localises units, sizes and currency, then runs a QA pass before import.

Segments lapsed customers by RFM and their own purchase cycle, then plans a win-back sequence per segment with honest offers taken from the store's real policy.

Reviews Performance Max asset groups and listing groups for overlap, thin assets and all-products catch-alls, and recommends restructures without editing the account.

Computes break-even and target ROAS from contribution margin after COGS, fees, fulfilment and returns, using only the store's numbers and showing every step.

Plans staged markdowns from weeks of cover and sell-through, never below the margin floor you set, with EU 30-day lowest-price references on every reduction.

Drafts review-request mails timed after delivery, not purchase, asking every customer the same way — no incentives for positive reviews, no review gating.

Builds a one-page weekly store report — revenue, orders, AOV, conversion rate, ad spend, MER and returns — against last week and last year, stating any missing source.

Mines the Shopping search terms report for wasted spend and new winners, proposing negatives by match type for a human to apply, never pushing them itself.

Audits GA4 ecommerce events against the recommended schema — required params, items[] fields, currency and transaction_id dedupe — and reports gaps without editing tags.

Triages Merchant Center disapprovals by reason and impact, maps each to the attribute or page fix, and proposes feed changes only for a human to approve.

Maps a store catalog to Heureka, Zboží.cz and Idealo XML/CSV feeds with correct elements, categories and delivery days, and reports rows it cannot fill from real data.

Pulls structured attributes (color, size, material, GTIN, dimensions) out of messy titles and descriptions into feed-ready columns, with a source and confidence per value.

Drafts "where is my order" replies from real order and carrier tracking data, states delays plainly, and never promises a delivery date the data does not support.

Walks the checkout step by step against known friction points — forced accounts, late shipping costs, surplus form fields — and ranks fixes by impact and effort.

Adapts one master product record into Amazon, Allegro and Kaufland listings within each title/bullet limit and category attribute set, flagging every missing value.

Reads a GA4 export and finds where the checkout funnel leaks — drop-off by step, device and source — with the one fix to try first.

Writes category page intros and buying-guide blocks from real product data and search queries — short above the grid, useful below it, no keyword stuffing.

Writes product page descriptions from the spec sheet in the store's voice — benefit first, facts only from the data, missing specs flagged instead of guessed.