Ads deliver value, but profitability and PMax's contribution still need checking

Store profile

The assortment spans several product categories; the niche and category names are withheld. Merchant Center has 3,464 approved products. Usable claims: the brand and verified SKU, material and size. Verify first: delivery from UAH 2,000, own manufacturing, “official store”, “premium” and “safe materials”. Do not claim without proof: “No. 1”, “the best”, medical benefits or fictitious discounts.

Campaigns and bidding

Table: scroll horizontally to see every column.

The general-assortment campaign's actual value per UAH of ads is below target. Category A and B campaigns are near their targets. Keep current budgets for now. Before changes, obtain margins and account for conversion lag. Also assess new-customer acquisition cost. Allow enough time after previous changes. Historical campaigns are considered separately: 17 paused and 8 removed campaigns spent UAH 0.00 in the audit period.

Products and conversion lag

The product sample contains all 50 rows returned by the query. It covers part of the catalog, so findings from this table cannot automatically be applied to all products.

Table: scroll horizontally to see every column.

Conversion lag: under 1 day: 88.01%; 1–3 days: 4.81%; 3–7 days: 2.84%; 7–14 days: 1.51%; 14–30 days: 1.84%; 30–90 days: 0.99%. The boundaries match Google Ads buckets. Shares were calculated across all 51 rows of four active campaigns (Source 1). Do not treat results from the latest 1–3 days as equivalent to data whose conversion lag has passed.

Devices, days and geography

Table: scroll horizontally to see every column.

Device totals reconcile with the account after rounding. Computers deliver a higher return but have a small share of spend. We recommend checking how easy it is to buy on a phone. We do not propose device bid adjustments for PMax or automated bidding. Saturday: UAH 12,480.19 spend, 73.78 conversions and UAH 103,534.04 value; value per UAH of ads was 8.30. Wednesday: UAH 12,506.63 spend, 77.03 conversions, UAH 101,586.43 value and 8.12 value per UAH of ads. Sunday: value per UAH of ads was 5.54 and cost per conversion was UAH 257.86. This comparison alone does not justify restricting the PMax schedule. The data cover all four active campaigns (Source 2). By hour: at 12:00, spend was UAH 7,524.94, with 33.74 conversions and UAH 45,419.21 value (6.04 value per UAH of ads); at 20:00, UAH 3,977.06, 25.97 conversions and UAH 50,937.03 value (12.81 value per UAH of ads). The comparison aggregates the same hours over the entire period in the account time zone. It is not one day's result. Lower noon return is a reason to check who clicks the ads and how they buy on a phone. It does not prove causality or justify disabling PMax hours. Source 2.

Table: scroll horizontally to see every column.

The city table combines rows for three PMax campaigns from the sample of 20 cities with the highest spend in the Source 3 query. This is a partial city sample, not the account's full geographic total. City and region data were considered separately. For Dnipro, first check logistics and margin, then run a limited test. Whether uncontrolled territories are excluded from ad delivery is unconfirmed. We therefore do not call the geographic settings erroneous.

Measurement and GA4

Google Ads optimizes bids only for purchases from GA4. Each conversion after a click is counted separately. Attribution is data-driven, with a 90-day window. Adding to cart and beginning checkout are secondary goals. They do not train bidding.

In GA4, we checked the site's data stream [store domain], UAH currency, time zone and link to [associated ad account]. The system counts each occurrence of key events. Technical event names, settings and the time zone are retained in the appendix.

For visits from Google ads, GA4 reports 10,845 sessions and UAH 557,719.00 purchase revenue. Google Ads reports UAH 617,556.54 conversion value. The 9.7% difference alone does not prove a measurement error. The systems attribute ad contribution differently. The channels report was limited to 30 rows.

In the GA4 item report, category is undefined for 1,469 units sold with UAH 1,133,770.00 revenue. These are product units, not order counts. Missing categories prevent full comparison of sales by category.

A developer needs to set up product category data in all ecommerce events. Then check the transaction number, amount, currency and absence of duplicate purchase counting in GA4. Technical fields and the validation tool are listed in the appendix.

PMax, creative assets, extensions and brand

Own-brand overlap

In the own-brand Search campaign, “Brand A” queries generated 15.00 conversions and UAH 25,419.00 value; “BrandA” generated 1.00 and UAH 4,000.00. The campaign also entered auctions for non-brand queries: “[non-brand query A]”, “[non-brand query B]”, “[non-brand query C]”. Its impression share <10% is therefore not an exact measure of own-brand query coverage. We could neither confirm nor rule out PMax taking conversions from Brand Search. Check own-brand exclusions across all PMax campaigns, queries with different brand spellings, new-customer acquisition settings and product overlap. Until then, we cannot claim that PMax generates additional sales or increase budgets solely on average value per UAH of ads. We reviewed 30 PMax asset groups, 12 active. Network labels in their names do not prove Display Network delivery. The original labels are in the appendix. Search themes: “[category theme A]”, “[category theme B]”, “[category theme C]”; there is also [audience A]. These signals help the system learn but do not restrict ad delivery to the specified audience. An active general-assortment PMax group “[active group A]” has Google's “Excellent” rating. Its original name is in the appendix. Anonymized copy includes “[Category A] premium”, “Brand A — [comfort claim]”, “[Natural material] for [audience]”, “[Category A] for [audience] — Brand A”; long headlines include “[Material and product purpose] — Brand A provides [safety and comfort claim]” and “[Category A] premium quality. Free Nova Poshta delivery from UAH 2,000”. Keep specific category wording; use “premium”, “safe” and delivery claims only after checking them on store pages. The “Excellent” rating does not measure ad performance. We do not assign conversions to individual ad assets. The own-brand Search ad has Google's “Poor” rating. Its original name is in the appendix. Anonymized excerpts: “Brand A — [Category A]”, “Brand A | Official store”, “Everything for [target need] at Brand A”; description: “Brand A — [Category A] from the manufacturer. Free Nova Poshta delivery from UAH 2,000.” This rating applies to the whole ad, not one headline. Check the claims against store pages. At account level, 8 sitelinks, a structured snippet and a promotion are active. Callouts, phone and business details exist at campaign level. We account for inheritance and do not claim that sitelinks are missing. Only impressions are available for placements.

Table: scroll horizontally to see every column.

Spend and conversions for these placements are unknown. Impressions alone do not justify recommending exclusions.

Merchant Center

Merchant Center approved 3,464/3,464 products. Rejected, pending or limited products: 0; product-level issues: 0. At account level, the company address is missing and rectangular and square logos do not meet requirements. Add the address and compliant logos. Approved products do not need resubmission for review. A missing barcode or “out of stock” status in isolated examples does not itself prove a defect. A supplemental feed must preserve the product identifier, language and primary feed label. Technical names are in the appendix. Build product labels after obtaining margin data. Add only genuine barcodes.

Search phrases and draft ad copy

SerpApi search patterns: “[category + manufacturer]”, “[category + buy intent]”, “[category + city]”, “[category query about a set]”, “[category query about a bundle]”, “[category + audience qualifier]”.

Table: scroll horizontally to see every column.

Replacement copy is a test draft. Verify brand fit, store-page evidence and legal validity of the claims before using it.

Creative concepts

• 1.91:1 1200×628: show a real Brand A product in use [category withheld], one focal subject, studio lighting and space on the right; on-image text “Brand A”; overlay “[comfort claim for the audience]”.
• 1:1 1200×1200: show a real product close up [colour, category and size withheld], warm lighting, one focal subject and clear space; on-image text “Brand A”, “[product size]”; overlay “[everyday-use property]”.
• 4:5: show a real boxed set [category withheld], one focal subject, 1–2 out-of-focus supporting objects and clear space; on-image text “Brand A”; overlay “[gift for the intended occasion]”.

For every image: exclude competitor logos, celebrities, simulated interfaces and before/after comparisons. Do not show impossible properties, Cyrillic in generated artwork or products in a catalog lineup. These are drafts. Verify brand fit and legal validity of the claims before use.

GA4 audiences

After checking product category data, consider GA4 audiences for purchasers over 180–540 days; cart abandoners over 30 days; checkout abandoners over 14 days; and product viewers without purchase over 14–30 days. Send these audiences through the existing Google Ads link. Add a category signal only to the relevant group. Uploading a customer list through Customer Match requires consent verification and store-owner approval.

Plan

Table: scroll horizontally to see every column.

Change controls

Table: scroll horizontally to see every column.

Campaign structure

Keep the own-brand Search campaign and remove queries without the store name. Keep PMax campaigns until product overlap is checked. Consider consolidation only after analysing sales economics. Fix company details in Merchant Center while preserving product approvals.

Budgets

Table: scroll horizontally to see every column.

Plan by week

• Weeks 1–2: measurement, Merchant Center and own brand.
• Weeks 3–4: complete product classification and the product allocation map across campaigns.
• Weeks 5–7: run one test.
• Weeks 8–10: remove weak elements after accounting for conversion lag.
• Weeks 11–13: scale only with margin data and stable results after changes.

Do not make more than 1–2 major changes at once.

Review coverage

• 4/4 active campaigns
• 50 product rows, part of the catalog
• 30 asset groups/12 active
• 100 own-brand search queries, complete sample
• GA4 channels, partial 30-row sample.

Totals were reconciled again. Active and historical data, and city and region data, were considered separately.

Source types

Google Ads reports by campaign, conversion, product, device, geography, day and hour, conversion lag, asset group and search query; GA4, Merchant Center and SerpApi. Internal identifiers and account access details are not published.

Evidence sources

• Google Ads conversion lag report across 51 rows of four active campaigns.
• Google Ads day and hour report for four active campaigns.
• Partial sample of 20 cities with the highest spend for three PMax campaigns.

Appendix: technical labels from the source data

These labels are retained to check settings and data transfer.

• CPA — cost per conversion. ROAS — conversion value per unit of ad spend, not profit.
• D-9 — label for the PMax and own-brand check in the source report.
• purchase — purchase, the primary Google Ads goal from GA4.
• MANY_PER_CLICK — counts each conversion after a click.
• DDA — data-driven attribution of ad contribution.
• add_to_cart and begin_checkout — adding to cart and beginning checkout, secondary goals.
• ONCE_PER_EVENT — counts each occurrence of a GA4 key event.
• Europe/Kiev — time zone of the checked GA4 data stream.
• google/cpc — source of visits from Google ads in the GA4 report.
• (not set) — product category is undefined.
• itemsPurchased and itemRevenue — product units sold and their revenue.
• item_category in the items array — product category in ecommerce events.
• transaction_id, value, currency — transaction number, amount and currency. Validation in DebugView must also rule out duplicate purchase counting.
• +GDN/-GDN — labels in group names that do not prove Display Network delivery.
• EXCELLENT and POOR — Google's original rating names for “Excellent” and “Poor”.
• IS — impression share.
• tROAS — target conversion value per unit of ad spend in the bid strategy.
• GTIN=null and out_of_stock — missing barcode and “out of stock” status in isolated examples.
• offer_id, language, feed_label — product identifier, language and feed label that the supplemental feed must preserve.
• In the source plan, effort for the address and logos is labelled S. For product category, own brand in PMax, product overlap and Dnipro, it is M. Dnipro has medium priority; the other actions have high priority.