Margin-first
eCommerce growth
Blended ROAS hides the products losing you money and the discounts buying revenue you already had. We rebuild feeds, category architecture and measurement so growth decisions are made on contribution margin instead.
What actually limits eCommerce growth
eCommerce performance marketing is constrained less by ad creative than by data quality and margin visibility. Product feed accuracy determines which items can be shown at all, category page architecture determines organic discoverability, and contribution-margin reporting determines whether scaling a campaign builds profit or buys revenue at a loss.
Retail search is a feed problem first
Before creative, bidding or budget, the question is whether your products are eligible, correctly categorised and richly enough described to be matched to a query at all.
In most retail accounts we audit, a meaningful share of the catalogue is either disapproved, missing GTINs, using manufacturer boilerplate as the description, or sitting in the wrong Google product category. Those items cannot win impressions regardless of how the campaign is structured, and Performance Max will quietly spend the budget on the products that are already selling.
The organic half is a mirror image. Category and collection pages are where non-branded retail search lands, yet they are usually thin templates with a product grid and no unique content, while faceted navigation generates thousands of near-duplicate URLs that dilute crawl budget. Fixing category architecture typically returns more than another round of product-page tweaks.
Relative revenue contribution, indexed. Directional pattern across Oneskai retail engagements; your split is measured against your own analytics during the audit.
A retail purchase is fast and comparative
Consideration is compressed into hours or days, and price comparison happens in another tab while your page is still open.
Need or prompt
Triggered by a replacement need, a seasonal moment or an ad. The query is descriptive rather than branded, which is why category page quality decides whether you enter consideration.
Scan and filter
Buyers compare on price, delivery speed, returns policy and review volume simultaneously. Missing delivery information at this point removes you from the set more often than price does.
Validate
Reviews, sizing detail, material specifics and returns terms. Thin product content forces the buyer to leave and validate on a marketplace, where a competitor often captures the sale.
Friction test
Unexpected shipping cost, forced account creation and limited payment options are the dominant abandonment causes — all of which are fixable and none of which are advertising problems.
Repeat or churn
Retail economics depend on the second order. Lifecycle flows, replenishment timing and post-purchase content determine whether acquisition cost is recovered once or many times.
The reporting change that changes decisions
Most retail growth problems trace back to a measurement choice made early and never revisited.
Blended ROAS as the north star
- One ROAS target applied across every product and margin tier
- Branded search revenue counted as acquisition performance
- Discount codes treated as growth rather than margin transfer
- Feed treated as a technical task owned by nobody
- Category pages left as thin templates with a product grid
- Returns and shipping cost excluded from channel profitability
Contribution margin per product tier
- Margin-banded targets, so low-margin SKUs cannot hide behind hero products
- Branded and non-branded separated before any channel is judged
- Discount cost modelled as an acquisition expense, because it is one
- Feed monitored continuously with disapproval and coverage alerting
- Category pages built as ranking assets with genuine unique content
- Returns rate and fulfilment cost included in channel-level reporting
Retail channels, and the honest caveats
Every channel below earns its place from the audit findings, not from a standard package.
| Channel | Job in the funnel | Primary KPI | When we do not recommend it |
|---|---|---|---|
| Product feed management | Make the catalogue eligible and correctly matched | Approved SKU coverage | Never — this is foundational for every retail account |
| Performance Max | Scale demand across Google surfaces | Contribution margin | Feed quality is poor; PMax amplifies bad data |
| Category & faceted SEO | Capture non-branded discovery organically | Non-branded organic revenue | Catalogue is under a few hundred SKUs |
| Paid social prospecting | Create demand for visually distinctive products | New-customer CAC | Product is commodity and competes only on price |
| Lifecycle email & SMS | Recover the second and third order | Repeat purchase rate | Consumable replenishment cycle exceeds 18 months |
| Marketplace presence | Meet buyers already validating elsewhere | Incremental revenue | Margin cannot absorb marketplace commission |
| On-site CRO | Remove checkout and validation friction | Add-to-cart to purchase rate | Traffic volume too low for reliable testing |
Performance Max deserves its caveat. It is an amplifier: pointed at a clean feed with correct margin signals it works well, pointed at a broken one it spends efficiently on the wrong products.
Reported on profit, not revenue
Revenue-based reporting is the reason unprofitable retail accounts look healthy for three quarters before anyone notices.
Margin bands, not one target
A 3× ROAS target is generous on a 70% margin product and loss-making on a 25% one. Targets are banded by margin tier.
Returns belong in the maths
Paid social frequently drives higher return rates than search. Excluding returns makes the worse channel look better.
New customers, counted separately
Repeat revenue inflates acquisition performance. We split it out before judging any prospecting channel.
Retail reporting line
- Contribution margin after ad cost
- The number scaling decisions are made on
- New-customer CAC
- Separated from repeat revenue entirely
- Non-branded organic revenue
- True discovery performance
- Approved SKU coverage
- Share of catalogue actually eligible to serve
- Repeat purchase rate at 90 days
- Whether acquisition cost gets recovered
- Return rate by channel
- Channels differ sharply and it is rarely tracked
- Blended CAC to LTV ratio
- Reviewed quarterly, not monthly
- Discount depth as % of revenue
- Margin leakage, made visible
Sequenced so data is fixed first
Optimising campaigns on top of a broken feed and unreliable margin data is expensive guesswork.
Feed and data audit
Full catalogue diagnostic: disapprovals, missing GTINs, category mapping, attribute coverage and title structure. Margin data joined to product records so profitability can be measured at SKU level.
Measurement rebuild
Branded and non-branded separated, contribution margin wired into reporting, returns and fulfilment cost attributed by channel. This is the work that changes which decisions get made.
Campaign restructure
Margin-banded targets, hero and long-tail product segmentation, and negative sculpting so branded queries stop inflating acquisition performance.
Category architecture
Collection pages rebuilt as ranking assets, faceted navigation indexation controlled, and internal linking resolved so crawl budget reaches commercially useful pages.
Lifecycle and margin defence
Post-purchase flows to recover acquisition cost, continuous feed monitoring with disapproval alerting, and quarterly discount-depth review.
Scope, stated plainly
Our retail experience concentrates on mid-market direct-to-consumer and multi-brand retailers on Shopify and headless stacks. We are not the right partner for enterprise marketplace sellers operating primarily on Amazon, and we will say so early rather than take the engagement. Anonymised feed diagnostics and margin-reporting structures from live accounts can be walked through on a call.
If your growth constraint is Amazon buy-box strategy rather than owned-store performance, a specialist marketplace agency will serve you better than we will.Oneskai retail scope statement
eCommerce & Retail questions
Why does the product feed matter more than ad creative?
Because the feed decides eligibility. A product with a missing GTIN, a disapproval or the wrong Google category cannot be matched to a query no matter how good the creative is. In most catalogues we audit, a meaningful share of SKUs are ineligible or under-described, so feed remediation returns faster than any campaign change.
Is Performance Max worth running?
It is worth running on a clean feed with reliable margin signals, and actively harmful on a broken one. PMax optimises toward the conversion data it receives, so if branded traffic is counted as acquisition and low-margin products are treated as equal to high-margin ones, it will scale exactly the wrong thing very efficiently.
How do you handle faceted navigation and duplicate content?
By deciding deliberately which filter combinations deserve an indexable URL and which do not. Combinations with genuine search demand — colour plus category, for instance — get canonical, crawlable pages with unique content. The long tail of filter permutations is blocked or canonicalised so crawl budget reaches pages that can actually rank.
Should we discount to hit growth targets?
Discounting reliably moves revenue and frequently destroys margin, because a large share of redemptions come from customers who would have bought anyway. We model discount cost as an acquisition expense and report discount depth as a percentage of revenue so the trade-off is visible rather than assumed.
What size catalogue does this approach suit?
Feed and measurement work pays back at almost any size. Category and faceted SEO becomes materially valuable above roughly a few hundred SKUs, because below that the number of viable category queries is small enough that a handful of well-built pages covers the opportunity.
Do you work with Shopify, or custom platforms too?
Both. Most of our retail work runs on Shopify and Shopify Plus, with a smaller share on headless builds using commerce APIs. Platform matters less than data access — what we need is reliable margin data at SKU level and the ability to control indexation and templates.
How quickly do feed fixes show results?
Faster than most retail work. Disapproval and eligibility fixes can affect impression volume within days of the next feed fetch. Title and attribute improvements typically show within two to four weeks. Category page SEO is the slower half, generally three to six months to meaningful non-branded revenue.
Related capabilities
Sources & references
- Google Merchant Center Help, product data specification — required and recommended attributes.
- Google Ads Help, Performance Max best practices and asset group documentation.
- GS1, GTIN standards and allocation rules for retail product identification.
- Google Search Central, “Faceted navigation best practices” — indexation control guidance.
- Google Search Central, structured data for Product, Offer and AggregateRating.
- Baymard Institute, published research on checkout abandonment causes.
Find out what your catalogue is hiding
A feed and margin audit covering catalogue eligibility, category mapping, attribute coverage and contribution margin by product tier — with the disapprovals and under-served SKUs listed individually.