Ecommerce advertising is not simply a matter of generating more orders. Your product feed, Merchant Center data, conversion tracking, campaign structure, order value and margins all decide whether spend produces profit or just expensive revenue. We manage that whole system with the business economics in view, not just the account.
Our approach draws on digital marketing experience dating back to 2009. Retained ecommerce accounts are managed directly rather than handed off to a rotating account team once signed — which is the main practical difference in how the work gets done.
Ecommerce advertising is not harder than lead generation — it is a different problem. In lead generation, the account and landing page are most of the system. In ecommerce, decisions interact continuously with products, inventory, pricing, margin and checkout, none of which the account can see.
| Dimension | Lead generation | Ecommerce |
|---|---|---|
| The conversion | Usually a lead — an enquiry, a call, a form | Usually a completed transaction with a value attached |
| Quality control | A sales team assesses lead quality later, off-platform | Product, price, checkout and delivery decide the outcome immediately |
| The core metric | Cost per acquisition usually carries the decision | ROAS, average order value, contribution margin and product-level profitability all matter at once |
| The offer | Often a single fixed proposition | Hundreds or thousands of products with different margins, demand and stock positions |
That last row changes the work. A lead-generation account has one offer to get right. A store has a catalogue where the easiest product to sell is frequently the one that makes the least money — and an automated campaign will find it, favour it, and report the result as success.
Google's own documentation makes the dependency explicit: Shopping and Performance Max campaigns use product attributes and Merchant Center data to match products with relevant searches, and accurate product data plus reliable conversion measurement are treated as prerequisites rather than refinements.
margin · returns · fees
titles · category · GTIN
approvals · issues
Shopping · PMax · Search
purchase · value · GA4
ROAS · AOV · contribution
Profitability decides which products get advertised next — the loop runs back to stage one.
Stages one through three happen before anyone opens Google Ads. Stage six happens in your own records, not in an ad platform.
Most underperforming ecommerce accounts we look at are not badly configured at stage four. They are being run correctly on top of an unresolved problem at stage one, two, three or five, and no amount of campaign work reaches those.
Five checks, in this order, before a bid or a budget changes. Each one can invalidate the analysis that would otherwise follow it, which is why the order is not negotiable.
Selling price, gross margin, habitual discounting, return rate, shipping, and payment or marketplace fees. A high-revenue product is not automatically a high-value one: a ₹4,000 item at 18% margin with one return in eight contributes less than a ₹1,200 item at 55% margin that nobody sends back. Until that distinction is clear, any budget recommendation is a guess.
Titles, product type, Google product category, GTIN, price, availability, images and any outstanding product issues. Rich, accurate product data and active issue monitoring matter partly because disapproved or silently omitted items cannot be advertised at all, which reads identically to poor performance in a campaign report.
The purchase event, transaction value, transaction ID for deduplication, the Google Ads conversion action, GA4, Tag Manager, and what the store itself recorded. Smart Bidding spends against the signals it receives, so a missing or wrong value teaches it the wrong lesson. Correct tracking does not guarantee performance; incorrect tracking prevents it.
Google provides product-level reporting across Merchant Center-linked campaign types, so this needs no custom build. What matters is the split: which products genuinely sell, which consume spend without returning, which are seasonal, and which are out of stock but still running.
Spend, conversion value, ROAS, CPA, average order value and budget distribution. Last, deliberately — these are output metrics, and reading them before the four checks above tells you what happened without telling you why. That is how accounts get adjusted monthly for a year with nothing changing.
Anyone can read a Google Ads dashboard. The work that changes results is reconciling it against product costs and order records — that is where the disagreements sit, and the disagreements are the findings.
Four campaign types with four different jobs. There is no universal mix — the right combination depends on conversion volume, catalogue size, brand awareness and how much of your demand already exists in search. Each has a dedicated page for the implementation detail.
Matched using product data rather than a keyword list, which makes the feed the primary lever. Shopping puts the product, price and image in front of someone already searching for something close to it — usually the highest-intent inventory a store has, and the first place to look when a catalogue is under-monetised.
Uses Merchant Center product data plus assets and audience signals to serve across Google's eligible inventory automatically. It works well and removes most manual control, which is why the inputs matter more than the settings — product data, conversion values and asset quality are what you still decide.
Captures the intent product feeds express poorly — category and comparison queries, brand searches, and the problem-led phrasing people use before they know what product they want. It is also where negative-keyword discipline protects the rest of the account.
Re-engaging people who viewed products or reached checkout without buying. Efficient per conversion and easy to over-credit — it largely harvests demand the rest of the account created, so measure it separately rather than letting it flatter blended performance.
The feed is usually treated as an operations task someone set up once. For a store running product campaigns it is closer to ad copy — the material Google uses to decide what your product is and which searches it belongs in.
What actually needs to be right
To be clear about what this does: better product information helps Google understand and represent your products. It does not by itself improve ROAS — performance still depends on demand, competition, pricing, the product page, conversion rate and margin. Feed work removes a constraint; it does not create demand.
Shopping and Merchant Center implementation in depth: Google Shopping Ads and Merchant Center →
Anatomy of a product record
ROAS measures revenue against ad spend. It knows nothing about what your goods cost, what was discounted, what shipping consumed or what came back as a return. A campaign can hit a strong ROAS target and still leave very little contribution behind.
Illustrative example only — not a client result
Figures chosen to illustrate the arithmetic. Your own margin structure is the only one that determines your targets.
Reported ROAS here is 5.0× — a figure most stores would be pleased with. Contribution after advertising is ₹15,000 on ₹100,000 of revenue, or 15%. Both statements describe the same month.
The target should come from break-even ROAS: one divided by gross margin. At the 35% margin above, break-even sits near 2.9×. So 5.0× is genuinely profitable — and a 2.5× campaign in the same store loses money on every order while still looking like a positive return.
This is also why benchmark ROAS figures are close to useless. A store at 70% margin breaks even below 1.5×; one at 25% needs 4×. Two businesses reporting identical ROAS can be in opposite commercial positions, which is why margin figures matter more than a current ROAS number in isolation.
One caveat worth stating plainly: Google Ads can only optimise towards values it actually receives. Bidding to profit rather than revenue requires passing margin-adjusted conversion values, and that is a measurement project before it is a bidding one.
Campaign-level reporting treats a catalogue as one product. Look at economics and conversion performance per product and four situations appear, each calling for a different decision rather than a different bid.
High margin · converts well
The straightforward case. Push budget here while demand and stock hold, and check that scaling has not quietly moved the acquisition cost past what the margin supports.
High margin · converts poorly
The most valuable quadrant, because the economics already work. Look at traffic quality, price against competitors, the product page and friction in the path — the fix is usually not in the account.
Low margin · converts well
These products flatter every report they appear in, and automated bidding will favour them. The question is whether the revenue they generate leaves enough behind to be worth the budget it absorbed.
Low margin · converts poorly
Rarely worth continued paid acquisition in its current form. Sometimes the answer is exclusion; sometimes it is repricing, bundling, or leaving the product to organic and email.
These are diagnostic starting points, not automated rules. A low-margin product can be worth advertising if it reliably brings customers who return — but that has to be demonstrated from order history, not assumed because it is convenient.
What happens after an account is taken on, in order. The first two steps repeat monthly in miniature — feeds and tracking both drift.
Account history and structure, Merchant Center, feed, tracking, product economics and the landing pages campaigns actually send people to. The output is a written view of what is limiting performance, ordered by likely impact.
Purchase events, transaction values and deduplication validated, then reconciled against order records for a fixed period. Nothing automated is trusted until the numbers agree, because Smart Bidding inherits whatever error is present. Detail: ecommerce conversion measurement.
Which products, campaigns and constraints deserve attention first, judged on what is most likely to move contribution rather than on what is quickest to change. Usually three or four things, not thirty.
Search-term and negative-keyword review, product performance and budget allocation, feed issues, asset testing where the campaign type uses assets, and observations on landing pages passed back to your team or developer.
Monthly reporting against business economics — spend, conversion value, ROAS, CPA, average order value, product mix and contribution — rather than a screenshot of one platform metric that happened to improve.
More investment where demand, conversion performance and margin all support it. Efficiency normally declines as spend rises, so the useful question is where the return crosses break-even, not whether ROAS held.
Not sure whether the account is the problem?That is what the audit answers — and if the constraint turns out to be elsewhere, we would rather say so than sell a retainer.
Request an auditCampaign strategy does not change by platform. What changes is how the catalogue reaches Merchant Center and how reliably purchases are measured, which decides whether the strategy can be executed at all.
The Google and YouTube channel connects the catalogue to Merchant Center and handles much of the conversion setup, which makes it quick to establish and easy to leave partially configured. Worth verifying: whether variants and availability came across as expected, whether purchase values include or exclude shipping and tax, and whether a second tag from a theme or app is duplicating conversions.
More configuration and more control. Feed generation and conversion tracking usually run through plugins, which is fine until several of them are doing overlapping work — the recurring problems are inconsistent product data between feed and site, and duplicated or missing purchase events on custom checkouts.
Nothing is provided, so nothing is assumed. Feed generation, Tag Manager, GA4 and purchase tracking all have to be built, which means they can be built correctly and against your actual margin data from the outset. Recommendations are delivered in a form a developer can act on directly.
The same product data serves more than one channel — worth knowing before paying to build it twice. It also underpins Meta Ads for ecommerce, and the product and category pages it points at are the same ones ecommerce SEO works on.
Paid traffic tends to expose weaknesses rather than compensate for them. In these situations, the recommendation is to fix something else first, sometimes instead of taking the account on at all.
Demand for the product is weak. Search advertising captures existing demand well and creates it poorly. If nobody is looking for the category, Google Ads is the wrong first channel.
Margins cannot support the acquisition cost. If break-even ROAS is higher than the category realistically achieves, the answer is pricing, bundling or cost of goods — not bidding.
The store has a serious conversion problem. Paid traffic arriving at a page that loses people converts a website problem into a monthly invoice.
Pricing is not competitive. Shopping results show your price next to everyone else's. That comparison happens before the click, and no campaign setting changes it.
Inventory is unreliable. Advertising products that cannot be shipped produces cancellations, refunds and distorted conversion data — three costs for one problem.
Merchant Center or product data has unresolved issues. Disapproved items cannot be advertised, so budget concentrates on whatever remains eligible rather than on what was chosen.
Conversion tracking cannot be trusted. This one is fixable quickly and must be fixed first, because every decision after it inherits the error.
There is not enough demand or budget to generate useful data. Automated bidding needs conversion volume to learn from. Below a certain level it cannot, and results stay unstable regardless of management quality.
If any of this describes your situation, the useful next step is diagnosis rather than a campaign build.
This is about whether paid acquisition is economically sensible for you right now, not about excluding anyone. Most stores get there; the question is whether they are there yet.
Clear product-market demand, a functional store that converts, product economics with room for an acquisition cost, fulfilment that can be relied on, an accurate catalogue, measurable purchases, and enough budget to produce data the bidding system can learn from.
This typically describes D2C brands and online retailers on Shopify or WooCommerce, in fashion, beauty and personal care, jewellery, home and lifestyle and consumer products — plus manufacturers selling direct, who often have the best margins in the room and the least developed advertising.
No validated demand for the product. Margins too thin to carry any realistic acquisition cost. A broken or high-friction checkout. Inventory that cannot be relied on. Tracking nobody trusts.
And one that is a mismatch rather than a stage: an expectation of guaranteed ROAS. Nobody can promise a return, because it depends on demand, competition, pricing, the store and margins as much as on the account. Anyone promising a number is either guessing or planning to explain later.
Five areas of ecommerce Google Ads work, each with its own dedicated page. Start with strategy if the account has not been reviewed yet.
How to set campaign priorities, budget allocation and channel mix from your product economics rather than from a generic ecommerce template.
Keyword strategy, match types and negative-keyword discipline for the category, comparison and brand searches a product feed expresses poorly.
How Performance Max is set up, fed and constrained specifically for product businesses, including asset groups and audience signals.
Purchase events, transaction values, deduplication, GA4 and Tag Manager configuration for stores, reconciled against real order records.
How credit for a sale is assigned across campaigns and touchpoints, and what that means for judging Shopping, Performance Max, Search and remarketing fairly.
Running and improving the Google Ads campaigns that sell a store's products: managing product data in Merchant Center, structuring campaigns, validating that purchases and their values are tracked correctly, analysing performance at product level as well as campaign level, and judging results against margin rather than revenue. The campaign work is the visible part; product data and measurement decide whether it can work.
There is no universal mix. Shopping and other Merchant Center-linked campaigns capture high-intent product searches using the feed. Performance Max extends reach across relevant inventory with more automation and less control. Search captures category, comparison and brand queries that product data expresses poorly. Remarketing re-engages people who did not buy. The right combination depends on conversion volume, catalogue size, brand awareness and existing search demand.
Central, for product campaigns. Merchant Center holds the data Shopping and Performance Max use to match products to searches, so incomplete or inaccurate data limits both what can be advertised and how well. Disapproved products cannot be advertised at all, which is why unresolved product issues often explain otherwise inexplicable performance.
Profit, using ROAS as the working indicator. ROAS measures revenue against ad spend and excludes cost of goods, shipping, discounts, returns and payment fees. The practical approach is to calculate break-even ROAS from gross margin — one divided by the margin — treat that as the floor, and judge campaigns against it. A 3× ROAS is excellent in one store and loss-making in another.
Yes, and Shopify is one of the more common platforms in this work. The Google and YouTube channel makes catalogue sync and conversion setup relatively quick, but implementation still needs checking: whether variants and availability transferred correctly, whether purchase values handle shipping and tax as intended, and whether a theme or app is duplicating conversion events.
Yes. WooCommerce setups vary more than Shopify ones, so the starting point is establishing how the feed is generated and how purchases are tracked. Common issues are product data disagreeing between feed and live site, and duplicated or missing purchase events where several plugins or a custom checkout are involved.
No percentage applies generally, and any rule of thumb is worth treating with suspicion. Budget follows from margin and average order value, which set the affordable acquisition cost, plus conversion rate, category competition and the volume automated bidding needs to learn. A store selling ₹800 items needs a very different budget from one selling ₹18,000 items.
Feed and tracking problems can often be resolved within the first few weeks, and where distorted data was the main issue, reported performance can change quickly. Meaningful movement in profitability usually takes two to three months, since campaigns need conversion data after structural changes and product-level patterns take time to become reliable. Neither a fixed ROAS nor a fixed date is promised.
We review the account, the product data, the conversion tracking and the business economics before recommending what should change. If Google Ads is not the main constraint, we say so — it is a more useful answer than a campaign rebuild that cannot work.
Share your website, the work already underway and the problem you want help with. We can discuss whether management, an audit or consulting fits the situation.