PPC for ecommerce is unusual among marketing channels in that you can be very good at it and still lose money. A campaign can hit a 4x return on ad spend, win a stack of impression share, and quietly destroy margin because the products it sells best are the ones with the thinnest markup. Most guides skip that part and go straight to bid strategies. This one starts with the arithmetic, because the arithmetic decides everything downstream — what you bid, what you exclude, and whether a "good" account is actually a good business.
What follows is the sequence we'd work through on a store spending anywhere from $5,000 to $250,000 a month: get the numbers right, get the data plumbing right, then get the campaign structure right. In that order.
Start with break-even ROAS, not target ROAS
Every ecommerce PPC account needs one number before it needs anything else: the return on ad spend at which a sale stops costing you money. It's a one-line calculation.
Break-even ROAS = 1 ÷ gross margin.
A store running a 40% gross margin breaks even at 2.5x. At 25% margin, 4x. At 60%, 1.67x. Run the numbers on your own catalogue with the ROAS calculator, and work out the underlying margin with the profit margin calculator if your figures live in a spreadsheet somewhere and nobody's checked them since the last supplier price rise.
Three things break this simple version, and all three matter:
Gross margin is not one number. It varies by product, often wildly. If your 18%-margin electronics accessories account for 60% of ad-driven revenue while your 55%-margin own-brand line accounts for 12%, a blended 4x target is far too loose on one and far too tight on the other. Segment the target by margin band, not by product category.
Returns eat the margin after the fact. Apparel stores with 30% return rates are not earning the ROAS their ad platform reports. If a category returns at 30%, its effective revenue is 70% of reported revenue and its break-even ROAS rises accordingly. Feed returns data back into your targets quarterly at minimum.
Repeat purchase changes the answer entirely. If the average customer buys 2.8 times over 18 months, a first-order ROAS below break-even can still be profitable. This is the argument for bidding aggressively on new customers, but only if you have the data to support it — model it properly with the customer lifetime value calculator rather than assuming a flattering repeat rate. And be honest about payback period: a business with 45 days of cash runway cannot fund an 11-month payback no matter how good the LTV looks. The break-even calculator is useful here for pressure-testing how long you can carry that gap.
Once you have per-segment break-even ROAS and a payback window you can actually finance, you have a target. Everything below is about hitting it.
Fix the product feed before you touch bids
Shopping campaigns are where most ecommerce ad budget goes, and Shopping is a feed-driven channel. Your feed is your keyword list, your ad copy, and your relevance signal all at once. A neglected feed caps performance in ways no bidding strategy can rescue.
The high-leverage fields, in rough order of impact:
Product title. Google matches queries against titles more heavily than any other attribute. The pattern that works is Brand + Product Type + key differentiating attribute + size/colour, front-loaded, within roughly 70 characters before truncation. "Nike Air Zoom Pegasus 41 Men's Running Shoe, Black, US 10" beats "Pegasus 41 — Shop Now" every time, and it isn't close.
Product type and Google product category. Custom product_type is your own taxonomy and it's what you'll use to slice campaigns and bids. Make it deep and consistent — Footwear > Running > Neutral Cushioned rather than Shoes. Google's own category taxonomy affects which surfaces you're eligible for.
GTIN and brand. Missing GTINs suppress eligibility on comparison surfaces and weaken matching. If your supplier data has gaps, chase them; it's dull work with an outsized payoff.
Custom labels. These are your five free slots for anything the bidding needs to see: margin band, stock level, seasonality, bestseller flag, clearance status. Pushing margin band into custom_label_0 is the single change that lets you set different ROAS targets by profitability instead of by guesswork. Do this one first.
Images. Clean, white-background, no promotional overlays (which get disapproved), and the primary image should show the product as the shopper expects to see it. Lifestyle shots belong in additional image slots.
Set a monthly recurring check on feed disapprovals in Merchant Center. Silent disapprovals — a price mismatch here, a missing shipping attribute there — remove products from auctions without ever showing up in your ads reporting as a problem. You just see revenue drift down.
Campaign structure: segment by economics, not by taste
The old advice was to split Shopping campaigns by product category. That's the wrong axis. Categories are a merchandising convention; they tell the bidding algorithm nothing it can act on.
Segment by economics instead:
- Margin band. High-margin products can afford lower ROAS targets and should be bid up. Low-margin products need tight targets or exclusion.
- Intent. Branded queries convert at a fraction of the cost of generic ones. Blending them together inflates your reported performance and hides the real cost of new-customer acquisition. Isolate brand traffic so you can see the difference.
- Product lifecycle. New products with no conversion history need a separate campaign with a manual push, or they'll be starved by algorithms optimising for what already works. Bestsellers, long-tail, and clearance each want different treatment.
For Performance Max specifically: use asset group and listing group structure to enforce these splits, and run separate campaigns where you need genuinely different ROAS targets. One campaign can only chase one target. If you're running a single Performance Max campaign across the whole catalogue with a blended 400% target, you're averaging your way into mediocrity — subsidising the weak products with the strong ones and calling the result a strategy.
Search campaigns alongside Shopping still earn their place for high-intent non-brand queries where you want copy control, and for defending your own brand terms against competitors bidding on them. Keep them tightly themed; the sprawling 200-ad-group accounts of a decade ago don't survive contact with modern match types.
Bidding: give the algorithm a target it can hit
Smart bidding works well when three conditions hold: enough conversion volume, accurate conversion values, and a target that is achievable. Break any one of them and it fails in ways that look like the algorithm's fault.
Volume. Target ROAS needs roughly 30–50 conversions in the past 30 days per campaign to work reliably. Below that, consolidate campaigns or use Maximise Conversion Value without a target until data accumulates.
Accurate values. Send actual transaction value, and if you can, send profit rather than revenue. Passing gross profit as the conversion value — revenue minus COGS, calculated at the line-item level in your purchase event — converts the bidding algorithm from a revenue maximiser into a profit maximiser. This is the highest-leverage change available to most ecommerce accounts and it takes an afternoon of developer time.
Achievable targets. Setting a 900% target on a campaign historically achieving 300% doesn't produce a 900% campaign. It produces a campaign that barely spends. Move targets in 10–15% steps and give each change a fortnight before judging it.
The other half of this is measurement integrity. Enhanced conversions, server-side tagging, and a consent setup that doesn't quietly drop a third of your conversions are prerequisites, not refinements. Under-reported conversions make every campaign look worse than it is, and smart bidding will duly bid down on your best traffic. If your tracking is uncertain, resolve that before changing a single bid.
Remarketing and the new-customer question
Remarketing carries the highest reported ROAS in nearly every ecommerce account, and a portion of it is incremental and a portion is people who'd have come back anyway. The honest way to run it is to treat cart abandoners and recent product viewers as genuinely valuable audiences, and treat sitewide 540-day lists as the low-value inventory they usually are. Our guide to building remarketing lists in Google Ads covers the audience configuration in detail.
The more consequential decision is whether to bid differently for new customers. Google's new-customer acquisition goal lets you either bid higher for new customers or restrict a campaign to them entirely. It's worth using if — and only if — your LTV data supports paying more for a first order, and your cash position can carry the payback. Turning it on because it sounds strategically sophisticated, without the LTV work behind it, is a fast route to a cash crunch.
The reports that catch waste early
Four checks, run on a fixed cadence, catch most of what goes wrong:
Search terms, weekly. Even in an automated account, search terms reveal what you're actually paying for. Negative out the informational queries, the competitor names you don't want, and the wrong-intent traffic. In Performance Max, the search categories report is thinner but still worth reading.
Product-level performance, monthly. Sort by spend descending and look for products with meaningful spend and zero conversions across a 60-day window. Exclude them or fix their listings. This report finds more waste per minute spent than any other.
Margin-weighted ROAS, monthly. Take your revenue by product, apply real margin, and recompute. Accounts frequently discover the campaign with the best reported ROAS is third or fourth once margin is applied.
Landing page conversion rate, ongoing. Traffic quality and page quality are separable problems, and PPC diagnostics often blame the former for the latter. If your product pages convert paid traffic at 1.2% while the category average is 2.8%, the fix is on the page, not in the account. Our guide to ecommerce conversion rate optimization covers what actually moves that number, and the conversion rate calculator is a quick way to compare segments before you commit to a test.
Where PPC fits with everything else
Paid search rarely operates alone. The stores that get the most out of it treat it as one channel in a system:
- Organic search covers the informational and comparison queries that convert too slowly to bid on profitably. If you're on Shopify, the technical groundwork in our Shopify SEO guide compounds while ad costs don't.
- Email captures the traffic PPC pays for but doesn't convert on the first visit. A capture-and-nurture flow measurably improves the effective ROAS of every paid click.
- Account structure hygiene, conversion tracking, and audience setup are shared infrastructure. If you're starting from scratch, our Google Ads account setup guide covers the foundations this article assumes.
The realistic target for a well-run ecommerce PPC programme isn't a headline ROAS number to screenshot. It's a channel that reliably produces profitable orders at a volume your operations can fulfil and your cash flow can fund — with reporting honest enough that you'd act on it.
FAQ
What is a good ROAS for ecommerce PPC? There's no universal number, because a good ROAS is entirely a function of your gross margin. A 2.5x return is excellent for a 60%-margin business and loss-making for a 25%-margin one. Calculate your break-even ROAS as 1 ÷ gross margin, then set a target above it that leaves room for overheads and returns. Compare that to your actuals rather than to an industry benchmark.
Should I run Performance Max or standard Shopping campaigns? Performance Max generally outperforms standard Shopping on reach and efficiency once it has conversion data, but it gives you far less visibility and control. Most accounts above roughly $20,000 a month in spend end up running Performance Max as the core with a small set of standard Search campaigns for brand defence and specific high-intent terms. Below that spend level, the reduced control matters less than the reach advantage.
How much should I budget to start with PPC for ecommerce? Enough to generate 30–50 conversions a month, which is the volume smart bidding needs to function. Work backwards: if your average order value is $80 and your realistic cost per acquisition is $25, that's $750–$1,250 a month as a floor. Starting materially below that means the algorithm never gets enough signal, and you'll conclude the channel doesn't work when the real problem was insufficient data.
Why is my ROAS high but my profit flat? Usually one of three causes. Your reported revenue includes returns you're later refunding. Your ad spend is concentrated in your lowest-margin products, so revenue looks good and contribution doesn't. Or a large share of your conversions are branded searches from customers who'd have bought anyway, meaning you're paying for orders you already had. Recompute ROAS with real margin applied and with brand traffic separated out — the answer is almost always visible in that one view.