How to optimise PPC campaigns: a margin-first playbook

PPC optimisation is the work of changing tracking, targets, keywords, ads and product priority so each euro of ad spend returns more profit, not only more revenue.

For an online shop, I do it in a fixed order: tracking, margin targets, search terms and ads, products, then one test.

A click’s price says little on its own, so the order starts with what an order is worth.

The benchmarks come from WordStream’s public 2024 report and the targets from 1 ÷ margin arithmetic, so you can check every number.

Where do PPC campaigns lose money?

PPC campaigns lose money in four places: the auction, traffic quality, the site and product economics. A high cost per click (CPC) is only a problem when the other three can’t pay for it.

In 2024, the average Google Ads CPC was $4.66 across industries, based on more than 17,000 campaigns from April 2023 to March 2024 analysed by WordStream. Its retail rows show why a cheap click is not a cheap sale:

Five of the six retail categories pay a below-average CPC, yet two of them pay an above-average cost per conversion.

Furniture is the extreme: clicks cost 29% less than average and each conversion 79% more, because only 2.53% of clicks converted.

Show the dataHide the data

Retail CPC and cost per conversion as a share of the all-industry average

Category	CPC, % of average	Cost per conversion, % of average
Shopping, Collectibles & Gifts	56%	63%
Sports & Recreation	50%	75%
Beauty & Personal Care	76%	73%
Apparel / Fashion & Jewelry	73%	125%
Furniture	71%	179%
Home & Home Improvement	149%	123%

Treat these rows as context, not a target. Your own limit is the break-even CPC: contribution per order × conversion rate, the most a click can cost before the order loses money.

At €25 contribution per order, each extra percentage point of conversion rate lets a click cost €0.25 more.

Show the dataHide the data

Break-even CPC by conversion rate, at €25 contribution per order

Conversion rate	Break-even CPC
1%	€0.25
2%	€0.50
3%	€0.75
4%	€1.00
6%	€1.50
8%	€2.00

Which signals point to the cause?

The three Quality Score components and the pattern of CPC, click-through rate and conversion rate point to the cause. Quality Score is Google’s 1–10 keyword-level estimate of expected click-through rate, ad relevance and landing page experience (Google’s Quality Score documentation).

Google says it is not used at auction time to determine Ad Rank, so read it as a diagnostic, never as a KPI.

Treat each row as a hypothesis. A sold-out size or a surprise delivery charge explains a low conversion rate more often than the landing page design does.

Judge every fix on cost per conversion, revenue and contribution profit.

Campaign averages also hide products: two products with the same ROAS can earn opposite profit, which is why ROAS can hide profit differences.

Why fix conversion tracking before changing bids?

Fix conversion tracking first because Smart Bidding optimises towards whatever you count as a conversion. Each common fault teaches it the wrong lesson, and it then buys more traffic like it:

Ad platform and shop reports rarely match exactly. Allow for reporting delay and attribution, but explain any material gap before you touch a target.

The full checklist is in Google Ads optimisation for Shopping and Performance Max.

Our guide to Google Ads conversion tracking and ad blockers separates enhanced conversions from server-side delivery. Neither replaces consent checks, correct order values or deduplication.

If purchases go missing in browsers with blockers, run the ad blocker conversion tracking check.

How much data does attribution need?

For data-driven attribution, Google recommends at least 200 conversions and 2,000 ad interactions within 30 days in supported networks. Below that, small differences between attribution models are noise, not instructions for moving budget.

A campaign can look weak on last click because it introduces shoppers who come back through another channel. Another looks strong because it takes the final click after other marketing did the work.

Google Ads attribution reports show those conversion paths, such as a keyword that assisted a conversion another keyword closed.

Attribution assigns credit, though, not cause: incrementality testing tells you whether the sales would have happened without the ads.

An incomplete profit value steers worse than an accurate revenue value, so pick the conversion value by what your data covers:

How do you set PPC targets from margin?

Set PPC targets from contribution margin, not from the sales price: break-even ROAS is 1 ÷ contribution margin, and the ROAS that keeps a profit is 1 ÷ (contribution margin − the share of revenue you want to keep).

Count product cost, fulfilment, payment fees, discounts and expected returns, with revenue and costs both excl. VAT. Your maximum affordable CPA is the contribution per order minus the profit you want to keep.

For example, a product with a 25% contribution margin breaks even at a ROAS of 4. To keep 10% of revenue after ads it needs 1 ÷ (0.25 − 0.10) = 6.67, a target ROAS of 667%.

Show the dataHide the data

Break-even ROAS and the ROAS to keep 10%, by margin

Contribution margin	Break-even ROAS	ROAS to keep 10% of revenue
20%	5	10
25%	4	6.67
30%	3.33	5
35%	2.86	4
40%	2.5	3.33
50%	2	2.5

The thinner the margin, the faster the required ROAS climbs: from a 40% to a 20% margin, break-even ROAS doubles from 2.5 to 5. That is why one account-wide target funds the low-margin products.

Check your own products in the break-even ROAS calculator, and agree the cost definition first: no calculator corrects a margin that leaves out returns.

How to set a target ROAS from your margin covers margin bands and how to move a target without unsettling the bid strategy.

What does the same ROAS earn on different margins?

The same ROAS earns a loss on a low-margin product and a profit on a high-margin one.

POAS (profit on ad spend) is ROAS × contribution margin, and break-even POAS is 1.

At a ROAS of 4, the 25% product from the target example sits exactly at 1, so any lower margin loses money.

Show the dataHide the data

POAS at a ROAS of 4, by contribution margin

Contribution margin	POAS at a ROAS of 4
20%	0.8
25%	1
30%	1.2
40%	1.6
50%	2

Google Ads reports gross profit for conversions with cart data (revenue minus cost of goods sold), but it has no POAS bid strategy. Maximise conversion value with a target ROAS steers on whatever value you send.

In Performance Max, the optional ROAS or CPA target belongs to the campaign (Google’s overview of Performance Max):

Products that need very different targets need separate campaigns. Each split thins the conversion data per campaign, so split on large margin gaps only.

Stock limits the target too. When replenishment won’t arrive before demand empties the shelf, lower the product’s priority or exclude it rather than selling what you can’t deliver.

How do you match keywords, search terms and ads to intent?

Match intent by reviewing search terms on a fixed routine, excluding only intent you are sure you can’t serve, and writing each ad group’s headlines around the need its queries share.

A query with no recorded purchase may be irrelevant, too recent or too rarely seen to judge.

How do you review search terms without cutting profitable demand?

Review search terms by tagging each one as converting, wasteful or unclear, and act only on the first two. Here is what that looks like for a shop selling waterproof footwear (illustrative queries):

The routine behind the tags:

Keep “best” and “reviews” queries even before they sell: those shoppers are still choosing which product to buy.

A high-spend query isn’t automatically wasteful either: check whether its orders clear the contribution limit and whether the landing page matches it.

Judge the cleanup by a shrinking share of clearly irrelevant spend over completed conversion windows, not by a lower CPC. A narrower traffic mix can cost more per click and still bring better orders.

How do you write ads that match search intent?

Write ads that name the need the ad group’s queries share, then give distinct, true reasons to choose you.

Say a shop sees strong demand for waterproof trail running shoes for women, but its ads lead with brand slogans:

Each illustrative headline fits the 30-character limit for responsive search ad headlines. The landing page should keep the same promise in its heading, filters, available sizes and delivery information.

Responsive search ads and Performance Max combine your headlines and descriptions, so each asset has to read well next to any other:

Read asset ratings alongside sales: a rating doesn’t prove a line caused profitable orders.

Test one messaging idea with a fixed asset set, not while also changing the landing page and the offer.

How do you prioritise products in Shopping and Performance Max?

Prioritise products by two numbers, ad clicks and return against the product’s own ROAS or POAS target, then write the result to a Merchant Center custom label (custom_label_0 to custom_label_4) your campaigns can split on.

Classify each product against its own economics, not one revenue threshold for every product, as in judging ecommerce PPC by product profit. Then check stock and merchandising before you settle the action.

A Star deserves room to grow, not an unlimited budget. Lowering its return target buys reach at lower efficiency, so watch total contribution after ads rather than volume.

Before you give a Ghost more budget, check its title and whether anyone searches for the product at all.

The labels are only as good as the margin, return and stock data behind them.

A stale stock feed can keep a sold-out product in a scale segment while its ad data still looks excellent. Give the label definitions an owner and a rule for missing inputs.

Doing this by hand means a spreadsheet per product, rebuilt every week from Google Ads, Merchant Center, your margins and your stock, which stops scaling at a few hundred products.

How do Product Metrics labels work?

Product Metrics labels work in two steps: Product Segmentation segments every product into the six segments above with Product Metrics ML, then product rules write labels to your Merchant Center supplemental feed.

Segmentation runs on ROAS straight away, or on POAS once you connect margins. You split campaigns or asset groups on those labels and set targets and budgets yourself.

For example, the Promote Star candidates rule sets custom_label_0 to star_candidate when all three example conditions hold:

A label doesn’t set a bid by itself. Campaign structure by profit covers where a separate campaign is worth the thinner data.

How do you test a change before scaling it?

Test one change in a Google Ads experiment with a written hypothesis and decision rule, then roll it out only to the products it was proven on.

A rising trend alone can be seasonality, conversion delay or the bid strategy still learning.

Google’s experiment guidance asks for a hypothesis, one variable at a time, one or two success metrics picked before the test begins, no edits to the base campaign unless sync is on, and a record of every test.

Write the protocol down before you touch a setting:

Don’t stop early because one arm leads, and don’t edit the tested assets mid-test. Read completed cohorts only, and check that a significant result is also worth enough money to implement.

Once a change wins, apply it only to the product group it was tested on, such as Stars or stock-limited items.

Final thoughts

A cheap click and a healthy campaign ROAS are both averages, and averages are where PPC accounts hide their losses.

The fix is to judge each click, query and product against what its order leaves after costs.

Start this week with one campaign: check its purchase signal, set its target from margin, and change one thing, with a named owner and a written decision rule.

If you want a broader account review before you start, Amax Marketing’s checklist shows how to improve Google Ads with Amax Marketing by checking account weaknesses before making isolated bid changes.

When your ad reports and your order economics disagree, AdCrafty’s margin-analysis guide can help you identify hidden costs behind healthy ROAS.

If you draft headline variations with AI, CreativeClaw’s guide to create ad variations with AI covers the workflow. Give the tool verified product facts and banned claims, then edit every line before it goes live.