Google Ads optimisation for Shopping and PMax

Google Ads optimisation for Shopping and Performance Max is the work of making the account’s inputs trustworthy before you change its settings: what counts as a purchase, what each purchase is worth and which products the spend reaches.

I work in this order: tracking, objective, structure, bidding, tests, products. Changing a target on an account with duplicate purchases or revenue-only values only automates the wrong goal, faster.

1. Why does Google Ads optimisation start with conversion tracking?

Google Ads optimisation starts with tracking because a bidding strategy can’t tell a duplicate purchase from a real one.

It optimises a wrong count as diligently as a correct one, so audit what counts as a purchase before you judge how efficiently the account buys them.

Reconcile Google’s numbers with backend orders

Compare Google-reported conversions and value with backend orders by day, market, device and product ID.

Reports rarely match exactly, so ask for an explanation of any material gap before you touch a bid.

Google’s guidance says to give each conversion its own value, then apply value-based bidding (Google’s optimisation tips for Performance Max with a Merchant Center feed).

It also recommends enhanced conversions; our guide to enhanced conversions and server-side tracking covers setup and checks.

If purchases vanish in browsers that block tracking, run the ad blocker conversion tracking check and decode a request with the SDK debugger.

Keep four numbers apart

Observed and modelled conversions are not independent datasets, so never add them together.

Orders and returns for the latest days are still arriving, so read the most recent week with that in mind.

2. How do you set an objective that includes margin?

Give automation contribution, not revenue: a sale can look strong in Google Ads and leave little after product cost, shipping, fees and returns.

Break-even ROAS is 1 ÷ contribution margin, so every margin has its own floor. The break-even ROAS calculator works it out per product.

Show the dataHide the data

Break-even ROAS by contribution margin

Contribution margin	Break-even ROAS
20%	5.00
25%	4.00
30%	3.33
40%	2.50
50%	2.00

Work through one order

Take one order of €100.00 excluding VAT as an example, with these illustrative costs.

POAS is ROAS × contribution margin, and break-even POAS is 1.0. The POAS vs ROAS post compares the two.

Pass margin to Google

If you don’t send profit as the conversion value, conversion value rules are the nearest lever: they raise or lower conversion values by customer type, device and location, as Google’s Performance Max feed guidance describes them.

Google’s own example is different margins for different types of customers. The rules don’t set a value per product, so product margins still need profit values or separate targets.

Set the target from margin, not from a revenue average; setting target ROAS from margin shows how.

3. How should you structure campaigns and choose a bidding strategy?

Consolidate campaigns that share an objective, split only for a real difference in budget, margin target, geography or product priority, then pick the bidding strategy for the constraint you have.

Structure around decisions

A campaign for every category looks organised but gives little control. Mixing products with incompatible economics makes one target hard to defend.

For feed and structure in depth, read how to optimise Google Shopping ads. To choose a structure by profit, see product segmentation for Google Shopping and PMax.

Keep a decision log of why each campaign exists and which decision its structure enables.

Then check demand matching.

The Google Shopping feed optimisation guide covers the feed side.

Match bidding to the constraint

Use Maximise conversion value to pursue purchase value within the budget. Use target ROAS when return is an explicit constraint and conversion values are dependable.

Keep manual control for teams that accept the management work.

Performance Max optimises for marginal return, the incremental value of additional conversions, so the average ROI or CPA of one channel can mislead (Google’s answers to top Performance Max questions).

4. How do you test one change at a time?

Run one variable at a time with a written acceptance rule.

A change log is not an experiment: if you change bidding, assets, product coverage and budget together, you can’t tell which change helped.

Google’s campaign experiments FAQ sets the ground rules.

Read the result without fooling yourself

Recent clicks haven’t produced their orders yet, so early movement can mislead.

Avoid overlapping experiments on the same traffic, and judge the result against contribution margin, not only the platform-reported outcome.

Urgent fixes, such as a broken destination, shouldn’t wait for a test. Log them so nobody mistakes them for evidence about another change.

To ask whether attributed sales would have happened anyway, read our guide to incrementality testing.

5. Which products deserve the next euro?

The products with the best return against their own break-even deserve the next euro, and a campaign average can’t show you which those are.

A campaign can meet its target while holding products you wouldn’t knowingly fund.

Show the dataHide the data

Four products, one shared target ROAS

Product	ROAS	Break-even ROAS
Product A (40% margin)	3.1	2.5
Product B (20% margin)	4.8	5.0
Product C (50% margin)	2.6	2.0
Product D (25% margin)	4.4	4.0

Judge each product against its own break-even

In the illustrative data, products A and C sit below the shared target, yet their POAS is 1.24 and 1.30, so both are profitable.

Product B clears the target at a ROAS of 4.8, but its POAS is 0.96, so it loses money on every sale.

The clicked product and the purchased product can differ.

An advertised product may introduce a customer to another item, so judging it only on its own sales can understate its role.

The opposite mistake is to treat every attributed basket as proof of demand: attribution records a path, not what would have happened without ads.

Give each product situation an action

Pricing or stock may be the right fix rather than a bid change, so involve merchandising.

Our Performance Max for ecommerce guide shows how to exclude or split by label.

Product Segmentation applies these situations to every product at once, so you don’t have to work through them by hand.

Account for reporting changes

Google has widened what product reporting covers.

Log the date when you compare periods across the June 2026 change. Better reporting helps diagnosis, but it doesn’t replace product economics.

6. What should you fix first?

Fix the action that removes the biggest uncertainty at the least risk, and diagnose before you expand.

None of these actions promises a lift.

When evidence is incomplete, don’t exclude on ROAS alone, don’t react to immature windows, and don’t treat channel averages as allocation instructions. Make the decision smaller instead.

Final thoughts

Optimisation is mostly a question of what the account is allowed to learn from. Fix the inputs and every later decision gets easier to read.

This week, pick one product group that attracts spend and write down the question you can’t yet answer.

Is a heavily clicked item overpriced, or does a low-return product start valuable cross-sales?

Give tracking questions to a measurement owner, pricing and stock questions to merchandising, and record the baseline before you change anything.

Further reading

If you are newer to Google Ads, The Advertising Suite’s guide to how to use Google Ads effectively covers the account basics behind this order of work.

Agencies running this review across many client accounts can start from AdCrunch’s resource on scaling optimisation across client accounts, which turns account work into a repeatable process.

To hand the tracking audit to an agency, find a D2C paid search agency through Next Point Digital.

Make transaction-level validation part of the scope: ask for an explained reconciliation, not just confirmation that a tag fires.