Home › Resources › Checklists › eCommerce PPC
eCommerce PPC Checklist
Most eCommerce PPC audits start with campaign settings and end with a list of things to pause. This one starts with whether the numbers the account reports are true, because every optimisation after that inherits the answer. Forty-two checks, in the order they are worth running.
CHECKS
42 checks
TIME TO RUN
One day
LAST UPDATED
September 2026
LEVEL
Marketer or in-house PPC
An account optimised against the wrong conversion number gets efficiently worse.
- Before you start
What this covers, and what it deliberately does not
Scoped to Google Ads and the measurement underneath it. The channels it leaves out are covered properly elsewhere.
This is a Google Ads account audit: Search, Shopping, Performance Max, and the conversion tracking and profitability maths that decide whether any of it is working. That is where the majority of eCommerce paid budget sits and where the majority of avoidable waste is.
Amazon advertising is a different system with different economics and is covered by the Amazon listing checklist and the break-even ACoS calculator. Paid social is not audited here, but the profitability section applies to it unchanged.
- Google Ads at admin level, with change history access
- Google Analytics 4 and Merchant Center for the same store
- Your actual gross margin, or at least a defensible estimate
- Ninety days of data, or the account is too new to audit
Run the sections in order. Everything from section two onward assumes section one passed.
SECTION 01
7 checks
Measurement you can trust
This section comes first because it invalidates the rest. An account bidding toward a conversion value that is wrong will reliably optimise itself toward the wrong customers, and it will do so confidently.
Confirm there is exactly one primary purchase conversion action
Multiple purchase actions marked primary double-count revenue and inflate every ROAS figure in the account. It is the single most common measurement fault we find.
Check the conversion value passed is revenue, not basket value before discount
If the value sent ignores discount codes, the account believes promotional traffic is more valuable than it is and bids accordingly.
Verify the conversion count matches the orders in your platform
Pull one week from Google Ads and the same week from your store admin. A gap larger than a few per cent needs explaining before anything else on this list matters.
Check which secondary actions are included in Conversions
Newsletter signups and add-to-basket counted as conversions will pull an automated bidding strategy away from revenue toward cheap events.
Review the attribution model and know which one is in force
It changes every comparison you are about to make. You do not need a particular model; you need to know which one produced the numbers on screen.
Confirm consent mode and tag behaviour in your largest markets
Consent handling decides how much of your conversion data is measured rather than modelled. Modelled data is legitimate, but you should know what proportion it is.
Check for cross-domain or subdomain tracking breaks at checkout
If the checkout sits on a different domain and tracking is not linked, the account loses the conversion and the session that produced it.
SECTION 02
7 checks
Account structure and hygiene
Structure exists to make the account controllable and readable, not to satisfy a rule about how many keywords belong in an ad group. Most inherited accounts are structured around a theory nobody remembers.
Read the change history for the last ninety days
Before judging performance, find out what was changed and when. Most sudden shifts have a cause sitting in the log rather than in the market.
Check campaigns are separated by anything you would budget separately
Brand and non-brand at minimum, then whatever you would genuinely move money between: margin tiers, categories, territories. If you would never budget them apart, they do not need separating.
Confirm brand traffic is isolated and measured on its own
Brand campaigns flatter every blended figure they are mixed into. They are often worth running; they are never worth hiding inside a non-brand average.
Check geographic targeting is set to presence, not presence or interest
The default setting serves ads to people merely interested in your location. On a store that cannot ship there, it is pure waste.
Review ad schedules and device modifiers for rules nobody remembers setting
Inherited accounts collect these. Each one is a decision made against data that may be years old.
Confirm the account, not the agency, owns the data and the assets
Conversion actions, audiences and historical data should live in an account you own. If they do not, that is a commercial problem, not a technical one.
Look for campaigns with no spend, no impressions and no end date
Dormant campaigns hide budget, confuse reporting and occasionally reactivate themselves when a shared budget frees up.
SECTION 03
7 checks
Search campaigns and search terms
The search terms report is the only place in the account where you can read what people actually typed. It is also the report most often skipped, because it is long and the conclusions are uncomfortable.
Read ninety days of search terms sorted by cost, not by conversions
Sorting by conversions shows what worked. Sorting by cost shows what you paid for. The second list is where the money is.
Check the proportion of spend on terms containing your brand
If most of the non-brand budget is quietly buying brand variants, the account is cheaper than it looks and smaller than it claims.
Look for research-intent terms being paid for at purchase-intent prices
How to, what is, and versus queries convert differently. They are not worthless, but they should not be funded from the same bid as a product query.
Confirm a negative keyword list exists and is maintained at account level
A negative list that has not been updated in six months is a list of last years waste.
Check match types against what broad match is actually matching
Broad match with a strong signal works. Broad match without conversion data or negatives is an unrestricted budget.
Confirm every ad group has at least three assets and a pinned headline where it matters
Responsive search ads with two headlines are not being tested. Pinning matters for price claims, delivery promises and anything legally constrained.
Check the landing page for each ad group is the page that answers the query
Category queries should not land on the homepage, and product queries should not land on a category. This one check often explains a poor conversion rate entirely.
SECTION 04
7 checks
Shopping, the feed and Merchant Center
Shopping is the one channel where the ad copy is your catalogue data. An account with a neglected feed cannot be optimised through bidding, because the bidding is working with the wrong material.
In Shopping, the product feed is the campaign. Everything else is a budget instruction.
Check the Merchant Center diagnostics for disapprovals and warnings
Disapproved products cannot serve at any bid. Warnings are the disapprovals of next quarter.
Compare the number of products submitted against the number actually serving
The gap is usually missing attributes, price mismatches or out-of-stock items, and it is invisible from inside Google Ads.
Confirm feed titles lead with the terms people search, not internal naming
The title is the single strongest relevance signal in Shopping. Brand plus product type plus the defining attribute, in that order, beats a catalogue code every time.
Check GTIN, brand, condition and product type are populated
Missing identifiers restrict eligibility and comparison. Product type is your own taxonomy and is the attribute most often left empty.
Verify feed price and availability match the live product page
Mismatches cause disapproval, and a shopper who lands on a different price than advertised is lost and annoyed.
Confirm the feed refreshes often enough for your rate of change
A daily feed on a store that reprices hourly will spend money on prices that no longer exist.
Segment Shopping by product rather than bidding the whole catalogue equally
A flat structure spends the same on your best margin item and your worst. The feed optimisation guide covers how to split it.
SECTION 05
7 checks
Performance Max and automation
Automated campaign types are not the problem people say they are. The problem is handing them a bad signal, no exclusions and no way to see what they did, and then judging them on a blended number.
Check what Performance Max is actually being credited with
Run the asset group and listing group reports rather than the campaign total. A campaign that mostly serves against brand queries is reporting your existing demand back to you.
Confirm brand terms are excluded where you want them measured separately
Brand exclusions are available at campaign level. Without them, no meaningful comparison between Performance Max and anything else is possible.
Review the search themes and audience signals you are supplying
These are instructions, not targeting. A weak or generic signal produces a campaign that finds the cheapest conversions rather than the best ones.
Check the asset groups have real creative, not one image and a logo
Thin asset groups restrict which inventory the campaign can serve on, which usually means it collapses back to Shopping inventory.
Confirm account-level and campaign-level exclusions for placements and content
Without them, video and display inventory absorbs budget on placements nobody would buy deliberately.
Check whether Performance Max is competing with your own Search campaigns
Where both are eligible, the one with the stronger signal wins the auction. That is fine if it is a decision and expensive if it is an accident.
Set a review period before judging any automated campaign
Automated bidding needs a learning period, and changing the target mid-learning restarts it. Decide the window in advance and leave it alone.
SECTION 06
7 checks
Budget, bidding and profitability
ROAS is a ratio of revenue to spend, and revenue is not margin. This section is about whether the account is targeting a number that leaves any money behind, and whether the media plan has anywhere to grow.
Calculate your break-even ROAS from gross margin, not from revenue
Break-even ROAS is one divided by gross margin. Until you know it, every target in the account is a guess dressed as a decision. The break-even ROAS calculator does the arithmetic.
Check the target ROAS in each campaign against that break-even figure
A target below break-even is a decision to buy revenue at a loss. Sometimes that is deliberate; it should never be accidental.
Include returns in the margin you are bidding against
A category returning a quarter of its units has a materially different break-even from one that returns nothing. Blended margin hides this completely.
Compare platform-reported revenue with revenue in your own accounts
Every ad platform claims conversions that other platforms also claim. Your store admin is the only figure that reconciles.
Check whether campaigns are limited by budget or by target
A campaign limited by budget has demand you are not buying. A campaign limited by target is bidding itself out of the auction. The fix is different in each case.
Look at how much spend sits on products that cannot be profitable
Low-margin and heavily-returned lines will absorb whatever budget the structure lets them. Excluding them is usually worth more than any bid adjustment.
Decide the one number the account is managed against, and publish it
Contribution after media is the honest candidate. Whatever it is, it should be agreed with whoever owns the P and L rather than chosen inside the ad account.
- Getting it wrong
Six ways PPC audits go wrong
Not mistakes in the account — mistakes in the auditing, which are more expensive because they produce confident conclusions.
Auditing performance before auditing measurement
Every judgement made about a campaign inherits the conversion data underneath it. Check the data first or spend a day being precisely wrong.
Reading blended ROAS with brand inside it
Brand search converts at a rate no acquisition campaign can match. Mixed into an average, it makes a failing account look adequate and hides which half is which.
Pausing everything that looks inefficient
Upper-funnel and category terms rarely convert last-click. Cutting them improves the report for a month and reduces the demand feeding everything else.
Changing targets during the learning period
Each change restarts learning. An account edited weekly never leaves it, and its performance data never becomes trustworthy.
Treating ROAS as if it were profit
A 4x ROAS on a 20 per cent margin loses money. Break-even ROAS is one divided by gross margin, and it is the only target that means anything.
Auditing the ad account and ignoring the feed
In Shopping the feed is the ad. An account audit that never opens Merchant Center has skipped most of the campaign.
If several of these apply, redo the measurement section before acting on anything else you found.
- Sequencing
What to fix first
Ordered by how much each one changes the value of the work that follows it.
Order
What to do
Why here
Typical effort
First
Fix conversion tracking and reconcile it with your store
Nothing below this line is trustworthy until it is done, and it is usually a day of work. Every other item on the list depends on it.
A day
Second
Separate brand from non-brand and establish break-even ROAS
Together these decide what good looks like. Without them you cannot tell an efficient account from a flattered one.
One to two days
Third
Clear the feed: disapprovals, titles, identifiers, availability
Highest-return work in most eCommerce accounts, because it raises eligibility and relevance at once rather than shifting budget around.
Three to ten days
Fourth
Work the search terms report and build the negative list
Direct waste reduction, and it compounds. Slower than it looks because it deserves to be read rather than skimmed.
Two to three days
Last
Restructure campaigns and retune bidding targets
Restructuring resets learning, so it should happen once, after the measurement and the feed are sound, not as a first instinct.
One to two weeks
Restructuring first is the most common sequencing error. It is visible, it feels like progress, and it costs you the learning period twice.
- What next
Where to go from here
Four routes, depending on what the checklist turned up.
01
TARGETS LOOK WRONG
Enter your margin and it returns the ROAS below which every additional order costs you money. Prints the formula so you can check it.
02
FEED IS THE PROBLEM
Titles, attributes, structure and the segmentation that lets you bid differently on products that earn differently.
03
TRAFFIC IS FINE, SALES ARE NOT
Paid traffic arriving on a product page that does not answer objections is an expensive way to fund a bounce.
04
WANT IT DONE
The same audit run by the specialists who manage these accounts daily, with the account, the data and the assets staying in your name.
- Questions
eCommerce PPC FAQs
What marketers ask while working through this.
How often should I audit a PPC account?
A full audit twice a year, and a short measurement check every quarter. Accounts drift through accumulated small decisions rather than sudden failures, and the drift is only visible against a baseline.
What is a good ROAS for eCommerce?
There is no general answer, and anyone offering one is guessing at your margin. Break-even ROAS is one divided by your gross margin: at 25 per cent margin it is 4x, at 50 per cent it is 2x. Good means comfortably above your own break-even after returns.
Should I run Performance Max?
Usually yes, with brand excluded so you can measure it, real assets in every asset group, and a defined review window. The complaints about it are mostly complaints about running it without those three things.
Why does Google Ads report more revenue than my store?
Attribution windows and cross-platform double-counting. Every platform claims conversions other platforms also claim. Your store admin is the only figure that reconciles, and it is the one to manage against.
Do I need to separate brand campaigns?
Yes, if you want to know anything about acquisition. Brand converts at rates non-brand cannot approach, so any average containing both describes neither.
How long should I leave a change before judging it?
Long enough for the bidding strategy to exit learning and accumulate conversions, which on most eCommerce accounts means two to four weeks. Decide the window before you make the change, not after you see the first week.
Does this checklist cover Amazon ads or paid social?
No. Amazon has different economics and is covered by the break-even ACoS calculator. Paid social is not audited here, though the profitability section applies to any channel unchanged.
- Keep going
Related reading
GUIDE
Titles, attributes and structure, and why the feed is the campaign rather than an input to it.
TOOL
The ROAS below which every extra order costs you money, from your own margin and return rate.
TOOL
The Amazon equivalent, with referral fees and FBA costs taken out before the target is set.
Ran the audit and want it checked?
The free audit covers the account and the feed with the specialists who manage eCommerce paid media daily, plus five other disciplines, ranked into one constraint.