eCommerce PPC & Paid Advertising Guide
Paid media is the only channel where you can lose money faster than you can read the report. This guide covers the arithmetic that sets your targets before any campaign is built, the account structure that survives platform automation, and how to tell real incremental revenue from revenue you would have had anyway.
CHAPTERS
8 chapters
READING TIME
38 min read
LAST UPDATED
18 September 2026
LEVEL
Intermediate
The most expensive decisions in a paid account are made before anyone opens the ads manager.
EXECUTIVE SUMMARY
What this guide argues, in five points
Paid media has become mostly automated. Bidding, placement and increasingly targeting are decided by the platform. What is left for a marketer to control is the inputs — the profit target, the product data, the account structure and the creative — and those are exactly the four things most accounts get least attention.
- Break-even ROAS comes out of your contribution margin. Until that number is written down, every target in the account is somebody’s preference.
- For a store, the product feed is the campaign. Shopping performance is decided by catalogue data far more than by bids or budgets.
- Structure the account by margin and by intent, not by the platform’s recommended setup. Automation optimises inside the boundaries you draw.
- Creative is the main remaining lever on paid social. Concepts, not colour variations, and taken from customers rather than competitors.
- Platform-reported ROAS double counts. Blended contribution and periodic incrementality tests are the only honest scoreboard.
The eight chapters below are the detail behind those five points, in the order we work through them when we take over a paid account.
- Contents
Eight chapters
Read it in order the first time. After that, the chapter index is the useful part.
01
Break-even ROAS falls out of contribution margin. Until it is written down, every target in the account is somebody’s preference.
4 min
02
Automation optimises inside your boundaries. Structure is now a strategic decision, not an admin one.
5 min
03
There are no keywords in Shopping. The product data is the targeting, and most catalogues are missing the fields that matter.
6 min
04
Shopping usually wins the product query. Text ads earn their place on the questions a product listing cannot answer.
4 min
05
Search harvests demand. Paid social creates it. Holding both to one ROAS target strangles the second.
5 min
06
On a marketplace the listing is the landing page, paid sales feed organic rank, and the fees change your break-even.
4 min
07
Bidding is automated. Budget is not, and an account without a written pacing model underspends and then panics.
5 min
08
Every platform claims the same order. Blended contribution and periodic incrementality tests are the only honest scoreboard.
5 min
- Before you start
Who this guide is for
Written for the people who sign off the media budget, whether or not they build the campaigns.
Founders and heads of eCommerce
You want to know whether the ROAS being reported is real, and what a defensible target actually looks like on your margin.
In-house paid media managers
You are running several platforms and want a structure and a measurement model you can defend in a board meeting.
Teams reviewing an agency
You want the questions a paid media partner should be able to answer before you sign anything.
NOT WRITTEN FOR
People looking for a bidding trick. There is not one. Once bidding is automated, the remaining levers are the profit target, the product data, the structure and the creative — and all four are slower and duller than a trick.
CHAPTERS
CHAPTER 01
4 min
Start with the profit maths, not the platform
Break-even ROAS falls out of contribution margin. Until it is written down, every target in the account is somebody’s preference.
The most common question in a new paid account is what the ROAS target should be. It is not a question of preference. Break-even ROAS falls straight out of contribution margin, and once it is written down most of the arguments about budget stop.
Break-even ROAS in one line
Break-even ROAS is one divided by your contribution margin percentage. At 30% contribution, break-even is 3.33. Anything below that is buying revenue at a loss; anything above it is contributing to overheads and profit. It is arithmetic, not strategy.
Contribution margin
30%
After goods, fees, fulfilment and returns
Break-even ROAS
3.33
One divided by contribution margin
Target ROAS at 10% profit
5.00
Leaves margin after paying for the media
Break-even at 1.9 orders per customer
1.75
If you can finance the payback period
The last row is the one that changes strategy. It is only available to businesses that actually know their repeat rate.
KEY TAKEAWAY
A 4x ROAS is excellent on a 40% contribution product and a slow way to go bankrupt on a 20% one. The number means nothing without the margin behind it.
Three targets, not one
- Break-even. The floor. Below it you are paying to give product away.
- Target. Break-even plus the profit you actually want. This is what campaigns are optimised towards.
- Growth. A deliberately lower target on new-customer campaigns, justified by repeat value and financed as an investment rather than reported as a loss.
Accounts that hold one blended target end up either starving growth or subsidising it invisibly. Three targets, written down, applied to different campaign types, makes the trade-off explicit and survivable when somebody asks why new-customer ROAS looks poor.
Our break-even ROAS calculator does this arithmetic and shows its working, including the repeat-purchase version.
CHAPTER 02
5 min
Account architecture that survives automation
Automation optimises inside your boundaries. Structure is now a strategic decision, not an admin one.
Automated bidding optimises inside whatever boundaries you draw. That makes account structure the main lever a marketer still controls: campaigns are no longer a place to manage bids, they are a place to separate things that should not be averaged together.
Separate by margin, intent and priority
Separate when
Why
Example
Contribution margin differs materially
One target cannot serve a 15% and a 45% product
Accessories split from core range
Intent differs
Branded, category and competitor traffic behave nothing alike
Branded search kept out of the growth campaign
Business priority differs
Clearance, new launches and hero lines need different treatment
New season isolated so it is not starved
Stock depth differs
Spending into a product that will sell out is waste with a delay
Low-stock lines held in their own group
Do not separate for the sake of tidiness. Every split divides the conversion data that automated bidding needs, and a campaign with too little data optimises badly.
KEY TAKEAWAY
The right number of campaigns is the smallest number that keeps genuinely different things apart. Most accounts we inherit are either one campaign doing everything or forty doing nothing.
Branded search, honestly
Branded search will always look like the best-performing thing in the account, because it is harvesting demand that other channels created. It usually deserves a small defensive budget, particularly where competitors bid on your name, but it belongs in its own campaign and it should be excluded from any report used to judge acquisition performance.
Feeding automation properly
- Send the right conversion signal. Optimising to purchase value with margin-adjusted values beats optimising to raw revenue on a mixed-margin catalogue.
- Give a campaign enough conversions to learn. If a campaign cannot reach a meaningful weekly conversion count, it is too narrow.
- Resist constant changes. Every significant edit restarts learning. Decide, then leave it alone long enough to judge.
- Use exclusions as a structural tool. What a campaign is not allowed to spend on is as much a decision as what it targets.
CHAPTER 03
6 min
Shopping, and the feed that decides it
There are no keywords in Shopping. The product data is the targeting, and most catalogues are missing the fields that matter.
On a store, Shopping is usually the largest paid channel and the feed is the campaign. There are no keywords to write: the platform matches queries to products using the data you supply. Every hour spent on titles, attributes and images returns more than an hour spent on bids.
Shopping performance is a catalogue data problem wearing a media buying costume.
The fields that decide matching
Field
What it decides
Common failure
Title
Which queries the product is eligible for
The internal product name, with the attributes buyers search by missing
Product type
Category context the platform uses to classify
Left blank, or a single flat value for the whole catalogue
Attributes
Eligibility for filtered and refined searches
Size, colour, material and gender absent on most of the range
Image
Click-through rate, more than anything else in the listing
Lifestyle crop where competitors show the product on white
Price and availability
Whether the listing serves at all
Stale feed; disapprovals nobody is monitoring
GTIN and brand
Matching confidence and comparison eligibility
Missing on own-brand ranges where it could be populated
A title rewrite across a catalogue routinely moves Shopping revenue more than any bid strategy change we have ever made.
KEY TAKEAWAY
Write titles in the order buyers search: brand, product type, then the attributes that distinguish it. The internal SKU name belongs nowhere near the feed.
Segmenting the catalogue
Feeding the whole catalogue into one campaign averages your best products with your worst. Segmenting by margin, by price band or by performance tier — using custom labels — lets the high-contribution range carry a different target from the clearance tail, which is the entire point of chapter two applied to Shopping.
- Label by contribution tier, not by revenue. A high-revenue, low-margin product does not deserve the best target in the account.
- Isolate best sellers so they are not starved by the long tail competing for the same budget.
- Watch the disapproval queue weekly. Silent disapprovals are the most common cause of a Shopping campaign quietly halving.
- Check feed freshness against stock. Advertising something you cannot ship is a refund plus a media cost.
Our Google Shopping campaign template carries a naming convention, a label scheme and a structure you can copy straight into an account.
CHAPTER 04
4 min
Paid search beyond Shopping
Shopping usually wins the product query. Text ads earn their place on the questions a product listing cannot answer.
Text search on a store sits in an awkward position. Shopping usually wins the product query outright, so paid search earns its place on the queries Shopping cannot serve: comparisons, problems, compatibility questions and category terms where the buyer is still deciding what to buy rather than where to buy it.
Where text ads still earn their place
- Category terms with a considered decision behind them, where a landing page can answer a question a product listing cannot.
- Compatibility and fitment queries, which are extremely commercial and poorly served by product listings.
- Competitor terms, where you have a genuine, defensible comparison — and where the landing page makes that comparison honestly.
- Defensive branded coverage, at a small budget, reported separately from acquisition.
KEY TAKEAWAY
Send category queries to the category page, not the home page and not a product. The most common paid search failure on a store is a good query landing on a page that asks the buyer to start again.
Landing pages are part of the media buy
The page you send paid traffic to is a media decision, not a web decision. A category page that fails the standards in our eCommerce SEO guide will also waste paid budget, because the same thinness that stops it ranking stops it converting. Fixing the page improves both channels at once, which is why conversion work usually beats the next bid increase.
Search terms, still
Automated campaign types have made the search terms report less complete, not less important. Reading it weekly and building a real negative list remains one of the few genuinely manual levers left, and on most inherited accounts it is where the first ten per cent of wasted spend is found.
CHAPTER 05
5 min
Paid social and demand creation
Search harvests demand. Paid social creates it. Holding both to one ROAS target strangles the second.
Paid social does a different job from search. Search harvests demand that already exists; paid social creates it, by putting a product in front of somebody who was not looking. That difference explains why the same ROAS target applied to both channels usually strangles one of them.
Once targeting is automated, creative is the variable that still decides outcomes.
Concepts, not variations
Producing forty assets that all make the same argument in different fonts is one test, not forty. A concept is a distinct reason to buy — a problem, a demonstration, a comparison, a proof, an objection answered. Test concepts against each other, then produce executions of whichever wins.
Concept
What it argues
Works best when
Problem
Here is the thing that annoys you and why
The product solves a nameable irritation
Demonstration
Watch it work, unedited
The benefit is visible in under five seconds
Comparison
Against the obvious alternative, honestly
You genuinely win on something buyers care about
Proof
People like you bought it and stayed
Trust is the barrier rather than awareness
Objection
The reason you are hesitating, answered
Price, fit, delivery or returns is the blocker
Take the arguments from reviews, support tickets and returns reasons. Customers phrase them better than a brief does.
KEY TAKEAWAY
Judge prospecting on new-customer cost and incremental revenue, not on blended ROAS. A prospecting campaign held to a retargeting target will be switched off before it ever works.
Retargeting, in proportion
Retargeting reports the best numbers in almost every account and creates the least new revenue. Most of the people it converts were coming back anyway. Keep it, cap it, and do not let its ROAS set the standard the rest of the account is judged against — that comparison is the single most common reason a growing account stops growing.
CHAPTER 06
4 min
Retail media and marketplace ads
On a marketplace the listing is the landing page, paid sales feed organic rank, and the fees change your break-even.
If a meaningful share of your category’s buyers start their search inside a marketplace, retail media is not optional — it is where the shelf is. It also behaves differently from search and social: the audience is already in a buying moment, and the competition is the product listed next to yours rather than a brand elsewhere on the internet.
What changes on a marketplace
- The listing is the landing page. Ad performance is capped by listing quality, images and reviews, none of which are advertising decisions.
- Ranking and advertising interact. Paid sales feed organic rank, which is why launch periods justify targets you would never accept as a steady state.
- Fees come off the top. Referral and fulfilment fees change the contribution margin, and therefore the break-even from chapter one.
- You do not own the customer. Plan for that in the unit economics rather than assuming the repeat value you get on your own store.
Our article on cutting Amazon ACoS without losing volume covers the campaign architecture behind this chapter.
CHAPTER 07
5 min
Bidding, budget and pacing
Bidding is automated. Budget is not, and an account without a written pacing model underspends and then panics.
Bidding is largely automated. Budget is not, and budget is where most of the remaining human judgement lives: how much, split how, paced how, and with what rule for stopping. Accounts without a written pacing model reliably underspend in January and panic-spend in November.
Pacing, plainly
- Plan monthly, review weekly. Daily budget reactions are noise, and they restart learning on every campaign you touch.
- Shape the year to your demand, not to the calendar. A seasonal catalogue spending twelve equal months is wrong twice a year.
- Hold a reserve. Ten to fifteen per cent unallocated covers the campaign that unexpectedly works, which is the cheapest incremental revenue you will find.
- Write the stop rule before launch. What has to be true after thirty days for this to continue. Agreeing it afterwards never happens.
KEY TAKEAWAY
Underspending is a real cost, not a saving. An account that ends the month 18% under budget bought fewer customers than it planned to and will be asked why revenue is flat.
Scaling without breaking learning
Large budget increases reset the optimisation the platform has done. Smaller, more frequent increases, applied to the campaign types that already clear target, hold performance better than a single step change — and when performance does degrade with scale, that is usually the audience genuinely being exhausted rather than a bidding problem to solve.
Our PPC budget template carries monthly pacing, a forecast and variance tracking, so the conversation is about numbers rather than impressions.
CHAPTER 08
5 min
Measuring paid media on profit
Every platform claims the same order. Blended contribution and periodic incrementality tests are the only honest scoreboard.
Every advertising platform is incentivised to claim credit, and each one does. Add up the revenue reported by search, social and retail media on most accounts and the total exceeds what the business actually took. That is not a tracking bug to fix; it is the structure of platform reporting, and the reporting model has to expect it.
Reconcile paid media to the accounts, not to the sum of the dashboards.
Three layers, three jobs
Layer
Question it answers
Cadence
Platform reporting
Is this campaign better or worse than last week?
Daily, for steering only
Blended contribution
Did the business make money after all costs?
Weekly and monthly
Incrementality testing
Would this revenue have happened anyway?
Quarterly, per channel
Using layer one to answer layer two questions is how budgets get approved for channels that are mostly taking credit.
KEY TAKEAWAY
The single most useful paid media report is blended: total media cost, total new customers, total contribution. It cannot be gamed by attribution windows and it matches the bank.
Running an incrementality test without a data science team
1
Pick one channel and one clear question. Usually retargeting or branded search, because both are suspected of taking credit.
2
Turn it off, or hold it out geographically, for long enough to matter. Two to four weeks, covering a full purchase cycle.
3
Watch total business revenue, not the channel. The channel will obviously fall. The question is whether the total does.
4
Accept the answer. A channel that costs money and moves nothing is the cheapest budget you will ever find.
EXPERT VIEW
“We ask two questions of every paid media report. Does the sum of the channels equal the money in the bank, and does the report say what we should stop doing. Most reports fail both.”
Dazzle Commerce · Paid media team
- Worked examples
Three decisions, before and after
The three arguments that come up on almost every paid account we take over.
TARGETS
One ROAS target across a mixed-margin catalogue
Before
Single 4x target across accessories at 45% margin and hardware at 18%
After
Targets set per contribution tier: 2.4x on accessories, 6.0x on hardware
A blended target overspends on the thin products and underspends on the profitable ones. Splitting by contribution moved profit without changing total budget.
SHOPPING
Bidding harder instead of fixing the feed
Before
Budget raised 40%; titles still the internal SKU names, attributes largely empty
After
Titles rewritten to brand, type and attributes; size, colour and material populated
The campaign was not losing auctions, it was not entering them. Eligibility is a data problem, and no bid fixes a product the platform cannot classify.
REPORTING
Judging prospecting against retargeting ROAS
Before
Prospecting switched off at 1.6x because retargeting reported 9x
After
Prospecting judged on new-customer cost and an incrementality holdout
Retargeting mostly re-converts people prospecting already reached. Comparing the two is how an account stops acquiring customers and then wonders why it stopped growing.
Anonymised from accounts we manage. The shapes recur; the numbers will not match yours exactly.
- Take these with you
Three checklists
Written to be worked through rather than read. Nothing here needs a tool you do not already have.
Before you spend
Two hours, once, then after any pricing change.
- Contribution margin calculated per product group, not blended
- Break-even ROAS written down and agreed
- Separate break-even, target and growth numbers defined
- Branded search isolated and excluded from acquisition reporting
- Conversion tracking reconciled against orders in the platform
- Stop rule written for every new campaign before launch
The feed, monthly
An hour that regularly outperforms a bid strategy change.
- Titles follow brand, product type, then distinguishing attributes
- Product type populated across the whole catalogue
- Size, colour, material and gender filled where applicable
- Primary images consistent with what competitors show
- Disapproval queue cleared and the cause fixed, not just resubmitted
- Custom labels reflect current contribution tiers and stock depth
Every month
Ninety minutes with the accounts open, not the dashboards.
- Blended media cost, new customers and contribution reported together
- Platform-reported revenue reconciled against actual revenue
- Search terms reviewed and negatives added
- Pacing checked against plan, with the reserve still intact
- Creative concepts refreshed rather than recoloured
- One incrementality test planned, running or read
Copy these into your own tracker. There is no download form and no email wall.
- Frameworks
Two frameworks worth keeping
The two decisions that get reargued every quarter, written down so they stop being reargued.
FRAMEWORK ONE
The target-setting table
One target across a mixed-margin catalogue is the most expensive simplification in eCommerce paid media. Set break-even from contribution, then decide the profit you want on top, then decide separately what you are willing to pay for a first order you expect to repeat.
Contribution margin
Break-even ROAS
Target at 10% profit
18%
5.56
8.33
25%
4.00
5.71
35%
2.86
3.89
50%
2.00
2.63
Growth campaigns can run below break-even on the first order where repeat value and cash flow both support it. That is a financing decision, and it should be written down as one.
On marketplaces, subtract referral and fulfilment fees from contribution first. The break-even column moves a long way.
FRAMEWORK TWO
The paid media priority order
When an account needs everything doing, the order matters more than the list. Each step assumes the one above it has cleared, because optimisation at step five is wasted while step one is wrong.
1
Measurement integrity. Conversion tracking reconciled, branded separated, contribution reportable. Nothing below can be judged without it.
2
Profit targets. Break-even, target and growth numbers set from contribution margin per product group.
3
Product data. The feed, because on a store it is the targeting and it caps everything Shopping can do.
4
Account structure. Separate by margin, intent and priority so automation optimises inside sensible boundaries.
5
Landing pages. The multiplier on every campaign at once, and usually cheaper than the next budget increase.
6
Creative. Concept testing on paid social, where the remaining variance actually lives.
7
Scale. Increase budget once the six above hold, in steps that do not reset learning.
New platforms sit outside this order. Settle these seven on the channel you already run before opening another.
- In summary
Key takeaways
Eight sentences you could hand to somebody who is not going to read the guide.
1
Break-even ROAS is one divided by contribution margin. Until that is written down, every target in the account is a preference.
2
Hold three numbers, not one: break-even, target, and a deliberately lower growth target that is financed rather than hidden.
3
On a store the feed is the campaign. Titles, product type and attributes decide eligibility, and no bid fixes a product the platform cannot classify.
4
Structure by margin, intent and priority. Automation optimises inside the boundaries you draw, so the boundaries are the strategy.
5
Send category queries to category pages. A good query landing on a page that asks the buyer to start again is wasted budget.
6
Judge prospecting on new-customer cost, never against retargeting ROAS. That comparison is how growing accounts stop growing.
7
Underspending is a cost, not a saving. Plan monthly, review weekly, and hold a reserve for the campaign that unexpectedly works.
8
Every platform claims the same order. Reconcile to the accounts and run one incrementality test a quarter.
If you want a single first action: calculate contribution margin on your top product group and compare the break-even ROAS to the target currently set in the account.
- Go deeper
Where this work sits in the practice
Three services that pick up the chapters above.
Google, Shopping, Meta and marketplace ads structured on contribution rather than on reported ROAS.
Retail media run alongside listings, pricing and Buy Box, because on a marketplace they are the same job.
The landing page multiplier from chapter four, run as a sized testing programme.
- Keep reading
Related guides
The rest of the guide library, all published in full — and the PPC budget calculator if you want the spend in this guide sized against a target.
Complete guide · 37 min
Where paid media sits in the wider mix, and how to tell whether traffic is your constraint at all.
Complete guide · 38 min
The organic counterpart: the category pages paid search should be landing on anyway.
Complete guide · 38 min
Retail media in context, where advertised sales feed ranking and the fees change your break-even.
Want this run on your account rather than read about it?
Our paid media engagements open by setting break-even ROAS from your contribution margin, then rebuilding structure and feed around it.