Ecommerce Paid Media: The Complete Guide

The short version

Ecommerce advertising has an unusually convincing wrong answer: reported ROAS. It is easy to read, updates in real time, and is wrong in three specific directions at once.

1. It is gross of returns, and returns vary by product, channel, and creative.
2. It credits demand you already had, especially on branded and retargeting.
3. It says nothing about whether the customer comes back, which is where most of the money is.

Almost every expensive ecommerce mistake is a decision made on that number without correcting for one of the three.

Start here

• Break-Even ROAS — the number below which everything else is decoration
• Contribution Margin — what actually pays for advertising
• Returns: The Number Missing From Your ROAS

What a customer is worth

This side of the equation decides what you can pay, and it is usually estimated rather than measured.

• Repeat Purchase Rate: The Metric That Decides What You Can Pay — measure it by cohort and by acquisition source, or it will mislead you
• First-Order Profitability: When Losing Money On Order One Is Correct — a financing question, not a marketing one
• LTV and Payback Period
• AOV

The single most useful split: repeat rate by first-order discount depth. Full-price and deep-discount customers usually behave differently enough to change channel strategy.

What you are actually causing

• Incremental ROAS: The Only Return That Is Actually Yours
• Incrementality
• Demand Capture vs Demand Creation

Run incrementality on your best-reported channel first. Branded search and tight retargeting report beautifully and frequently cause the least.

The catalog

Where automated campaigns quietly spend on products that cannot pay for themselves.

• Not Every SKU Should Be Advertised — sort by contribution, not revenue
• Product Feed Quality: The Unglamorous Thing That Decides Shopping Performance — in shopping, the feed is the ad
• Shopping Feed and Performance Max

Excluding the products that lose money outperforms almost any bid optimization applied on top of them.

Price and offer

• Discount Dependency: How Promotions Become The Business — it works every time, which is the problem
• The Offer: The Lever Most Teams Never Touch — usually more headroom than the fortieth creative variant
• Social Proof: Why Most Of It Does Nothing

Creative

• The Hook: The First Two Seconds Decide The Rest
• The Creative Brief: What It Must Contain To Be Worth Writing
• Ad Placement Specs And Safe Zones
• Creative Testing

Seasonality

• Peak Season: What To Decide Before The Auction Gets Expensive
• Seasonality

The work that decides peak happens in the cheap weeks before it. Once CPMs rise, the only lever left is whether to keep bidding.

The order to work in

1. Establish break-even from real margin, net of returns.
2. Measure repeat rate by cohort and by acquisition source.
3. Exclude products that cannot pay for their own advertising.
4. Test the offer before the fortieth creative variant.
5. Run incrementality on whatever reports best.

Steps 3 and 5 are the ones most often skipped, and they are usually where the largest single improvement is sitting.

Practical rule

Correct reported ROAS for returns, incrementality, and repeat behaviour before you rank anything on it. Uncorrected, it is a plausible number pointing in the wrong direction.