What does an ecommerce marketing agency actually do?

The useful version starts with the arithmetic rather than the channels. It establishes contribution profit by cohort — revenue after product cost, shipping, payment fees and acquisition — because that is the only number that says whether growth is worth having. Then it separates incremental from blended acquisition cost, so a weak channel cannot hide inside a strong one. Then it works on repeat rate and owned channels, which is the part of the demand base that compounds. The less useful version reports revenue and return on ad spend, both of which can rise while the business loses money on every order.

Why is blended CAC misleading for a DTC brand?

Because it averages a channel that creates demand with channels that harvest demand you already had. Branded search, email and returning direct traffic would have delivered many of those customers anyway, so blending them in makes paid acquisition look more efficient exactly when organic is doing well. Decisions made on the blend then overfund the channel closest to existing demand and underfund the one actually creating new demand — the opposite of what the number appears to recommend. Incrementality answers a different and more useful question: what does one more dollar produce.

What should a Shopify brand measure instead of revenue?

Four numbers. Contribution profit by cohort, after every variable cost including acquisition. Repeat rate at thirty, sixty and ninety days split by acquisition channel, because the shape is legible early and tells you whether a channel bought a customer or a transaction. Incremental rather than blended acquisition cost. And the share of revenue arriving through owned channels — email, SMS, returning direct — because that is the only demand you are not renting. Revenue and return on ad spend can both rise while all four of these deteriorate.

How much does discounting actually cost a DTC brand?

More than the margin on the discounted orders, because a predictable promotional calendar teaches customers that full price is for people not paying attention. The damage shows up gradually as a rising share of revenue sold on discount and a falling response to each successive offer, which is usually met with a deeper offer. By the time the dependency is visible in the margin line, unwinding it costs a season of revenue — which is the reason most brands that recognise the problem still do not fix it.

Is retention cheaper than acquisition in ecommerce?

Usually, and the comparison is often made too loosely to act on. The specific version worth measuring is contribution profit per cohort with and without a retention intervention, because that isolates what the work actually produced. The structural point stands regardless: a first order on a discounted entry offer is roughly break-even at best in most categories, so whether the cohort is profitable is decided after it, and a brand that only funds acquisition is buying transactions rather than customers.

Ecommerce marketing stops working when acquisition is the only lever

Acquisition cost rose faster than gross margin, so a structure that worked at a lower cost per order now loses money on the first purchase and does not say so. Revenue and return on ad spend can both rise for a year while contribution profit falls the whole time.

Five things that are true of DTC operators, four numbers worth reporting by cohort, seven questions for anyone bidding on the work, and three situations where the answer is not to hire.

Published August 23, 2026 · 9-minute read

Five things that are true of a DTC brand

The arithmetic changed and the reporting did not

Acquisition cost rose faster than gross margin across most categories, so a structure that worked at a lower cost per order now loses money on the first purchase and never says so. The dashboard still shows revenue and return on ad spend, both of which can rise while contribution profit falls.

Growth that is measured on revenue will continue right up to the point where cash runs out, and every report along the way will look encouraging.

Blended CAC hides which channel is actually working

Blending paid, organic, email and word of mouth into one acquisition number means a channel that would fail on its own is subsidised by one that would have delivered those customers anyway. The blend improves whenever the free channels do well.

Decisions made on blended cost systematically overfund the channel closest to existing demand and underfund the one creating new demand, which is the opposite of what the number appears to say.

The second order is where the business actually is

Most first orders on a discounted entry offer are roughly break-even at best. Whether the cohort is profitable is decided by the repeat rate over the following months, which is knowable early from the shape of the first thirty days.

Judging a campaign at thirty days measures the wrong thing at the wrong time. Judging it never is worse, and is what usually happens.

Discounting trains the customer to wait

A predictable promotional calendar teaches buyers that full price is for people who are not paying attention. The effect appears gradually as a rising share of revenue on discount and a falling response to each successive offer.

By the time the promotional dependency is visible in the margin, unwinding it costs a season of revenue, which is why it usually is not unwound.

The product page is the conversion asset nobody owns

Descriptions, imagery, sizing, reviews and stock accuracy do more for conversion than most campaigns, and responsibility for them is usually split between merchandising, operations and whoever built the theme.

It is the cheapest available improvement in most catalogues and it stays undone, because no single person is accountable for it and it never appears on a marketing plan.

The four numbers a Shopify marketing agency should be measured on

By cohort rather than by month. A monthly view mixes customers acquired under different economics and reports the average as though it were a trend.

Contribution profit by cohort, not revenue: Revenue after product cost, shipping, payment fees and acquisition, tracked by the month a customer first bought. This is the only number that says whether the growth is worth having, and it is routinely absent from the reporting stack entirely.
Repeat rate at 30, 60 and 90 days: Measured per cohort and per acquisition channel. The shape is legible early, which makes it the fastest honest read on whether a channel is buying customers or buying transactions.
Incremental rather than blended acquisition cost: What an additional dollar produces, not what the average dollar appears to have produced. Blending lets a weak channel hide inside a strong one, and the blend flatters exactly when organic demand is doing the work.
Share of revenue from owned channels: Email, SMS and returning direct traffic as a proportion of the total. This is the only part of the demand base that compounds rather than being rented, and it is the closest thing to an asset the marketing function builds.

Choosing an ecommerce digital marketing agency

What a firm asks for in the first week tells you what it works on. One that asks for contribution profit by cohort and repeat rate by channel is working on whether the growth is worth having. One that asks only for ad account access is working on return on ad spend, which can improve while the business loses money on every order.

So ask all seven of these, of everyone, including us.

1What is contribution profit by cohort, after shipping, payment fees and acquisition?
2What is our repeat rate at 30, 60 and 90 days, split by acquisition channel?
3What does an incremental dollar of spend produce, as distinct from the blended average?
4What share of revenue is on discount now, compared with two years ago?
5Who is accountable for product page conversion, and what have they changed this quarter?
6What proportion of revenue comes from owned channels rather than rented ones?
7Which of your deliverables would still be working twelve months after we stopped paying you?

The first is the one most brands cannot answer without a week of work. That difficulty is the finding, not an obstacle to it.

When you should not hire anyone

When contribution profit per cohort is unknown. Scaling spend against a number nobody has calculated is how a growing brand runs out of cash while the dashboard looks healthy.
While the product pages are weak. Traffic into a page that does not convert makes the acquisition problem look worse than it is and hides the cheaper fix.
When the category genuinely does not repeat. Some products are bought once, and for those the honest answer is a different business model rather than better marketing.

Common questions

What does an ecommerce marketing agency actually do?: The useful version starts with the arithmetic rather than the channels. It establishes contribution profit by cohort — revenue after product cost, shipping, payment fees and acquisition — because that is the only number that says whether growth is worth having. Then it separates incremental from blended acquisition cost, so a weak channel cannot hide inside a strong one. Then it works on repeat rate and owned channels, which is the part of the demand base that compounds. The less useful version reports revenue and return on ad spend, both of which can rise while the business loses money on every order.
Why is blended CAC misleading for a DTC brand?: Because it averages a channel that creates demand with channels that harvest demand you already had. Branded search, email and returning direct traffic would have delivered many of those customers anyway, so blending them in makes paid acquisition look more efficient exactly when organic is doing well. Decisions made on the blend then overfund the channel closest to existing demand and underfund the one actually creating new demand — the opposite of what the number appears to recommend. Incrementality answers a different and more useful question: what does one more dollar produce.
What should a Shopify brand measure instead of revenue?: Four numbers. Contribution profit by cohort, after every variable cost including acquisition. Repeat rate at thirty, sixty and ninety days split by acquisition channel, because the shape is legible early and tells you whether a channel bought a customer or a transaction. Incremental rather than blended acquisition cost. And the share of revenue arriving through owned channels — email, SMS, returning direct — because that is the only demand you are not renting. Revenue and return on ad spend can both rise while all four of these deteriorate.
How much does discounting actually cost a DTC brand?: More than the margin on the discounted orders, because a predictable promotional calendar teaches customers that full price is for people not paying attention. The damage shows up gradually as a rising share of revenue sold on discount and a falling response to each successive offer, which is usually met with a deeper offer. By the time the dependency is visible in the margin line, unwinding it costs a season of revenue — which is the reason most brands that recognise the problem still do not fix it.
Is retention cheaper than acquisition in ecommerce?: Usually, and the comparison is often made too loosely to act on. The specific version worth measuring is contribution profit per cohort with and without a retention intervention, because that isolates what the work actually produced. The structural point stands regardless: a first order on a discounted entry offer is roughly break-even at best in most categories, so whether the cohort is profitable is decided after it, and a brand that only funds acquisition is buying transactions rather than customers.

Before you take any of this on faith

The method is visible rather than asserted. howtothink.ai is a public knowledge graph built solo and running in production — 1,700 atomic lessons, roughly 9,700 generated pages, 3,300+ graph edges. It is our own build rather than a client result, which is the point: it is the part you can inspect yourself before anyone asks you for money.

What we actually do for DTC operators

Scoped to contribution profit, because revenue and return on ad spend can both rise for a year while the business loses money on every order.

Cohort economics computed from the order record rather than assembled by hand each quarter.
Incremental rather than blended acquisition cost, so a weak channel stops hiding inside a strong one.
Repeat rate at 30, 60 and 90 days split by channel, which is the fastest honest read on whether a channel buys customers or transactions.
Product data reaching every surface from one place, and lifecycle messaging running on behaviour rather than a calendar.

How retention marketing gets built

The architecture behind it — cohort economics computed from the order record rather than assembled by hand each quarter, product data that reaches every surface from one place, and lifecycle messaging that runs on behaviour rather than on a calendar — is written up in full, with the deployment order and an acceptance test on every phase.

The retention operating model Other industries

Ready to talk instead? Book the 30-minute consultation, or take the three-question diagnostic first. No email required for the diagnostic.

Also: the same problem with physical stores · the deployment blueprint