The e-commerce growth engine: how to run marketing like a product team

Most stores run marketing like a calendar. A sale in March, a new campaign in May, a redesign when there’s time. Each one launches, nobody checks what it did, and the next idea starts from scratch.

Software teams stopped working like that years ago. They work from one goal, one backlog, small weekly releases and a review on a fixed day. This is the same method for an online store, the way we run it for clients.

Why campaigns stall and products don’t

A campaign is a project with an end date. When it ends, the learning ends with it. A product team never really finishes: it ships, measures and ships again, so every week builds on the one before.

The difference comes down to three habits:

• One goal at a time. Everyone knows which number matters this quarter.
• One list of work. Ideas wait in a backlog, not in someone’s inbox.
• A fixed rhythm. Work ships every week, and results are read on the same day every week.

Step 1: Pick one number for the quarter

Choose a number that pays the bills and moves within a week, so you can see your work move it. For most stores it is one of these:

• New-customer revenue
• Cost per new customer
• Contribution margin after ad spend

Write the goal as a sentence with a date: “New-customer revenue from [$40k] to [$55k] a month by the end of March.” If you have used OKRs you will know the shape: one objective, and a few results you can count.

Step 2: Keep one backlog

Every idea goes into one list, whoever it comes from: the founder, the ad team, a customer email, a competitor’s new landing page. Each idea then gets the same three scores out of 10, so the loudest idea doesn’t win by default.

Score every idea the same way

Impact: how much could it move the number? Confidence: how sure are we? Ease: how fast can we ship it? Here is a slice of a sample backlog:

Re-score the top ten every Monday. A test result often changes how sure you are about the ideas next to it.

Step 3: Ship small, every week

Take the top two or three ideas and ship them this week. Each one should be small enough to finish in days and change one thing only, so the result can tell you why it happened.

Same ad, two landing pages · [14] daysSample data

A · Product pageConversion [2.1]%

B · Bundle pageConversion [2.9]%

Use the platform’s own test tool where it has one. Meta’s A/B test splits the audience into random groups that don’t overlap, so nobody sees both versions. That gives a cleaner read than switching ads on and off by hand.

Write the stop rule first

A test ends when one of these is true. Agree on it before launch, so nobody argues about it on Monday.

• Each version reached [50] orders.
• The test ran for [14] days.
• After [7] days, one version costs [30]% more per order than the other.

Step 4: Measure what you shipped

A test is only as good as the numbers behind it. Before the first release, check three things:

1. Orders in your ad accounts roughly match the orders in Shopify or WooCommerce for the same days.
2. Every campaign, ad set and email has a name you can still read in a report six months from now.
3. You can see new and returning customers separately.

Then judge the whole store, not just each platform. Total revenue divided by total ad spend (often called MER, or blended ROAS) tells you whether the marketing pays off overall, even when two platforms claim the same sale.

Here is a worked example. A store spends $10,000 on ads in a month and takes $42,000 in revenue, so its MER is 4.2. Meta reports $30,000 in sales and Google reports $25,000: $55,000 between them, more than the store sold. Both platforms count the same orders. MER can’t count them twice.

If it isn’t in Monday’s report, it didn’t happen.

Step 5: Report every Monday

One page, every Monday at 09:00. No dashboard to log into and no deck. It answers four questions, in this order:

1. The one number: where it is against the goal.
2. What won: each winner, its test and its lift.
3. What we cut: each loser, and what it taught us.
4. Where the money goes: this week’s budget changes, before they happen.

The losers stay in the report. Knowing what doesn’t work for your store is half the value of testing, and it stops the same idea coming back in six months.

Who does what

A small team can run this. What matters is that every job has one owner:

• The number: the founder or head of marketing. Sets the goal and signs off the budget moves.
• The backlog: one person keeps it scored and in order, and says no to ideas that try to skip the queue.
• The tests: whoever builds the ads, emails and pages ships them, with the stop rule written down.
• The report: whoever owns the numbers writes Monday’s page, and nobody edits the losers out.

A 12-week cycle, week by week

We run the method in 12-week cycles. The weeks overlap on purpose: testing starts before the plan is finished, and scaling starts as soon as the first winner is clear.

Five mistakes that stall the engine

1. Changing the goal mid-quarter. The tests you already ran stop adding up to anything.
2. Testing two things at once. A winner can’t tell you which change won.
3. Calling a test early. Early numbers swing a lot. Wait for the stop rule.
4. Hiding the losers. When cut tests leave the report, the same ideas come back.
5. Skipping a Monday. The rhythm is the method. One missed report soon becomes three.

What to do this week

1. Write down your one number and its goal for the next 12 weeks.
2. Put every open idea into one list and score it.
3. Ship the top idea, with one change and a stop rule.
4. Next Monday, write the one-page report, even if it’s short.

If you’d rather we ran the cycle with you, book a strategy call and bring last quarter’s numbers.