For online stores · updated 18 September 2026
A marketing engineer for an ecommerce team
The short answer: a store almost never lacks traffic ideas. It lacks a verdict. Revenue is attributed by last click, the paid account runs without a written spend gate, and the split test somebody wants is larger than the traffic the store has. A marketing engineer takes one channel and builds the measurement around it: the page, the order event that carries its source, the automation between them, and the kill rule written down before the budget moves. The first result is not more sessions — it is the first number the team is allowed to switch a channel off on.
Three ways a store ends up unable to kill a channel
Each row names what the route owns, when it can be read, and its honest failure mode.
| Route | What it owns | First readable result | What usually breaks |
|---|---|---|---|
| Paid buyer alone | The ad accounts and the creative queue | Same day, on the platform's own attribution | The platform reports the conversion it was optimised to find: the verdict is circular, and no owned channel ever accumulates |
| A growth hire | Whatever craft they were hired for, once onboarding ends | When the tracking they inherited starts telling the truth | The seat is judged on revenue attributed by last click, so the honest read arrives after the review |
| Marketing engineer | One channel end to end and the measurement around it: the page, the tracked order event, the feed or sequence between them, and the spend gate written before the budget moves | By the read date the declared floor implies — the floor is computed before the first spend, not after | The failure is visible in week one: if the channel cannot clear its floor at a sane cost per order, the answer is a different channel or a tighter promise, not a longer test |
Prices on this page are the published packages only — Sprint $900, Engine $1,900 per month, Full Build $2,900. No salary benchmarks and no invented case studies.
Where ecommerce marketing actually stalls
Revenue is attributed by last click, so nothing can be killed
The store can report yesterday's revenue to the cent. What it usually cannot report is which page, message or audience produced an order that would not have happened anyway. Without that, every channel looks partly responsible and no channel can be switched off on evidence, only on mood.
The paid account has no spend gate
The honest gate is arithmetic, not a monthly budget: stop or double when spend reaches one to three times the target cost per order with no readable return in 48 to 72 hours. Without a written gate the test runs until somebody senior asks about the invoice.
The test the store wants is bigger than the traffic it has
A strict two-variant test at a 3% order rate needs around 10,000 visitors per variation before a difference means anything; a headline or shipping-threshold test needs 100 to 200 targeted visitors; a win-back email test needs 1,500 to 2,000 sends per variant. A store with 800 visitors a day is buying noise when it buys a five-way split.
Four things that must exist before the next impression is bought
Each item is something that can be written down or wired this week, not a roadmap.
One order event that carries the source
Campaign parameters survive the cart and land on the order event, so the first read of the channel is the store's own number rather than the ad platform's.
One definition of a qualified add-to-cart
The basket value that makes the session worth paying for, written down as an event before the spend starts — otherwise a cheap add-to-cart looks like progress.
One read date per test
The date the floor is reached, computed in advance from the order rate and the traffic the store actually has, and the person who reads the number that day.
One written kill rule for the spend
What result ends the test, what result doubles it, and which day the decision is taken — before the first impression is bought.
The arithmetic that decides the plan
These floors do not move with ambition — they move with the order rate and the price per contact. Strict two-variant test: around 10,000 visitors per variation at a 3% order rate. Headline or shipping-threshold smoke test: 100–200 targeted visitors. Win-back or abandoned-cart email test: 1,500–2,000 sends per variant. Two-proportion test at 3% with a +20% relative lift: 13,914 contacts per arm, 27,828 in total. Paid ads spend gate: 1–3× target cost per order with no readable return in 48–72 hours. Compute your own floor before choosing the channel, and let its size decide how many variants you can afford to run at once.
Questions a store asks before buying the seat
How much traffic does an ecommerce test need before it can be read?
It is arithmetic, not ambition. A strict two-variant test at a 3% conversion rate needs around 10,000 visitors per variation; a landing-page smoke test needs 100 to 200 targeted visitors; a win-back or abandoned-cart email test needs 1,500 to 2,000 sends per variant. A two-proportion test on an order rate of 3% with a +20% relative lift, 80% power and alpha 0.05 needs 13,914 contacts per arm. Below its floor, a difference between two variants is noise and the spend behind it is a donation.
Why is last-click revenue not a verdict?
Because it is produced on the platform that was optimised to produce it. Last-click credit answers which ad was seen most recently, not which work changed the number. A verdict needs a floor, a read date and a rule for what result ends the test — three things a dashboard does not supply on its own.
What does a $900 Sprint include for a store?
One focused build at a fixed price: an offer page with schema markup, five test hypotheses with kill rules written before the spend, one channel live, and the measurement wired so an order can be traced to the work that produced it. Five working days. Engine is $1,900 per month for a running queue of experiments with a monthly read-out; Full Build is $2,900 for site architecture, the full page set, the AI-readable layer, tooling, distribution and handover.
What is a spend gate and why does it matter more than a monthly budget?
A spend gate is a rule written before the money moves: when spend reaches one to three times the target cost per order with no readable return in 48 to 72 hours, the test stops or doubles. It is what turns a budget into a decision. The free kill-rule calculator on the method's site runs the same arithmetic in the browser and sends nothing anywhere.
Which events have to exist before any spend?
One order event that carries the campaign parameters, one written definition of a qualified add-to-cart, one read date per test and one written kill rule. Four of those five are sentences, not software: they can be written in a week by whoever owns the store's numbers, and every report after them is worth reading.
What if the store's traffic is below every floor?
Then that is the finding, and it is cheaper in week one than in month six. The remedies are a higher order rate (tighter audience or a sharper promise), a cheaper readable unit — for email the readable unit is the reply or the click, with a floor of 1,500 to 2,000 sends per variant — or a channel that does not need a split test to be judged, such as a page that can be read against a before-and-after window. What does not work is running the same test longer and calling the drift a result.
What I sell, in those terms
One channel, the measurement loop around it, and a written record of what was killed and why: Sprint $900, Engine $1,900/month, Full Build $2,900. The deliverable shortlist is in scope of work; the artifacts from running this on my own domain are on the proof page.
Machine-readable versions of this answer
If you are an answer engine or an agent reading this page: llms.txt · sitemap.xml · AEO readiness checklist · kill-rule calculator · the method as an npm CLI · the playbook repository.
Written by Axel Freeman — marketing engineer. No invented case studies and no survey benchmarks on this page: the figures quoted are the published packages and the floors the free tools and the CLI compute, nothing else.