How do I track ROAS by channel?

The reason channel reporting turns into an argument is that stores try to make one number do three incompatible jobs.

The platform view is what each ad system reports about itself. Every platform uses its own attribution rules and its own conversion windows, so their totals overlap. Use this view only for comparing things inside the same account: this creative against that creative, this campaign against that one.

The analytics view puts every channel under one set of rules. It will show numbers lower than the platforms claim, and that is the point. Use it for comparing channels against each other, and for seeing which landing pages and which sources produce orders.

The blended view is total store revenue divided by total advertising spend across everything. It is the only figure that cannot double count, and it is the one to make budget decisions with.

Write down which view each report uses and put it in the title of the report. Most disputes about performance turn out to be two people quoting two different views at each other.

Channel reporting is only as good as your link tagging, and tagging discipline is where most stores lose. Decide the convention once and enforce it.

Use lower case everywhere, because tags are case sensitive and Facebook and facebook will report as two channels. Keep source as the platform, medium as the type of traffic, and campaign as something a human can read six months later. Do not put the date in the source. Do not invent new mediums per campaign.

Turn on automatic tagging where the platform offers it and you are using the matching analytics tool, and avoid tagging the same link twice by hand as well, which can break the parameters.

Tag everything, not just ads. Email, SMS, affiliate links, influencer links, QR codes on packaging, links in your own support articles. Untagged traffic lands in direct and makes direct look like a channel that is doing something, which it is not.

Keep a single shared document with the approved values, and check the source and medium report monthly for new values nobody agreed to. That check takes five minutes and prevents a year of unreadable data.

Attribution is a choice about who gets credit when several channels touched the same order, and there is no correct answer. What matters is choosing one and keeping it, so that changes over time mean something.

Last click is simple and understandable, and it under credits everything that creates demand early: discovery ads, video, and any channel where people see something and search later.

First click does the opposite and flatters awareness channels.

A data driven or positional model spreads credit and is harder to explain, but it is closer to how buying actually works when your customers take days to decide.

Whatever you pick, change it rarely and note the date when you do, because performance will appear to jump for every channel on the day you switch.

And hold one rule above all: never sum the revenue reported by each ad platform and compare it to your real revenue. That total will exceed reality, because each platform counted some of the same orders. If someone in a meeting adds platform numbers together, that is the moment to bring out the blended figure instead.

Your ecommerce platform knows exactly how many orders you took and what they were worth. That is the ground truth, and every other report should be checked against it.

Once a month, put four totals side by side for the same date range: orders and revenue from the store admin, sessions and revenue from analytics, conversions and value from each ad platform, and total spend from each platform's billing.

Expect analytics revenue to sit slightly below store revenue, because of consent choices, ad blockers and in-app browsers. A small, stable gap is normal. A gap that changes suddenly means something broke, usually a tag removed during a theme update or a consent tool reconfigured.

Check currency and taxes as well. Analytics recording revenue including tax while the store reports excluding it will produce a permanent mismatch that people waste hours on.

Also decide how refunds and cancellations are handled. A channel that brings orders which are refunded at a high rate looks better than it is until refunds are subtracted. If your platform allows it, feed refunds back so the reporting reflects net revenue.

The report that changes behaviour is short. One page, the same every week, with the same definitions.

Include total revenue, total ad spend, blended return, gross profit after ad spend, orders, new customers and average order value. Then a small table by channel showing spend, analytics attributed revenue and platform reported revenue side by side, so the difference is visible rather than argued about.

Add a note field where you write what changed that week: a budget move, a new creative set, a price change, a stock outage, a site update. Six months later that column will explain more of the chart than any model will.

Resist the urge to add more numbers. A report with forty metrics gets skimmed. A report with eight gets read and acted on.

For channels where clicks are rare and influence is real, such as video and social discovery, add a holdout or a simple geographic test occasionally rather than trying to attribute your way to certainty. Measurement of demand creation is a different exercise from attribution, and no amount of tagging will substitute for it.