How to Tell Which Marketing Channel Is Actually Working: UTMs, GA4 and Attribution for Founders Who Are Not Analysts

Here is how to know which marketing channel is working, in three steps. One: put UTM parameters on every link you control, so traffic arrives labeled instead of anonymous. Two: read the Traffic acquisition report in GA4 (Reports, then Acquisition, then Traffic acquisition) and sort by your money event, not by sessions. Three: add a required free-text "How did you hear about us?" field to your checkout or signup, because customers will tell you things no tracker can see. Where the analytics and the answers agree, spend more. That is the whole system.

You do not need an attribution platform or a data warehouse. Those exist for companies spending millions across forty channels. You are answering one question: of the handful of things I am doing, which ones bring people who pay? That is answerable with free tools you already have, in about an afternoon.

This guide walks through each step properly: what UTM parameters actually are and the naming conventions that keep your data readable, which GA4 report answers "where do customers come from" (there are two that look identical and answer different questions), why last-click measurement quietly lies to you and why that is acceptable at your size, why the numbers from Meta, GA4 and Stripe will never match and how much divergence is normal, and how to see the new stuff: visitors arriving from ChatGPT, and GA4's built-in AI analyst that now answers plain-English questions. Written for a founder doing their own marketing, probably at 11pm, with real money on the line.

What are UTM parameters and how do you actually use them?

UTM parameters are short labels you append to a URL so analytics tools know where the click came from. When someone visits yoursite.com/?utm_source=newsletter&utm_medium=email&utm_campaign=aug-launch, GA4 files that visit under newsletter, email, aug-launch instead of guessing from the referrer. No code, no tags on your site. The label lives in the link itself.

Google's own campaign URL documentation lists the parameters GA4 reads. Five do almost all the work:

Google's guidance: always set source, medium and campaign; term and content are optional extras. The three links you will build most often:

• Email newsletter button: ?utm_source=newsletter&utm_medium=email&utm_campaign=aug-launch
• Instagram bio link: ?utm_source=instagram&utm_medium=social&utm_campaign=bio-link
• Meta ad: ?utm_source=facebook&utm_medium=paid&utm_campaign=aug-prospecting&utm_content=ugc-video-1

The rule that makes this work: tag every link you control, and never tag internal links (links from one page of your site to another), because that overwrites the visitor's real origin mid-session.

Do this now: list every place you post links (newsletter, bio links, ads, guest posts, your email signature) and build a tagged URL for each one tonight. Untagged links are votes you threw away.

Which GA4 report shows where customers come from?

GA4 has two acquisition reports that look nearly identical, and picking the wrong one is the most common way founders confuse themselves.

• Traffic acquisition (docs) answers: where did this visit come from? It uses session-scoped dimensions like Session source and Session medium, and counts every visit, new or returning.
• User acquisition (docs) answers: where did this person come from the very first time? It uses First user source / medium, locked to each user's first-ever visit.

For "which channel is working", start with Traffic acquisition. Step by step:

1. Open GA4, click Reports in the left rail, then Acquisition, then Traffic acquisition.
2. Set the date range to the last 28 or 90 days. Ignore anything shorter; small sites need time for patterns to be real.
3. The table opens on Default channel group. Switch the first column's dropdown to Session source / medium: channel groups are Google's guess, source/medium is your actual tagging.
4. In the Key events column, pick your money event (purchase, signup, lead) and sort by it.
5. Read that column, not sessions. 300 visits and 20 purchases beats 5,000 visits and 2.

Then check User acquisition once a month for a different question: which channel introduces us to people? If a source dominates first visits but rarely closing visits, it is your discovery channel. Do not kill it for failing to close.

Do this now: open Traffic acquisition, switch the dimension to Session source / medium, sort by your key event, and screenshot the table. That screenshot is your baseline for every future "is this working" argument.

What UTM naming conventions keep data clean?

GA4 treats UTM values as case sensitive: utm_source=Instagram and utm_source=instagram become two separate rows, and your report splits one channel into fragments. Every messy analytics account dies from a thousand of these cuts. The conventions:

• Lowercase, always. Every parameter, every time. This one rule prevents most of the damage.
• One name per channel, forever. Pick facebook or fb, never both. Pick newsletter, not sometimes email-list.
• Hyphens, not spaces. Spaces become %20 in URLs and look like garbage in reports. aug-launch, not Aug Launch.
• Source is the place, medium is the type. Source: instagram, newsletter, google. Medium: one of a tiny fixed set, email, social, paid, cpc, referral. Do not invent new mediums per campaign.
• Campaign names that sort. Start with the date: 2026-08-launch, 2026-09-black-friday-teaser. Future you reads reports chronologically.
• Keep a shared sheet. One spreadsheet with columns for URL, source, medium, campaign, and the final tagged link. Every link gets built from the sheet, even when it is just you, because 11pm you freelances.

Do this now: create that spreadsheet, write your five allowed mediums in the first row as law, and rebuild your current live links from it.

Why last-click attribution lies to you, and why that is fine at your size

Every table in the last section quietly assigns each conversion to one channel. GA4 offers three attribution models: data-driven (a machine-learned split of credit), paid and organic last click, and Google paid channels last click. The older models, first click, linear, time decay, position-based, were removed in November 2023. Session dimensions in Traffic acquisition are effectively last-click thinking: the visit that converted gets the trophy.

Last click rewards the closer and erases the introducer. Someone hears about you on a podcast, sees your posts for a month, googles your name and buys. Google gets the credit; the podcast that did the persuading shows nothing. A 2025 Refine Labs analysis put numbers on the gap: their attribution software credited web search with 78 percent of conversions, while customers themselves said search drove only 12 percent, with about 85 percent naming sources like social content, podcasts and word of mouth that tracking barely sees. GA4 also excludes direct visits from credit unless there is nothing else, so your told-a-friend traffic partly vanishes into other channels' totals.

Here is the permission slip: at your size, this is fine. Rand Fishkin has argued for years that chasing provable attribution is a boondoggle for companies under serious scale, because high-fidelity tracking costs more to build and maintain than the value it returns. With 30 conversions a week, the difference between attribution models is statistical noise. What you need is a consistent ruler, applied the same way every month, so you can see direction: channel A is growing, channel B never converts. Last click, plus the correction in the next section, is that ruler.

Do this now: write down your breakeven cost per customer from actual margin. Every attribution debate gets shorter when there is a number to check channels against.