Blog

How to find which traffic sources bring paying customers

AnalyticFast Team · 2026-10-31

A practical guide by the AnalyticFast Team. Examples are illustrative; this article reports no original research or customer performance results.

A source with the most visitors may not bring the most customers. To decide where to spend your next afternoon, connect observed visits to verified payments and read the result alongside traffic volume, refunds and missing attribution.

1. Define the payment you want to measure

Choose a consistent financial event: a successful payment, not a Buy click, signup or checkout session creation. Decide whether the report is about new customers, all collected payments or net revenue after refunds. Keep currencies separate and write down the reporting timezone. Those choices prevent a launch-day traffic report from being compared with an unrelated payout total.

2. Label the links you control

Use a small UTM convention: source for the platform, medium for distribution, campaign for the initiative, and content for the individual placement. Keep a link register so x and twitter do not become accidental competing labels. Test the final landing URL after shorteners and redirects. Never put an email address or token in campaign parameters. Google’s campaign URL guide explains the parameter roles.

3. Carry browser context into checkout

Install one tracker, verify a real public-page visit, then read its optional attribution IDs before your server creates checkout. Put those IDs in the provider’s supported custom metadata fields. A referrer may identify a platform, but it cannot reliably identify which person completed an external checkout. Consult the provider-specific handoff; arbitrary URL parameters are not supported everywhere.

4. Verify payment, retry and refund behavior

On a separate analytics test site, arrive with a unique campaign and complete a sandbox payment. Confirm the provider’s signed delivery and the payment’s source in the dashboard. Replay the same delivery and issue a refund. Net revenue should change only for the distinct financial events. For subscriptions, include a later renewal test instead of assuming that the initial payment covers it.

5. Compare a small set of decision metrics

The following numbers are illustrative, not customer results. Suppose a community post brings 200 visitors, four new customers and $120 net revenue; a broad launch brings 1,000 visitors, five customers and $150 net. The first has a 2% paid conversion rate and $0.60 revenue per visitor; the second has 0.5% and $0.15. The broader launch still brings more total revenue. Your next experiment depends on effort, cost and whether either result repeats.

Inspect both first touch and last non-direct touch. Discovery channels and conversion-oriented messages can receive different credit for the same payment. Do not add model totals together. Give delayed purchases a consistent observation period and keep unknown-source revenue visible.

6. Change one thing and keep a decision log

Record the channel, audience, placement, time spent, spend if any, and the question you tested. Avoid promoting a channel after one high-value purchase or rejecting it before typical buyers have time to return. Attribution describes observed paths; a controlled experiment is needed for a causal lift claim. The useful result is a better next decision, with its uncertainty written down.

Add your website to AnalyticFast, or start with the installation guides and payment integrations.