Knowing where your traffic actually comes from
Per-visit records instead of aggregates, first touch instead of last click, and why a working channel can read as zero.
Most analytics answers how much traffic you got. The questions that change decisions are narrower: did that campaign produce a real person, what did the people who bought have in common, and is anyone on the site right now.
Those need per-visit records, not counts, and they need the origin captured at the moment of arrival rather than inferred later.
First touch beats last click. By the time someone buys, the referrer usually says direct, because they came back a second time. Capture the origin on the first visit, keep it, and write it onto the order. Then in two years you can still ask what produced your best customers, using the system you trust most.
Expect a large unknown bucket, and be honest about it. A substantial share of visits carry no referrer at all. Direct is not a channel, it is the absence of evidence, and reporting it as people typing your address from memory is how you talk yourself into the wrong conclusion.
This matters most for AI referrals. Traffic from an assistant frequently arrives with no referrer and lands in direct, so a channel that is genuinely producing customers reads as nothing and gets defunded. The instrument cannot represent the thing, so it reports absence.
The general rule: when a number says nothing is happening, check whether the measurement is capable of detecting it.
Need help with any of this?
These notes are free and always will be. If you would rather someone just set it up, or you are stuck on something similar, get in touch at hello@opsira.io.