The visit that converts is rarely the visit that mattered.

For most SaaS businesses the conversion is a trial start or a booked demo. The mis-definition is usually not about which one — it is about picking the version that is easy to count rather than the version that correlates with revenue.

A raw signup is trivial to instrument and, in a product-led motion, close to worthless as a ranking axis. Low-friction signup forms attract exactly the traffic that will never activate, so channels optimised for signup volume outrank channels producing customers. The same failure in a sales-assisted motion looks like counting every demo request, including the ones that no-show and the ones that were never qualified.

The fix is to move the definition one step closer to revenue: an activated trial rather than a started one, a held demo rather than a booked one. That costs you reporting latency, because the qualifying state arrives later than the click. It buys you a ranking that survives contact with the sales team.

The path is longer than the attribution window. A considered B2B software purchase involves several visits over weeks, often from more than one device, and usually with a long gap in the middle while an internal conversation happens. Default attribution windows were designed for shorter cycles. When the window is shorter than the real path, the discovery channel gets no credit at all — and the channel that gets all of it is whatever brought the buyer back at the end, which is reliably brand search or direct.

Content is impossible to defend. Everyone knows the blog contributes. Almost nobody can show it, because the article was read in session one and the signup happened in session four. Last-click reporting shows content producing nearly no conversions, which is how content budgets get cut in businesses where content was the thing that worked.

Demos booked off-site are invisible. Where the demo is booked through an embedded scheduler or on a vendor domain, the conversion often does not fire. The meetings appear in someone's calendar and never in the analytics, and the gap is systematically larger for the channels that send higher-intent traffic straight to the booking page.

• The conversion event set at the state that correlates with revenue — activated trial or held demo — rather than the state that is easiest to fire.
• Every booking and signup path instrumented, including embedded schedulers and flows that complete on a vendor domain.
• An attribution window derived from your actual observed path length rather than a platform default.
• Multi-session paths held together across the longest window the collection layer can honestly support, with the limits of that stated rather than hidden.
• Content grouped by role in the path — discovery, evaluation, decision — rather than by URL structure.
• Behavioural scoring across every session in the path, so a visitor's accumulated engagement is visible rather than only their last visit.

Source ranking against the conversion event, with first-touch and converting-touch shown side by side rather than as a choice. For a long-path business, the disagreement between those two columns is the most useful thing on the page: it is where you find the channels that start deals and never get credited for them.

Below that, content performance grouped by role in the path, so an article that never converts anyone directly but appears in most converting paths shows up as what it is.

Then the visitor list, ranked by conversion likelihood. In a sales-assisted motion this is the view that gets used daily — accounts showing sustained engagement without having booked anything are the clearest outreach list a marketing site can produce.

Multi-session measurement is bounded by browser storage. Someone who researches on a work laptop and signs up on a personal one is two visitors, and no collection architecture resolves that without asking people to identify themselves.

For most SaaS businesses this is a manageable amount of noise. For a few — enterprise sales, very long cycles, heavy mobile-to-desktop switching — it is large enough that source attribution should be treated as directional rather than precise. We will tell you which of those you are on the call, because it changes how much weight the ranking deserves.