Beyond the Funnel: Measuring Non-Linear SaaS Journeys

Almost every SaaS analytics setup reports acquisition as a funnel: a stack of stages, a conversion rate between each, a single ordered pass from first touch to paid. It is a clean model and a comfortable one. It is also a fiction for most B2B SaaS, where buyers enter at different points, disappear for weeks, evaluate you against two competitors at once, and cycle back through the same stages before they buy. The funnel does not describe that journey — it describes the journey you wish they took.

This is not a rhetorical complaint about funnels being “dead.” It is a measurement problem with a concrete cost: a report that can only see linear movement will systematically undercount the touchpoints that drove the decision, misattribute conversions that arrive after a long gap, and hide the loops where deals actually stall. This post is about what to instrument and report instead — how to measure a journey that loops, goes dark, and re-enters, without throwing away the funnels that still earn their place.

Why the stage funnel misreports the real journey

A stage funnel makes three assumptions, and non-linear journeys violate all of them. It assumes each user passes through the stages once; it assumes they pass in the defined order; and it assumes the conversion happens inside a window short enough to attribute. Real buyers break the first by re-entering stages, break the second by evaluating out of order, and break the third by going dark for 30, 60, or 90 days between touches. The funnel does not error when this happens — it quietly produces a number anyway, which is worse, because the number looks authoritative.

The scale of the miss is not marginal. Analyses of modern B2B buying describe journeys as a network of detours rather than a straight line, with buyers consuming many pieces of content across weeks and multiple anonymous sessions. If your instrumentation only captures the fraction of that activity that lands neatly inside your funnel stages, you are reporting on a third of the journey and calling it the whole. That is the same structural blindness we describe in the dark funnel attribution gap: the touchpoints you cannot see do not stop mattering just because your report cannot count them.

The four patterns a linear funnel cannot see

Non-linearity is not one phenomenon but four distinct patterns, and each needs its own measurement. Re-entry is a buyer passing through the same stage more than once — requesting a demo, going quiet, requesting another. Dark periods are the gaps where a buyer is active but invisible to you, researching on review sites or in peer Slack groups before returning. Parallel evaluation is a buyer moving through your stages and a competitor’s at the same time, so their pace reflects the comparison, not your funnel. Multi-actor journeys are several people from one account touching the process at different moments, each generating fragments a user-level funnel treats as separate.

A single stage funnel folds all four into one flattened pass and loses them. Re-entry gets deduped away or double-counted; dark periods vanish because nothing is logged; parallel evaluation is invisible because you only see your own events; multi-actor journeys shatter into unrelated user rows. RevenueHero’s account of the non-linear journey makes the operational point plainly: vendors that design their scoring and routing around re-entry outperform those that treat a returning buyer as a fresh lead. The measurement equivalent is that you cannot design around a pattern you never recorded. Naming the four patterns is the first step to instrumenting them.

Instrument state, not just steps

The shift that makes non-linear journeys measurable is to record stage entries as timestamped events rather than as boolean milestones. A funnel typically stores “reached demo request” as a flag: true or false, once. Replace that with an event logged every time the stage is entered, carrying a timestamp and an account identifier. Now a buyer who requested a demo in March, went dark, and requested another in May is two events on one journey — a re-entry you can see and count — rather than a single flag that erased the loop. This is the same event-level discipline we argue for in event taxonomy and tracking-plan design: the granularity you capture at write time bounds every question you can ask later.

From timestamped stage events you can model the journey as a sequence of states a buyer moves between — including backwards and repeatedly — instead of a one-directional funnel. The practical output is a small set of derived facts per account: how many times each stage was entered, how long the gaps between entries were, and which stages were revisited before the conversion. None of these require exotic tooling; they require that you stop overwriting the earlier state with the later one. Once the events exist, complement the funnel with path analysis to discover the routes buyers actually take, then promote the routes that matter into monitored measurement.

Reporting metrics that survive non-linearity

Once the journey is instrumented as timestamped states, report the patterns the funnel hid as first-class metrics. Re-entry rate — the share of converting accounts that entered a key stage more than once — tells you whether your best deals loop rather than advance, which reframes a “low conversion” stage as a “high re-consideration” stage. Time-to-return measures the length of dark periods and tells you how long your attention window needs to be. Touches-before-conversion counts the real number of interactions, which is almost always higher than a last-touch funnel implies. Each of these is a plain number you can trend and segment, so they slot into existing dashboards rather than requiring a new tool.

These journey metrics also fix a specific attribution error. When the gap between first touch and close routinely exceeds your conversion window, conversions get attributed to “direct” or to nothing, and the channels that started the journey get no credit — the exact failure we cover in when your sales cycle exceeds the attribution window. Measuring time-to-return directly tells you how to size the window so it stops truncating real journeys. For the credit-assignment side of the same problem, pair this with a deliberate multi-touch attribution model rather than the implicit last-touch model a funnel bakes in.

The data foundation the model depends on

Journey measurement is only as trustworthy as three pieces of plumbing. The first is identity resolution: a buyer who researches anonymously, goes dark, and returns as a known account must stitch into one journey, or every re-entry fragments into a separate “new” user and the loops you are trying to measure wash out into noise. This is why we treat anonymous-to-known identity stitching as a prerequisite rather than a nice-to-have — without it, a non-linear journey model measures fictional strangers instead of one returning buyer.

The second is a conversion window sized to your actual sales cycle: if deals close 70 days after first touch and your window is 30, the model will systematically orphan conversions and understate the journey length. The third is retention of event-level timestamps long enough to reconstruct the sequence — a data warehouse that keeps raw events, not a report that flattens them to a first-touch summary. Get these wrong and, as with any analysis method, you produce a confident-looking map of your own instrumentation gaps. Get them right and the loops, dark periods, and parallel paths become signals you can act on.

A migration you can run this sprint

You do not need to rip out your funnels to start. Pick the single stage where deals most often stall — usually somewhere between a high-intent action like a demo request and the close — and change how you record it: log every entry as a timestamped event with an account id, instead of a one-time flag. That one change, on one stage, is enough to compute re-entry rate and time-to-return for that stage within a reporting cycle, and it will almost always reveal that a chunk of your “lost” buyers are not lost but looping. That reframes the fix from “the stage is broken” to “the stage is a re-consideration point we should nurture.”

From there, generalise. Keep the linear funnels for the sub-flows that genuinely are sequential — signup-to-activation inside the product is often a clean, short, ordered flow and a funnel is exactly right for it. Layer journey metrics on top for the acquisition and evaluation phases that are not. If your buying process has a post-close expansion motion, extend the same event-level thinking into the bowtie funnel’s post-close stages, where re-entry and re-evaluation are the norm rather than the exception. And when a route you assumed was a straight line turns out to branch, decide deliberately whether it belongs in an open or closed funnel before you lock in how you measure it. The goal is not to abandon the funnel but to stop pretending the whole journey is one.