The path to purchase is not a funnel

Every team has seen the funnel. Awareness at the top, a narrowing middle, a conversion at the bottom. It is clean, it is easy to draw, and it has shaped how marketing budgets get split for decades. The problem is that almost nobody buys that way, and pretending they do leads to a long list of confident decisions built on a shape that was never real.

We are not arguing that stages do not exist. People do move from not knowing you, to considering you, to buying. What we are arguing is that the movement is not orderly, it is not one-directional, and a large share of it happens in places your analytics will never record. If your picture of the journey is wrong, every number you hang off it is wrong too.

The funnel was always a metaphor, not a map

The original funnel was a sales-training device from over a century ago. It was meant to describe a salesperson's mindset, not to model how a market behaves. Somewhere along the way it stopped being a teaching aid and became the data model. Tools were built to report against it, dashboards were laid out to mirror it, and budgets were carved into top, middle, and bottom as if those were real, separable buckets.

A metaphor is fine until you start dividing real money by it. The moment you say "we are spending too much at the top and not enough at the bottom," you have treated a drawing as a measurement. The honest version of that sentence is usually "we do not know which of our spend created demand and which just collected it."

What a real journey looks like

Watch a single genuine purchase closely and the tidy triangle falls apart. The journey is almost always four things the funnel is not.

Nonlinear. People loop. They consider, leave, forget, come back warmer, compare you to two competitors, drop off for a month, then convert on a Tuesday after a podcast mention. There is no single arrow.

Multi-session and multi-device. The research happens on a phone at night, the decision on a laptop at work, the purchase on a third device entirely. Each of those can land in your data as a different stranger unless you have stitched identity deliberately.

Mostly invisible. The touches that move people are often the ones you cannot tag: a recommendation in a group chat, a comment under someone else's post, a conversation with a colleague.

Demand-led, not channel-led. The channel that gets the last click is frequently not the channel that created the want. Branded search is the clearest example, and we will come back to it.

Google's own research has a name for the part the funnel skips. They call it the messy middle: the long, looping stretch between a trigger and a purchase where people swing between exploring options and narrowing them down, sometimes for weeks, sometimes across dozens of sources. That middle is where most of the real persuasion happens, and it is the part your stage-based reporting compresses into a single bar.

The touches you never see

Here is the uncomfortable part for anyone who reports on marketing. A meaningful slice of what drives purchases is structurally unmeasurable with standard tracking, and that slice has grown.

Dark social is the catch-all term for sharing that happens in private channels: direct messages, group chats, email forwards, closed communities. A link passed between two friends in a messaging app arrives at your site with no referrer, so your analytics files it under direct traffic. It looks like someone who typed your URL from memory. It was actually a recommendation, which is the most valuable touch there is, and you have just recorded it as noise.

Word of mouth and offline conversation leave no digital trace at all. Someone hears about you at dinner and searches your name three days later. Your data sees an organic branded search and a conversion. It does not see the dinner.

Branded search deserves special attention because it is the great misattributor. When someone searches your brand name and clicks an ad or a listing, that click looks like a conversion source. It is almost always a lagging indicator. The brand search is the receipt for demand that some other touch created. Paying to capture branded search and then crediting that spend with the sale is one of the most common ways teams convince themselves a channel works when it is mostly intercepting people who were already coming.

Why getting the journey wrong costs real money

This is not an academic complaint. A wrong journey map produces wrong budget decisions in a predictable direction. Channels that capture existing demand, retargeting, branded search, bottom-of-funnel placements, get over-credited because they sit closest to the conversion. Channels that create demand get under-credited because their influence happens earlier, in sessions that were never connected to the eventual sale.

So the budget drifts toward harvesting and away from planting. For a while the numbers look great, because harvesting is efficient by definition. Then growth stalls, nobody can say why, and the dashboards still show healthy returns on the channels that were only ever collecting a harvest someone else planted. The map said those channels were the engine. They were the exhaust.

How we map the journey properly

You cannot eliminate the invisible touches, but you can stop letting them corrupt everything else, and you can build a picture honest enough to budget against. That work has four parts.

1. Stitch identity before you trust a single path

If one person across three devices and four sessions shows up as four people, your journey data is fiction before you even start. The first job is durable identity: a logged-in or first-party identifier carried server-side so sessions connect into a person rather than scattering into anonymous fragments. Most broken journey analysis is really broken identity, and it is fixable.

2. Build an event taxonomy that matches behavior

Off-the-shelf tracking measures pageviews and a purchase and very little of the considered behavior in between. A journey worth reading needs events that capture intent: comparison views, depth of engagement, returns after a gap, the specific actions that in your business actually precede a decision. You define those from how people really buy, not from what the tag fired by default.

3. Blend the quantitative with the qualitative

The data will never show you the dinner conversation. A two-question post-purchase survey will. "How did you first hear about us" and "what nearly stopped you" recover exactly the touches your tracking cannot see, and they correct for the branded-search illusion better than any model. The strongest journey maps triangulate behavioral data with what customers tell you directly.

4. Use holdouts to separate cause from credit

The only reliable way to know whether a channel creates demand or just collects it is to turn it off for a slice of your audience and watch what happens. If sales hold steady without it, that channel was harvesting. If they fall, it was planting. A holdout settles arguments that attribution models will debate forever, and it does it with evidence instead of assumptions.

What good looks like

A team that has done this work does not have a perfect view of every customer's path, because that does not exist. What they have is a journey model that is honest about its blind spots, identity that holds people together across sessions, events that capture real intent, survey signal that catches the invisible touches, and the discipline to test causation rather than assume it.

From there, budget conversations change. Instead of "we are under-invested at the bottom of the funnel," the sentence becomes "we have measured which spend creates demand and which captures it, and here is what happens when we shift the balance." That is a defensible decision. The funnel chart never was.

The path to purchase is not a funnel. It is a map, most of it is faint, and the value of a good analyst is drawing the parts that can be drawn and being clear about the parts that cannot. Budgets built on the honest map beat budgets built on the pretty one every time.