A modern sales funnel is most useful as a set of commercial states, not a literal recording of the order in which someone researches. Buyers may compare, pause, ask colleagues and return to earlier questions. Keep the stage definitions stable while allowing the route between them to vary.
Google’s 2020 research on the “messy middle” describes repeated exploration and evaluation between an initial trigger and a purchase. That model helps explain why a straight diagram can hide real behaviour; it is not proof that every purchase follows an identical loop.
What a nonlinear journey can look like
In a fictional software purchase, an employee sees a colleague’s recommendation, reads a comparison, tries the product and shares a security question with IT. Procurement then asks for pricing. The employee returns to an introductory guide to explain the tool internally. A visit to introductory content at that point does not necessarily mean the opportunity has moved back to an early stage.
Separate three kinds of information
- Observed behaviour: a recorded page visit, trial or submitted question.
- Commercial state: an opportunity accepted by sales, a proposal issued or a decision pending.
- Interpretation: a hypothesis about the person’s motivation.
A pricing-page visit is evidence of a visit, not confirmation of budget or buying authority. Ask the relevant question before changing the sales state solely because a behavioural score increased.
Design for returns and multiple people
Make comparisons, implementation details and terms easy to revisit and share. Keep approved information consistent across the website, demonstrations and proposals. Record which buying question remains unresolved, who can answer it and which stakeholder needs the answer.
For account-based sales, distinguish people from companies. Several contacts may contribute to one buying decision; merging every interaction into a single supposed person creates a false history.
Measure without claiming a complete journey
Use stage timestamps and cohorts to study progression. Retain unknown sources and offline influences as unknown. Attribution reports allocate credit according to a model; they do not establish what would have happened without a channel.
A useful review asks why suitable opportunities stalled, which evidence helped a decision and where handoffs failed. Pair those observations with costs and outcomes. More touchpoints are not automatically better if they do not help the buyer resolve a question.
