Funnel Reports: Building the View You Will Actually Use

Build your funnel report from the steps a buyer actually takes, in order: session, product or service view, add or enquiry start, checkout or form start, completion. Save it as a standard view, with device comparison one click away, and read it the same way every time: step rates first, in order, against the previous period and the baseline. A funnel report fails when its steps are the tool's defaults rather than your path, when it mixes devices into a blend, or when it is rebuilt ad hoc each month so its definitions wander. Build once, verify the counts against known orders, then let repetition do the work. The report's job is to name the weak step, reliably, every month.

Walk your own funnel as a customer and list the commitments in sequence. For a store that is usually view, add, begin checkout, purchase. For a service business: service page view, form start, form submit, qualified conversation, where the later steps may live partly outside analytics and should be noted as such rather than faked. Keep the count of steps small; every added step must be a place where a decision or a drop can happen. Where a step cannot be measured cleanly, say so on the report's companion note instead of inserting a proxy that will later be mistaken for the real step. An honest gap in a funnel is information; a fabricated step is a future argument.

Use your analytics tool's funnel or exploration builder as its current documentation describes, since interfaces and step-matching rules change: Google Analytics support documents the current funnel exploration for that stack. Set the step match rules deliberately, deciding whether a visitor must complete steps in order, and choose the counting basis, users or sessions, then write it down. Verify the final step against ground truth: completed orders in your platform for the same period should reconcile closely, and any gap gets investigated now, using tracking traps, not after a quarter of decisions has been built on the view. Save the report, name it plainly, and forbid casual copies that fork its definitions.

The funnel graphic invites a single judgement, big or small; the step table rewards a better one. Read each conversion between adjacent steps and compare it with its own history. The weakest step relative to its neighbours and its own past is the candidate for work, read alongside traffic volume reaching it, because a weak step few reach may matter less than a mediocre step everyone meets. Watch also for a step improving while the next deteriorates by the same amount, the signature of a step made easier in a way that passes uncommitted visitors forward. That pattern is not progress; it is displacement, and only step-by-step reading, rather than total-watching, catches it.

Keep the saved funnel whole-site, then apply comparisons inside it: device first, because device asymmetry is the commonest actionable pattern, then your largest source where the question demands it. Resist building twelve saved funnels for every segment; they will diverge in maintenance and nobody will know which is canonical. One funnel, standard comparisons, in a fixed order, is the design that survives a year of monthly reviews like those in the five reports. When a segment's funnel becomes a standing decision input, for example mobile checkout work running for a quarter, that is the moment it earns a saved view of its own, with an owner and a retirement date.

Tools offer choices that quietly change the story: an open funnel counts visitors who enter at any step, a closed funnel only those who begin at the first; date ranges can assign conversions to the session that started the journey or the one that completed it. Neither choice is wrong, but each must be chosen once, documented, and left alone, because flipping them between months manufactures movement. Note the choices in the definitions document from analytics setup. When the platform changes a default in an update, and platforms do, the note is what lets you spot that your funnel moved because its ruler changed, not because your visitors did.

End every funnel reading with a written line: the step chosen for attention, or an explicit no action, with the reason. Connect the step to its evidence companions: recordings for what the loss looks like, in recordings and heatmaps, and the post-purchase ranking for what buyers say nearly stopped them. The funnel alone justifies investigation, rarely a specific fix. That restraint is a feature: it keeps the monthly review honest about the difference between locating a loss and explaining it. Over quarters, the sequence of chosen steps becomes a map of the business's education, each step attended in turn, none revisited without new evidence. That is what a working funnel report is for.