What the funnel framework is
The funnel shape describes the stages a buyer passes through from attention to action; the AIDA framework applies the same sequence inside a single message. As a marketing framework, the funnel is less a theory of persuasion than a measurement structure. It answers three questions: what are the stages, who is in each one, and what share of the people in one stage reach the next. Everything else the funnel is used for, from budgeting to diagnosis, rests on those three answers.
The marketing funnel basics guide introduces the stages and what each is for. This page treats the funnel as a thinking tool. What it makes visible is that outcomes are multiplicative. If each stage passes a fraction of the previous one, the final count is the top volume multiplied by every rate in turn, and improving any single rate improves the end. That simple structure is what lets a team locate a problem instead of arguing about it.
What the funnel hides is just as important. It does not explain why people move or stall, it treats every person in a stage as equivalent, and it says nothing about time, about people who leave and return, or about customers who bring others in. Treated as a map of how customers think, it misleads; treated as a set of counters with clear definitions, it is one of the most useful tools in marketing. The author's marketing funnel expertise page describes how this framing is applied in practice.
Stages and definitions
| Stage | Entry rule (who counts) | What moves someone to the next stage | Where the count comes from |
|---|---|---|---|
| Exposure | Has had a measurable opportunity to see the business: an impression, a listing view, a mention | Attention and a reason to act | Ad platforms, search impressions, social reach |
| Visit | Has arrived on an owned property: site, app, store, profile | Finding something relevant on arrival | Analytics sessions or users, footfall |
| Engagement | Has done something beyond arriving: a second page, a scroll depth, a video view, a search | Interest in a specific need the business can meet | Analytics events defined in a tracking plan |
| Lead or intent | Has identified themselves or signalled intent: sign-up, enquiry, add to basket, trial start | Enough confidence to commit a small step | CRM, forms, commerce events |
| Conversion | Has completed the transaction the business counts: order, contract, subscription | Price, trust and the removal of friction | Order system, CRM closed-won, billing |
| Retention | Has returned or renewed within a defined period | Value delivered and a reason to come back | Billing, repeat-order data, product usage |
| Advocacy | Has referred, reviewed or recommended | Satisfaction worth talking about | Referral tracking, review platforms, surveys |
The column that matters is the entry rule. A stage without a written rule for who counts cannot be measured consistently, and most funnel arguments are really arguments about definitions: whether a newsletter sign-up is a lead, whether a trial that was never opened counts as intent, whether a return within a year is retention. Write the rule, name the data source, and record the date it was agreed. The TOFU, MOFU and BOFU framework groups these stages into three bands and maps content to each; the funnel framework keeps them separate because the rates between them are where diagnosis happens.
Stage definitions are also business-specific. A retailer's intent stage is a basket; a business-to-business firm's is a demo request; a subscription app's is a trial. The table is a template, and the first task when using the framework is to rewrite it in the business's own terms.
Metrics between stages
The funnel's native metric is the rate between adjacent stages: of those who were in one, how many reached the next within the period. Each rate is a CVRShare of visits or clicks that convert. in the general sense, and the familiar site-level conversion rate is just one of them, usually visit to conversion. Reporting every adjacent rate, rather than the overall one, is what turns the funnel from a scorecard into a diagnostic. An overall rate can fall while most stages improve, if one stage collapses; only the adjacent rates show which.
Two more measures belong beside the rates. Volume at each stage, because a rate without a count is uninterpretable: a stage that converts well from a trickle is not a strength. And time between stages, because a funnel that converts the same share more slowly is a different funnel, and because lag decides how long a change takes to show at the bottom. Cost attaches naturally too: spend divided by the count at any stage gives a cost per stage, with CACCost to acquire one new customer. being the version at conversion.
Choose the rates that become a KPIA metric tied directly to an objective. carefully. A team held to the overall conversion rate will be tempted to shrink the top of the funnel to flatter it. A team held to the adjacent rates, with volume alongside, has nowhere to hide. Segmenting the rates by channel, device, new against returning and product line is the next step, since an average rate conceals that one segment is broken and another is fine.
Finding the constraint
With volumes and rates in hand, the funnel becomes a search for the constraint: the stage where an achievable improvement adds the most to the final outcome. The lowest rate is a candidate but not the answer. The question for each stage is how far its rate sits below what is realistically achievable, how much volume passes through it, and what an improvement would cost. A stage with a low rate that is already near its ceiling is not a constraint; a stage with a respectable rate that is well below what comparable pages or segments achieve, and that carries most of the volume, usually is.
The arithmetic can be done on a single sheet. Model the funnel as it is, then change one rate at a time to a defensible target and read the effect at the bottom. The stage whose plausible change moves the final number most is where effort goes first. Repeating the exercise after each fix finds the next constraint, which is why funnel work is a cycle rather than a project. The funnel and checkout optimisation hub applies this biggest-leak logic to the stages a website controls. For stages that live in memory rather than on a site, the guide to reading a brand funnel from tracker results compares awareness, consideration and preference conversions with each competitor's in the same survey.
Constraints are not always rates. A stage can be capacity-bound: sales cannot handle more leads, stock cannot fulfil more orders, support cannot onboard more accounts. Pushing more volume into a capacity-bound stage lowers its rate without anyone having done anything wrong. The funnel shows the symptom; the cause is found by asking the people who own the stage.
Worked funnel (illustrative)
A subscription software company locates its constraint
A subscription software company for small accountancy practices lays out a quarter's funnel. Visits: 96,000. Trial sign-ups: 2,880 (3.0%). Activated trials, defined as creating a client record in the first 7 days: 1,210 (42%). Paid conversions: 290 (24% of activated trials). The team's instinct is to fix the 3.0% visit-to-trial rate, which looks lowest. The modelling says otherwise. Lifting visit-to-trial to 3.5% with no other change adds 48 paying customers a quarter. Lifting activation from 42% to 55%, which product analytics suggests is achievable because 31% of unactivated trials never completed the import step, adds 90. Lifting activated-to-paid from 24% to 28% adds 48. Activation carries the volume and sits furthest from its plausible ceiling, so it is the constraint. The team rebuilds the import step, re-measures after 8 weeks at 53% activation, and the next pass of the model shows visit-to-trial has become the constraint. Cost per paying customer fell from £310 to £241 without any change to spend.
Criticisms and alternatives
The standard criticism is that people do not move in a straight line. They loop, skip stages, leave for months and return, and consult sources the funnel cannot see. The criticism is correct about behaviour and mostly beside the point about counting: if stages are defined as states rather than steps, a person is always in exactly one of them and the counts remain valid. What is lost is explanation, and the journey map is the tool that restores it.
A deeper criticism is that the funnel ends at conversion and treats customers as output. Everything they do afterwards, including bringing in the next customer, sits outside the model or in a retention stage bolted on at the bottom. The growth loops framework answers this by closing the shape: the output of one cycle becomes the input of the next, and growth compounds rather than merely accumulating. The flywheel framework makes a similar move with momentum and friction. Neither replaces the funnel's counting; both replace its picture of where customers come from.
A third criticism is that the funnel encourages buying the top and squeezing the middle, because that is what it makes measurable, while under-valuing brand work whose effect shows as a slow rise in rates everywhere. That is a fair description of how the funnel is often used and not a flaw in the framework. See-Think-Do-Care, which sorts the audience by intent and insists each cluster be measured on its own terms, is the usual corrective. In practice most teams run a funnel for counting, a journey map for explanation and a loop or intent model for strategy, and the mistake is expecting any one of them to do all three jobs. For the specific choice between this framework and the flywheel, the comparison of the marketing funnel and the flywheel has a decision table by planning question and business type.
Common mistakes
- Stages without entry rules. If nobody can say who counts as a lead, the lead-to-conversion rate is fiction.
- Reporting only the overall rate. The adjacent rates locate problems; the overall rate hides them.
- Rates without volumes. A stage that converts well from almost nobody is not a strength.
- Fixing the lowest rate. The constraint is the stage where an achievable change moves the bottom most, which the arithmetic finds and instinct does not.
- Ignoring time. Lag between stages decides when a fix shows and whether this quarter's bottom reflects this quarter's top.
- Reading the funnel as psychology. It counts; it does not explain. Pair it with research before deciding why a stage leaks.
- Changing definitions mid-period. A new rule for who counts as a lead resets every rate that touches it; record the date and compare only across periods that share the rule.
