Conversion Funnel
A conversion funnel is the structured sequence of stages a prospect passes through, from first awareness to a completed conversion such as becoming a qualified lead or customer.
Key takeaways
- Stages must be mutually exclusive and one-directional, or the counts double-count people.
- Build the funnel backwards from the paying outcome, not forward from traffic.
- More stages localise problems better but multiply the places tracking can break.
- Changing a stage definition reshapes the funnel without any change in real demand.
- Committee and returning buyers break the single-path assumption a funnel depends on.
In depth
A funnel is an ordered set of counts. You define entry criteria for each stage, count how many records satisfy them, and the drop between consecutive counts is the transition rate. The stages must be mutually exclusive and traversed in one direction, or the same person appears twice and the arithmetic breaks. Nothing about the funnel causes behaviour; it is a bookkeeping frame laid over a journey that people take at their own pace, in their own order.
Funnel shape is set by two dials: how many stages you define and how strictly each one is bounded. Few broad stages are easy to maintain but tell you little about where to act. Many narrow stages localise problems precisely and cost more tracking work, and each extra boundary is another place the data can be wrong. Stage definitions also change the shape by themselves: moving the qualification bar from a form fill to a scored threshold thins the lower stages without any change in demand.
Working funnels are built backwards. Start from the outcome that pays, define the stage immediately before it, and continue upward until you reach the first observable touch. Then attach one owner and one intervention per transition. A scorecard funnel is a compact version of this: page view, quiz start, quiz completion, contact submitted, qualifying tier reached. Because each of those is a recorded event rather than an inferred one, the counts reconcile and disagreements about whose number is right largely disappear.
The linear model fails where buying is not linear. Committee purchases involve several people entering at different stages, and a funnel keyed to individuals will count them as unrelated journeys. Buyers also loop backwards, leave for months and return through a different channel, which a strict stage progression records as churn plus a new lead. Self-serve products with no sales contact often have no meaningful middle at all. In those cases a loop or lifecycle view describes the reality better than a funnel.
Example in practice
How to measure it
Report four numbers per stage: entrants, exits to the next stage, the transition rate between them, and the median time a record spends in the stage. The rate says where you lose people; the dwell time says where they stall without leaving. A stage that keeps its rate but doubles its dwell time is degrading, and only the timing column shows it.
Then check the funnel against itself. The product of all transition rates should equal the end-to-end rate computed directly from first touch to closed outcome. When the two disagree, records are entering mid-funnel or leaving without being counted, and the stage arithmetic is unreliable until that is resolved. Run this reconciliation monthly, not once at setup.
Common mistakes
Teams frequently copy the awareness-interest-consideration-action labels straight from a textbook and then cannot say which recorded event marks each boundary. The result is a diagram nobody can query. Fix it by naming, for every stage, the exact event or field value that places a record inside it. If no such event exists, either instrument one or drop the stage; a stage you cannot count is decoration.
The other failure is comparing stage counts taken over different periods, so the top of the funnel is this month's traffic while the bottom is deals started last quarter. The percentages that come out of that are fiction. Use a cohort: follow the people who entered in one window through every later stage, even if that means waiting for the sales cycle to finish before the funnel reads correctly.
Frequently asked questions
What are the typical stages of a conversion funnel?
A common model uses awareness, interest, consideration, and action, though teams adapt the labels to their journey. The key is that each stage is measurable and reflects a real decision the buyer makes.
How is a conversion funnel different from a sales pipeline?
A conversion funnel describes the prospect's journey and the percentages who advance through it, usually from a marketing view. A sales pipeline tracks specific deals and their dollar value from a rep's perspective, so they overlap but answer different questions.
Where does a quiz fit in a conversion funnel?
A quiz typically occupies the interest and consideration stages, engaging visitors while collecting qualifying signals. The scored result and lead-capture form then carry them into the action stage as a qualified lead.
How many stages should a conversion funnel have?
Enough that every stage has a distinct owner and a distinct intervention, and no more. Three to five is the usual working range for a lead-generation funnel. Add a stage only when you can name the event that marks its boundary and the action you would take if that transition worsened. Otherwise you are adding reporting overhead without adding a decision.
What is the difference between a conversion funnel and a customer journey map?
A funnel is quantitative and one-directional: counts per stage and the rates between them. A journey map is qualitative and often loops: it describes what a buyer thinks, feels and does at each point, including detours a funnel cannot represent. Most teams need both, with the map explaining why a transition in the funnel is weak.
Can one person appear in two funnel stages at once?
They should not. If they do, the stage definitions overlap and the transition rates will exceed what actually happened. Make each stage a state a record is in, not a thing it has done, so entering a later state automatically removes it from the earlier one. Then the stage counts sum to the total population.
How do I model a funnel when several people from one company buy together?
Run the funnel at the account level rather than the contact level. Stages then describe the buying group: any contact reaching a stage advances the account, and the account count is what you report. Keep contact-level data underneath for engagement detail, but never mix the two in one chart, because the totals will never reconcile.
Where does a quiz sit inside a conversion funnel?
Usually across the middle. The landing page handles the top, the question flow occupies the consideration span where a prospect is working out whether the problem applies to them, and the scored result plus contact form closes the bottom. Its advantage is that the middle stops being a black box: every question answered is a recorded step.
How often should a funnel be redefined?
Rarely, and never in the middle of a comparison period. Redefining stages resets your history, so treat it like a schema change: pick a cutover date, recompute at least a few prior periods on the new definition if the data allows, and label every chart with which definition it uses. Constant redefinition is the main reason funnel trends become unreadable.