The sales funnel: from seeing a product to buying

A thousand people see a product and twenty buy. The overall conversion rate is 2%, and the meeting starts to argue about whether 2% is good. The number cannot answer that question, because it bundles five different losses. Some of the thousand never clicked. Some clicked and left without adding anything. Some added to cart and abandoned. Some ordered and failed to pay. Some paid once and never came back. The sales funnel is the accounting of those losses: it turns one disappointing ratio into a chain of counters, each with its own denominator and its own possible fix.

§3.4.1 describes the funnel as marketing arranged in time order or stages - awareness, interest, consideration, intent, evaluation and purchase - wide at the top and narrowing as prospects are disqualified or decide not to buy. The textbook stresses that marketers must stay in contact at each stage and keep leads engaged, with the goal of moving them to the bottom. Two details in that definition matter. First, the funnel focuses on steps taken by marketers or sellers, not on the buyer's inner state. Second, it represents the quantity and conversion of prospects through pipeline stages, which is why it can be counted.

For measurement, use six counters: exposure (impressions), click (visit), add-to-cart or inquiry, order, payment and repurchase. Define each by a specific event in the seller's analytics, not by a feeling. Order and payment must be separate, because failed payments, cash on delivery, fraud review and buyer remorse lose people after the order is created. Repurchase must be separate, because it measures retention rather than acquisition and belongs after the purchase stage. The first five counters roughly correspond to the textbook's stages, but they are not identical: the textbook stages are marketing stages, while these are events that a dashboard can count.

Run the arithmetic on a clean example. 10,000 impressions produce 500 clicks (5%); 500 clicks produce 100 carts (20%); 100 carts produce 40 orders (40%); 40 orders produce 30 payments (75%); 30 paid buyers produce 6 repurchases (20%). The overall conversion from impression to payment is 30/10,000 = 0.3%, or 0.05 × 0.20 × 0.40 × 0.75. Add repurchase and the end-to-end ratio is 6/10,000 = 0.06%. Each layer also has a drop-off: 95%, 80%, 60%, 25% and 80%. The arithmetic is simple; the discipline is keeping the denominator honest and not comparing a stage rate with an overall rate.

Diagnosis starts with two different questions. Which layer loses the most people in absolute terms? Which layer has the worst rate relative to its own baseline? In the example the top loses 9,500 people, but a 5% click rate on cold display traffic may be normal. The cart-to-order rate of 40% and the payment rate of 75% are more likely to be actionable. Then value a percentage point in money. At these rates, +1pp of click rate adds 100 clicks and about six paid orders, while +1pp of payment rate adds only 0.4 orders. The top of the funnel moves more volume but usually costs more to move, so the right comparison is cost per incremental order, not the size of the percentage.

Forecasting uses the same chain in reverse. Expected paid orders equal traffic × click rate × cart rate × order rate × payment rate. At the example's rates, 10,000 impressions yield 30 paid orders; to reach 45 at unchanged rates, the seller needs 15,000 impressions. Alternatively, a 10% relative improvement at each of the four conversion steps multiplies paid orders by 1.1 to the fourth power, about 1.46, which turns 30 into 44. Compounding explains why small improvements across several layers often beat one heroic fix, and why a forecast must state which rates are assumed stable.

§3.4.1 gives the funnel report two uses: forecast sales from current lead volume, and identify where deals get stuck so the process can be improved and the team coached. The order matters. Diagnosis should come before scaling, because if a downstream layer leaks badly, buying more traffic only multiplies the loss. A report that shows a beautiful top and a broken checkout is not a growth story; it is an instruction to stop spending until the leak is fixed.

The funnel is a model, not a law. Buyers loop: they click, leave, compare on another device, read reviews, return through search, and sometimes buy weeks later. Some enter mid-funnel from a shared link or a creator's recommendation. Repurchase bypasses the top entirely. B2B cycles are longer and involve several people. Returns can reverse a payment after the report closes. For all these reasons, the funnel is best used as a shared scoreboard with fixed definitions, not as a claim that every buyer walks down six steps in order.

Keep the funnel separate from Lesson 10's five stages. The five stages - need awareness, search, evaluation, purchase and post-purchase - describe buyer psychology and behaviour. The funnel's six counters describe seller-side events and their conversion. The five stages help you decide what message belongs at each moment; the funnel helps you see how many people were lost at each hand-off. A click count cannot tell you that a buyer is 'in consideration'; it can only tell you that a visit happened. Using one tool as if it were the other produces false precision.

Finally, protect the measurement from its own common failures. Fix event definitions and attribution windows before comparing months. Segment by channel, device and new versus returning buyers, because an aggregate rate can rise while every segment falls if the traffic mix changes. Check denominators: a 50% rate on four people is noise. Treat a paid order as revenue only after the return window. And remember what the funnel does not measure: logistics cost, commission, discount and margin. A funnel can look healthy while the business loses money. The next lesson takes the last counter, repurchase, and asks what makes a buyer come back.