What a leak really is
A funnel loses people at every stage, and some loss is expected. Not everyone who visits a site is a potential customer, and not everyone who joins an email list will buy. A “leak” is a stage where people leave at a noticeably higher rate than you would expect, or higher than you are willing to accept, given what you know about your business.
The goal of looking for leaks is not to reach 100 percent at every step, which is impossible. It is to find the one or two places where a fix is most likely to matter, so your effort goes somewhere sensible.
Start with a stage-by-stage table
You need counts for each stage of your marketing funnel over the same period. Choose stages that you can actually measure. For an online service business, they might be:
- visitors to the website
- visitors who viewed the services page
- people who started the enquiry form
- people who sent the form
- people who became customers
Gather the counts from your analytics tools, form software and sales records. Then calculate, for each step, the share of people who made it from the previous stage.
A worked illustration
These figures are invented for illustration, not benchmarks. Suppose one month produced these counts:
| Stage | People | Share of previous stage |
|---|---|---|
| Visited the website | 2,000 | not applicable |
| Viewed the services page | 600 | 30 percent |
| Started the enquiry form | 120 | 20 percent |
| Sent the form | 60 | 50 percent |
| Became customers | 12 | 20 percent |
Checking the arithmetic: 600 out of 2,000 is 30 percent; 120 out of 600 is 20 percent; 60 out of 120 is 50 percent; 12 out of 60 is 20 percent. The overall rate from visitor to customer is 12 out of 2,000, which is 0.6 percent.
What stands out? Half of the people who start the form do not send it. That is 60 people who showed real interest and then stopped. It might be a form that asks too much, a technical fault on phones, or a question the page did not answer. Because these people were close to acting, improving this step may be worth more than trying to attract more visitors at the top. That is only a hypothesis, though; the numbers show where, not why.
Compare, but sensibly
A single percentage is hard to judge in isolation. Useful comparisons include:
- Against your own past: is this stage better or worse than last month or last quarter?
- Against other stages: a step where people leave far more than at neighbouring steps deserves a look.
- Between segments: break the numbers down by traffic source, device or customer type. Visitors from a search result and visitors from a social post often behave differently. Customer segmentation can help you decide which splits are meaningful.
Be careful with external comparisons. Published “average” rates vary by industry, price, audience and how a stage is defined, and they are rarely comparable to your own measurement. Treat any outside figure as a rough conversation starter at most.
Beware small numbers
If a stage has 15 people in it, a change of three people can swing the percentage dramatically. Small counts bounce around from week to week by chance alone. Before concluding that something is broken or fixed:
- use a longer period, so more people are counted
- check whether a one-off event, such as a campaign or a holiday, distorted the month
- wait for a larger sample size before treating a difference as real
The same caution applies to any experiment you run later. The idea behind statistical significance is that some differences are just noise, and small funnels produce a lot of noise.
Find the reason behind the number
Numbers locate the leak; other evidence explains it. Some sources to combine with the figures:
- Direct feedback: ask recent customers what almost stopped them, and ask people who did not buy, where it is appropriate and polite to do so. Even a one-question survey can be revealing.
- Session observation: if you use tools that show recorded sessions or heatmaps, remember that they involve collecting data about visitors. Check the consent and privacy rules that apply to you, since they vary by country and change; this is not legal advice.
- Your own walk-through: go through the stage on a phone and a desktop as a first-time visitor.
- Conversations: sales calls, emails and chat transcripts often reveal the confusion that the figures cannot.
- Technical checks: slow pages, broken forms and tracking mistakes can all look like human drop-off. Test that your measurement works before trusting it.
Common causes by stage
The causes below are possibilities to investigate, not diagnoses.
- Top of funnel: the wrong audience is arriving, or the message does not match the source that brought them. Awareness stage content may attract readers who are not buyers.
- Middle of funnel: people cannot compare options easily or find the evidence they need at the consideration stage.
- Bottom of funnel: cost surprises, long forms, unclear next steps or missing reassurance at the decision stage.
- After purchase: poor onboarding or follow-up, which shows up as low repeat purchases rather than a visible drop.
Choose one fix at a time
When a leak is identified, resist the urge to change everything at once. Pick one cause, make a single change, and note the date. That way, if the numbers move, you have a clearer idea of why. Where you have enough traffic, A/B testing is a more careful way to compare a change with the original. Keeping a record of what you changed and what happened, such as in a testing log, protects you from repeating mistakes.
Keep expectations realistic
Fixing a leak will not guarantee more sales; results depend on the market, the offer and many things outside your control. Funnels also shift over time, so the biggest leak this quarter may not be the biggest next quarter. A regular, simple review, perhaps monthly, with the same table each time, tends to be more useful than an occasional deep investigation.