What funnel analysis tells you
Funnel analysis follows scanners through each step from exposure to conversion: seen, scanned, page loaded, action started, action completed. It shows exactly where people drop out, so fixes go to the step that leaks most.A funnel with only two steps hides the problem. Break the journey into the real steps a scanner takes.
How to compute funnel analysis
Step conversion = people who reach a step divided by people who reached the previous step, multiplied by 100. Overall conversion = conversions divided by exposures, multiplied by 100.Illustrative example numbers, not real-world data: 5,000 exposures lead to 400 scans, 120 action starts, and 60 completions. Scan-to-start is 120 / 400 x 100 = 30%, and start-to-complete is 60 / 120 x 100 = 50%, so the bigger leak in this example is between scan and start.
How to collect the data
Define the steps before launch, tag each step in analytics, and count the same cohort through the whole funnel rather than mixing different weeks.Keep the raw counts, not just the final percentage. Raw counts let you recompute the metric if you later fix the definition, and they make it possible to spot a denominator problem, such as a week where exposures were estimated instead of counted. Store the counts in one place with the date range, the placements included, and the name of the person who pulled them. That small discipline turns a spreadsheet into an audit trail.
Reading the result
Never judge funnel analysis as a single number. Break it down by placement, device, step, and week. A campaign that looks average overall may be strong in one segment and weak in another, and the fix is completely different in each case. Compare the same placement against its own past, and compare different placements only when their audiences and contexts are genuinely similar. A short trend over four or more comparable periods is worth far more than a single snapshot, because it separates a real shift from weekly noise.Also check the relationship between funnel analysis and the metrics around it. If scans rise while conversions fall, the new scanners may be less interested visitors, or the landing page may have changed. If cost rises while funnel analysis stays flat, you are paying more for the same outcome. Metrics make sense in groups, not in isolation, and the habit of reading two or three together prevents most misreadings.