A dashboard says the pricing page has a high exit rate. Two people look at the same chart. One concludes the prices are too high; the other concludes people found what they needed and left to think about it. The data supports both, which means it supports neither.
Counting is not understanding
Analytics is very good at questions of the form how many. How many sessions, how many reached the third step, how many arrived from search.
It is completely silent on why, and it is unreliable on who — ad blockers, tracking prevention and consent choices remove a meaningful and non-random slice of visitors. The people missing from the numbers are systematically different from the people in them.
So the honest use is narrow: use it to find where something changes sharply, then go and find out why somewhere else. A step in a form where most people stop is a fact worth having. The reason they stop will not be in the chart. It is in the form.
Cheaper things that tell you more
Watching one person use the thing, in silence, without helping, is worth more than a month of session data. It is also uncomfortable, which is why it gets skipped.
Reading what customers actually write is the other one. Support messages, quote requests, the questions that arrive by WhatsApp. If four people in a month ask whether you deliver outside the city, that is a page that does not exist, and no funnel report will ever say so.
And the sharpest signal: what people do instead. If the sales team keeps exporting to a spreadsheet, the product has a hole shaped exactly like that spreadsheet.
Measure a decision, not a wall
Before adding tracking to something, it is worth asking which decision the number will change. If there is no answer, the tracking is collecting data on behalf of a future conversation nobody has scheduled.
That is not free. It costs page weight, it costs privacy, and it costs the attention of whoever ends up staring at a dashboard of numbers that were never attached to a question.