Measurement
Signal Loss Moved the Job to the Page
Consent gates, blocked scripts and shorter cookie lifetimes broke the tracking most funnels were built on. The durable response is not another tag. It is collecting what you need on the page, from people who chose to give it.
Every year the measurement story gets a little worse. Consent banners remove a slice of traffic from analytics. Browsers cap script-set cookies. Ad blockers take another slice. Platforms report conversions your warehouse cannot reconcile. Teams respond by adding another measurement tool, and the picture does not get clearer.
The uncomfortable answer is that a lot of this is not recoverable by better tagging. What is recoverable is the information a visitor is willing to tell you directly, on the page, in exchange for something worth having. That is a design problem, not a tracking problem.
Separating the fixable from the gone
Not all signal loss is the same, and treating it as one problem leads to spending months on the part you cannot fix.
| What you lost | Cause | Recoverable? |
|---|---|---|
| Cross-site view of one person | Third-party cookie restrictions | No, design around it |
| Analytics on non-consenting visitors | Consent requirements | Partly, in aggregate |
| Long attribution windows | Shortened cookie lifetimes | Partly, with server-side |
| Knowing why someone converted | Never tracked it | Yes, ask them |
| Knowing who is a good fit | Never asked | Yes, ask them |
The bottom two rows are the interesting ones. They were never a tracking capability that got taken away. They are information you can collect at any time, and most funnels simply never ask.
Declared beats inferred, and it always did
Behavioural inference was always a proxy. Someone visits a pricing page three times, so you infer intent. Someone matches a lookalike, so you infer fit. Declared information skips the inference: the visitor tells you their company size, their timeline, the problem they are trying to solve.
- It survives every privacy change. Information given directly, with consent, in a form does not depend on a cookie or a third-party script.
- It is more accurate. Inferred company size is a guess. Selected company size is an answer.
- It improves the same session. You can change what the next step says based on what they just told you, which inference from a past session cannot do.
- It feeds bidding a better conversion. Reporting qualified leads rather than raw submissions changes what automated bidding learns to find.
How to ask without killing conversion rate
Five rules 01
Ask about them, not about you
“What are you trying to solve?” earns an answer. “How did you hear about us?” serves your reporting and everyone can tell.
Five rules 02
Front-load the easy questions
Starting with a low-effort choice gets people moving. Contact details belong after the visitor has invested a little.
Five rules 03
Show the payoff as you go
If each answer visibly narrows a recommendation, the next question feels worth answering rather than like an obstacle.
Five rules 04
Branch on the answers
Someone who says they are a two-person team should not then be asked about procurement. Branching is also what makes the exercise feel like a conversation.
Five rules 05
Pass what you learn onward
The answers should reach your CRM as fields and your ad platforms as conversion values, or you have collected data nobody uses.
Treat consent as part of the funnel, not a legal afterthought
The consent banner is usually the first interaction a visitor has with your brand, and it is usually the least considered element on the page. It is worth the same care as a headline.
- Do not let it cover the promise. A banner obscuring the headline costs you the message match you paid for.
- Say what you actually do. Specific, plain language earns more consent than a wall of legal boilerplate.
- Make the page work either way. If your funnel only functions for consenting visitors, you have designed for a shrinking share of traffic.
- Check it on a phone. Banners that are a strip on desktop routinely become a full-screen wall on mobile.
Frequently asked questions
- Q: Is server-side tracking the answer?
- It helps with cookie lifetime and blocked requests, and it is worth doing. It does not create consent you were not given, and it does not tell you why someone converted. Treat it as one repair, not the strategy.
- Q: Will asking questions lower our conversion rate?
- It usually lowers raw submissions and raises qualified ones. Judge it on the outcome further down the funnel, booked meetings or closed revenue, not on form fills, or you will conclude the wrong thing.
- Q: How many questions is too many?
- There is no universal number; there is a ratio. Every question needs to visibly earn its place by improving what the visitor gets. Three questions that shape a recommendation outperform two that do nothing.