Analytics Without Cookies: How to Measure Results When Data Is Shrinking
Analytics is gradually losing sight of what it used to see. Browsers restrict cross-site identifiers, ad blockers cut out part of the scripts, consent mode removes data from people who declined, and mobile platforms limit tracking between apps.
There are two common reactions. The first: ignore it and keep staring at numbers that no longer reconcile. The second: decide analytics is dead. Both are wrong, and here is what to do instead.
What exactly is lost
| What | Consequence |
|---|---|
| Recognising the same user across sites | The classic attribution model breaks |
| Long return windows | Someone returning after a month looks like a new visitor |
| Events from people who declined cookies | Understated conversion counts |
| Data stripped by blockers | A systematic skew towards the less technical audience |
| The link between devices | One client counted twice or three times |
The nature of the loss matters: data does not disappear at random, it shifts. You are seeing a different slice of the audience rather than a smaller one, which is exactly why year-on-year comparisons stop working.
Three approaches that work
Your own data instead of someone else’s
The main change of direction: value moves from tracking other people’s sites to what you know yourself. Registrations, orders, enquiries, purchase history, subscriptions. That data does not go anywhere and does not depend on browser settings.
In practice this means encouraging identification where it is appropriate, and connecting the site to your CRM.
Server-side collection
Events are sent from your server rather than from the browser. This is more resilient to blockers and gives you control over exactly what data leaves your systems. It does not bypass consent: if someone declined, you must not send data regardless of the method.
Modelling and aggregates
When part of the data is unavailable, platforms reconstruct the picture statistically. These are estimates, not exact figures. You can use them, but with understanding: the error grows on small volumes, and modelling does not work on a sample of ten conversions.
What to measure when precision is gone
The most useful shift in thinking is from exact to comparative.
- Trends instead of absolute values. If the collection method has not changed, a thirty percent drop is still a drop, even on an understated base.
- Channel share rather than count. The distribution between sources distorts less than absolute figures.
- CRM data as the anchor. You know deal counts and revenue precisely, regardless of browsers.
- Asking “how did you hear about us”. A crude but honest check on what analytics reports.
- Experiments. Comparing two variants works even on incomplete data, because the skew is identical in both.
- Branded search. Growth in searches for your company name is a signal that owes nothing to cookies.
What does not work
- Working around consent. Technical tricks that let you track people who declined are a legal risk, not a solution.
- Ignoring discrepancies. If analytics shows half as many enquiries as the CRM, that is not “margin of error”, it is broken measurement.
- Swapping one platform for another. Moving to a different tool does not recover data that was never collected.
- Comparing periods across a method change. After reconfiguring collection, old figures are no longer comparable. Record that in writing, or in six months nobody will remember why.
The practical minimum
- Configure consent mode correctly: no consent, no analytics scripts.
- Establish a shared identifier between the site and the CRM.
- Send deal statuses from the CRM back into analytics, not just enquiries.
- Log the date of every change to your collection setup.
- Add a “how did you hear about us” question to the form or the first call.
- Switch reports to shares and trends instead of absolute numbers.
- Reconcile analytics against the CRM quarterly and record the gap.
The last point matters most. A five percent discrepancy is normal; thirty percent is a reason to investigate rather than something to accept.
What comes of this
Paradoxically, restricting data makes analytics more honest. It used to be possible to build elaborate reports on identifiers without ever asking how much money any of it produced. Now precise data survives only where it always mattered most: in your own database, in the CRM, in revenue.
That returns you to the simple question any analytics should start from: can you state the cost of a closed deal per channel. The answer does not depend on cookies. It depends on whether you connected the website to the place where the money is recorded.
