Comparison with cookie-based tools

Note

This page compares identification models, not vendors. It is written so you can decide whether a cookieless product measures what you need. Where a cookie-based tool is genuinely better, it says so.

The one difference everything follows from

	Cookie-based	tracing.tools
Identifier	Written to the device, persists for months or years	Derived server-side, rotates every UTC day
Where it lives	localStorage or a cookie	Nowhere — it is recomputed per request
Scope	Often cross-site	One site
Survives	Browser restarts, days, months	Until midnight UTC
Consent rule	Storage/access on terminal equipment — usually needs consent	Outside what that rule regulates
Blocked by	ITP, ETP, ad blockers, consent refusal	Ad blockers, if the host is on a list

Everything below is a consequence of that row.

What the numbers mean

Metric	Cookie-based	tracing.tools
Visitors, single day	Distinct devices	Distinct browser-days — close to the same
Visitors, 30 days	Distinct devices over 30 days	Roughly the sum of daily distinct browser-days
Returning visitor	Across months	Within one UTC day only
Bounce rate	Comparable	Comparable
Time on page	Comparable	Comparable
Pageviews, sessions	Comparable	Comparable
Retention past D1	Real	Near zero unless you identify
Multi-day funnel	Works	Does not — “same day” is the widest honest window
Lifetime value per visitor	Works	Does not — revenue is credited to the visit it fired in

Do not compare 30-day visitor totals across the two models

They are different quantities. A cookieless 30-day visitor number is inflated against a cookie-based one by roughly the average number of days a person visits in a month. Compare pageviews, sessions and revenue — those are directly comparable.

What the cookie-based tool measures better

Be honest about this before choosing:

• Multi-day and multi-week retention without asking you to identify anyone.
• Long funnels — a trial started on Monday and converted on Friday.
• Cross-device journeys, where the tool also has a login identifier.
• Lifetime value per acquisition source.
• Last-touch or multi-touch attribution over a long consideration window.

If those are the questions you are paid to answer, a persistent identifier is the right instrument and this is the wrong product — unless you are prepared to call identify() on every authenticated page load, which recovers most of it.

What it measures worse

• Anything a consent banner refuses. The measured population is self-selected by whoever clicked “accept”, and the bias is not random.
• Anything ITP or ETP truncates. A first-party storage identifier capped at seven days, or one day, is not the persistent identifier the reports assume, and the tool rarely says which of your numbers it silently affected.
• Revenue. Revenue here is browser-reported too, with the same limits: a blocked script is a missed sale, and a refund issued later is invisible unless you send a negative amount. Reconcile against your processor for the books. → Revenue events

What this product adds on top

Not everything a cookieless tool does is a compromise:

Revenue by source with no setup	One event with an amount, credited to the visit that made it: source, campaign, landing page, country and device come for free
One currency at a time	Currencies are never added together, and the screen says which ones took money
Sessions with a full timeline	Every hit of a visit, with its properties
An honest retention screen	Identity coverage is reported alongside the grid
A copilot grounded in your own numbers	Never able to invent a metric

Migrating

You can run both for a period — the tags are independent and neither sees the other. Expect:

• Pageviews and sessions to land within a few percent, once bot filtering and exclusions are matched.
• Visitors to differ structurally. See above.
• Traffic sources to differ where the other tool does its own referrer classification into channel groups; this product reports the referrer domain and the UTM parameters as sent.
• Higher totals than a consent-gated tool, because no one is opting out of measurement by refusing a banner.

Choosing

Use a cookieless model when the questions are where did people come from, what did they read, what did they do, and what did that earn — over days and weeks.

Use a persistent identifier when the questions are who came back in March, and what did they spend over a year — and you are willing to pay for it in consent banners, bias and blocked traffic.

Or use this and call identify(), which buys the multi-day questions back for the population you can actually identify, and is honest about how large that population is.

Related

• How visitors are counted
• Cookieless & GDPR
• Retention