Cookie-free analytics solves one leak: a rejected consent banner takes nothing from a script that writes nothing to the device. It does not solve the other one. An ad blocker can still stop a browser script from loading, and when it does, that visit simply does not exist in your numbers. This post explains what blockers stop, why the visitors you lose are not a random sample, and how to measure from the edge or your server so a page load is counted whether a script runs or not.
What does an ad blocker actually stop?
Most blockers work from filter lists: patterns of script URLs, domains and request paths known to belong to ads and analytics. When a page asks for a matching file, the browser extension cancels the request before it leaves the machine. The script never loads, so nothing it would have sent is ever sent. There is no error on your side and no gap in the logs you look at. The visit just is not there.
This applies to Vero’s browser script as much as to any other. Being small and cookie-free does not put a script on an allow list. If someone tells you their analytics script cannot be blocked, ask where it runs. Anything that runs in the visitor’s browser can be stopped by the visitor’s browser.
Why are the visitors you lose not a random sample?
If blockers removed the same share of every audience, you could ignore them: your totals would be low but your comparisons would hold. Nothing says they do. Blocker use varies by audience, and if your readers are developers, people in IT or Hacker News regulars, it is safer to assume more of them run one, not fewer. A launch on a developer forum can then look smaller in a browser-only report than it was.
That skew is what makes the loss expensive. It does not only lower a total, it changes the mix: the source that brings the most technical visitors looks weaker than it is next to the source that brings the least. If you sell to developers, browser-only analytics can quietly tell you to spend less on the channels that know you best.
Three places to measure a visit
The browser script is the richest. It sees what only a browser can see: how far a reader scrolled, how long a page was actively in front of them, Web Vitals on their device, route changes inside a single-page app, clicks on outbound links and downloads. It is also the only one a blocker can stop.
Edge middleware runs on your hosting platform before the page is served, on Vercel and Next.js, Cloudflare or Netlify. It sees every request for a page, including from visitors whose browser never runs a script, and it can tell Vero about each one. A server-side call does the same from your backend, on Node, Bun or Deno, or any language that can send a POST. Neither has anything in the browser to block.
The edge and server section of the docs has the code for each. For pageviews, pick one per site. The browser script sends a pageview of its own on every load, so running the edge snippet and the script on the same pages counts each load twice. Choose the edge or the server where blocked visits matter most, and the script where you need the signals only a browser has. Money takes its own path either way, as the last section explains.
Edge tracking in practice
The middleware does three things for each page request. It checks that the request is a GET for an HTML page, so assets and API calls are left alone. It sends a pageview to Vero’s collect endpoint with your site token, the path and the referrer header. And it forwards the visitor’s IP, user agent and country in three headers, x-vero-ip, x-vero-ua and x-vero-country, so the daily visitor hash is computed from the visitor and not from your server.
Two values come from your dashboard: the site token, from the install snippet, and the ingest secret, under Connect your sales. Keep the secret on the server, in an environment variable such as VERO_INGEST_SECRET. It is what tells Vero that forwarded headers can be trusted. A public token alone could not do that, or anyone could invent visitors. The IP is hashed in memory and discarded, exactly as it is for the browser script, and known bot user agents are dropped before anything is stored.
One detail the minimal snippet leaves to you: campaign tags. The middleware sends the path, not the query string, so copy utm_source, utm_medium and utm_campaign from the request URL into the body if you want tagged links to keep their campaign. The collect endpoint accepts all five utm fields.
What you give up at the edge
Edge pageviews are counted from requests, so they know nothing that happens after the HTML is sent. No scroll depth, no active time, no Web Vitals, no outbound clicks. A single-page app that changes route without a new request records the first page only. That is why edge tracking is a floor for traffic and the script is the instrument for behaviour.
That is the trade-off to weigh before you switch. If you stay with the script, the tracker builder shows the exact size of the combination you pick. If numbers ever look wrong after a change like this, my data is not showing up lists the usual causes in order.
Money never depended on the browser
Revenue takes a different path. When someone pays through Stripe, Lemon Squeezy or Paddle, the provider sends a signed webhook from its servers to Vero. No blocker sits between them, so every sale is recorded whatever the visitor’s browser does. A checkout you build yourself can post each sale to the ingest endpoint from your server, with the same result.
What a blocked script can cost is the link between that sale and its visit. The checkout module adds the visit to Stripe Payment Links and Lemon Squeezy checkout links; if the script never loaded, the link goes out without it and the sale arrives without a source. Vero then gives it the source of that customer’s first attributed payment, if there is one, so a renewal keeps its channel. A brand new customer whose browser blocked everything is recorded as revenue with no source: counted, never invented. How to attribute revenue without cookies explains the rest of that chain.
Unsure of a term? Read the glossary

Vero is built by one person in Spain. I read every waitlist entry and pick the beta accounts myself.