How we count clicks

Our incentive, stated first

Inflated clicks would make us money. Every counted click burns a listing’s balance, and a drained balance has to be topped up. We therefore have a direct short-term interest in counting clicks that are not real. You should not take our word that we don’t.

What we offer instead of our word: filtered clicks burn nothing, burn is capped per hour, every rejection is logged with its reason and exportable by the listing owner, and you can turn on ?ref tagging to reconcile our numbers against your own analytics. If our count and yours disagree, you will be able to prove it.

What counts

A click is counted only if all of the following hold.

One visitor per listing per 24 hours. A visitor is an HMAC of IP address and user agent under a salt that rotates daily. We keep two salt generations so the window does not reset at midnight.
Not a known bot. Crawler user agents, headless-browser signatures, link-preview fetchers, and monitoring tools.
Not the owner. A click from the listing owner’s signed-in session does not count.
Not proxied. Requests arriving through a detectable proxy are recorded and excluded.

A counted click past the hourly burn cap — 20% of the balance — still appears in the public counter but costs nothing. We separate “we don’t believe this” from “we won’t charge for this” on purpose; conflating them is what makes other boards’ numbers unreadable.

What we cannot currently detect

Datacenter traffic filtering is not implemented. Properly telling a cloud server from a home connection needs an ASN dataset we have not wired up. Traffic we cannot classify is currently counted, because rejecting everything unverified would throw away most legitimate visits. Until this ships, our counts include automated traffic that does not identify itself as automated.

We would rather tell you this than let you infer it from a number that looks too good.

Also worth knowing:

IP and user agent are only ever stored as a salted hash, so we cannot tell you who clicked — only how many distinct visitors there were.
A shared IP behind one office NAT looks like one visitor. Ten colleagues clicking count once.
Raw click records are deleted after 90 days. Daily totals are kept indefinitely.

Board-wide totals

No aggregated days yet — totals appear once the first rollup runs.

Published so a sudden change in our filter rate is visible to you at the same time it is visible to us.

Checking us

Sign in and open your listings. Every listing has a CSV export of its full click log — timestamp, whether it counted, the rejection reason if it didn’t, what it burned, and how the visitor arrived. Filtered rows are included.

You can also enable ?ref tagging, which appends ?ref=meltboard&cid=… to your outbound link so each click is identifiable in your own analytics. It is off by default because some people want clean URLs — but it is the strongest check available, and we would rather you used it.

Changelog

1.0 — 2026-08-26 — First published. Datacenter filtering is not yet implemented and is documented as such.