Advertisers
Target what the page says, not what we guessed about the reader
Contextual targeting over the full IAB taxonomy, brand-safety controls no bid can override, billing on viewable impressions only, and server-to-server conversion tracking. This page explains how each one works, so you can check it rather than take our word for it.
The context classifier
Our own URL-to-category classifier over the IAB Content Taxonomy 2.2. The signal is the page itself, not an accumulated profile of whoever is reading it.
When the tag requests an ad it sends the page URL, its title, the meta description and the first few headings. That classifies the page into the IAB tree, and the result is cached. Which ad to serve is decided afterwards, in milliseconds.
The practical consequence is that you target the article someone is reading rather than an accumulated profile of that person. It is both a better signal for a brand and less personal data for anyone to hold.
IAB Taxonomy 2.2
The full tree, not a short list of interest buckets. The category assigned to the page is the one your targeting is matched against.
Geography checked twice
An in-process GeoLite2 database, cross-checked against the CDN's own headers. Disagreement between the two sources is itself a proxy signal, and it is measured.
Decided in microseconds
Ad selection adds no perceptible delay to the page. It runs entirely in memory, well inside a five-millisecond budget, so a slot fills before a reader notices it was empty.
Brand-safety controls
Four independent dimensions, not a switch. What the page is, what the ad depicts, what you tolerate, and what the publisher accepts.
Content is rated on a five-value scale and tagged with a fixed set of content flags. You declare the highest rating you accept and which flags you block; those settings are applied in the eligibility check, which knows neither your bid nor your budget.
| Value | Name | Meaning |
|---|---|---|
| 0 | general | Safe for anyone. It is the default everywhere. |
| 1 | teen | Mild themes. It is the ceiling on what machine classification can assign on its own. |
| 2 | mature | Adult themes, artistic nudity, strong language. |
| 3 | adult | Explicit content. Requires an attested age gate on the site. |
| 4 | extreme never served | Illegal content, non-consensual material, or anything involving minors. It is a refuse-and-report state, never monetizable. |
What you control
- The highest page rating your campaign may appear beside.
- Which content flags you block, both taste and regulated.
- The inventory tier: tier A placements only, or tier B as well.
- Your own per-source blocklist, to exclude one specific site without changing any other setting.
The viewable impression is billed, not the served one
“Served” and “viewable” are two different events here, with two different names, and only one of them is billed.
Under CPM the viewable impression is what is billed, to the MRC standard: at least 50% of the ad's pixels in view for one continuous second, in a foreground tab, on a page that has had a real user gesture. The viewability beacon produces it, and only once per token.
The ad request records an ad_served event. The beacon records an impression event. You can see both, and the ratio between them, in your report.
- The event is an impression, a click or a conversion. No-fill and postback rows are measurement, never money.
- If it is an impression, it was viewable to the MRC standard. An impression nobody could see is not an impression anyone pays for.
- If it is a click, a real user gesture produced it. A synthesised DOM click is the cheapest fraud there is.
- It did not come from a datacenter address. A datacenter address is not a person.
The invalid-traffic rate, disclosed
We do not promise it is zero. We tell you what it was and take it off.
Invalid traffic is excluded when the invoice is computed, by a rule that can be run again over the same events. That is what makes a correction possible at all: money that cannot be recomputed cannot be corrected.
Your dashboard shows the invalid-traffic rate beside the period's spend, computed the same way the invoice was, so the disclosed rate and the billed amount cannot disagree. When you were charged for something you should not have been, the correction appears as its own line — invalid traffic credit, with the period — rather than as a total that quietly came out lower.
Postback attribution
Server to server, signed, with no browser pixel and no dependence on third-party cookies.
We append a click identifier to your landing URL. When the conversion happens, your backend calls us with that identifier and an HMAC-SHA256 signature over the parameters. Every call is stored with its outcome, accepted or rejected, and you can read it in the dashboard: postbacks are not dropped silently.
Default window
168 hours
Configurable between 1 and 720 hours.
Correlation ceiling
30 days
A click stays correlatable for this long, and no window may exceed it.
Clock tolerance
5 minutes
How far the timestamp parameter may sit from our clock. It is the replay half of signature verification.
Regulated categories
We publish where every regulated category stands so you find out before writing to us rather than after a rejected campaign.
| Category | Status | Reason |
|---|---|---|
| Mainstream dating | Accepted | Open, with mandatory human review of every creative against the deceptive-pattern list. |
| Adult dating | Accepted | The label is accepted, the content is not: the category exists so it can be declared, and the creative rating cap excludes it from service. |
| Gambling and iGaming | Deferred | A licensed activity where the intermediary carries obligations of its own, different in every country. Switched off until a legal review exists that would let us price the risk. |
| Crypto assets and Web3 | Deferred | Consumer financial promotion carries criminal exposure in several jurisdictions. One fraudulent ad would end the brand-safe positioning that is the entire thesis. |
| Supplements with health claims | Deferred | Deferred until a claims-substantiation process exists. Today it is covered by the pharmaceutical flag. |
| Prescription pharmaceuticals | Prohibited | No toggle at all. It is prohibited in nearly every country, and it is refused in code, in a database trigger, and in a schema constraint. |
For advertisers
Target on the real content of the page, pay only for viewable impressions, and read the invalid-traffic rate beside your spend.
For publishers
One tag, control over which ads appear on your pages, and per-placement and per-context detail in the report.