Skip to main content
POST
Validate churn signals API

Frequently asked questions

The confidence score is statistical — it grades a churn from detection history. This endpoint is definitive: it loads the site again right now with a browser and consent-banner clicks, so it sees more than the bulk crawl did. Use it on the handful of churns you are about to act on, not on the whole feed.
confirmed — the technology is absent on a healthy re-detection, so the removal is genuine. false_positive — it is still there, so the churn was a prior crawl miss. masked — a header-detected origin web server is now behind a CDN or proxy, which is a detection change rather than a removal. inconclusive — the site was unreachable, blocked, or gave too little signal.
Roughly 2 credits per domain in fetch mode and 5 in browser mode, charged after completion. An inconclusive result is never charged, because that is our side failing rather than yours. Verdicts are cached for 48 hours per domain and technology pair, so a repeat call returns the stored verdict with cached: true at no charge and with no re-crawl. Inconclusive results are never cached, so a retry genuinely re-runs.
Up to 10 in auto or browser mode, and up to 100 in fetch mode. Fetch parallelises at roughly 1–3 seconds per item; browser takes 30–60 seconds each, which is why its cap is small enough to fit a synchronous response. Split larger browser batches on your side.
It routes technologies with a reliable detection method to the fast fetch path and sends everything else straight to a browser. If a fetch result comes back weak it escalates to a browser once. That keeps the average cost near the fetch price while still settling the hard cases.

Authorizations

Authorization
string
header
required

API key in format tapi_live_[32-char] (live) or tapi_test_[32-char] (test)

Body

application/json
items
object[]
required

The churn candidates to re-check. Maximum 10 items in auto or browser mode, 100 in fetch mode.

Minimum array length: 1
mode
enum<string>
default:auto

auto routes high-reliability technologies to a fast fetch and everything else straight to a browser, escalating weak fetch results once. fetch parallelises at roughly 1–3 seconds per item. browser takes 30–60 seconds per item.

Available options:
auto,
fetch,
browser
interact
boolean
default:true

Click consent banners so consent-gated tags fire.

Response

A verdict per submitted item

success
boolean
Example:

true

data
object