skim — install intelligence

Click-to-install time

Verdicts

Installs per hour

Traffic sources

source	clicks	installs	conv	avg ctit	avg score

Cohort retention (D1 / D3 / D7)

date	installs	D1	D3	D7

time	install	device	ip	source	attr	ctit	score	verdict	label

time	click_id	source	campaign	gaid	ip	label

Source economics — the money trail (flags: conv_inversion / no_retention / no_value)

source	clicks	installs	conv	D1 retained	purchases	verified rev	flags

Rule hits

Flagged device fingerprints

fingerprint	installs	gaids	avg score

Hottest IPs

ip	installs	gaids	avg score

Postback log

time	install	status	url / error

Attack simulation

Generates labeled traffic (genuine + attack classes) through the live collector, then measures how many of each class the fraud engine catches.

users 300

Previous runs

run	time	users	recall	precision	fpr

Catch rate by attack class

class	n	clean	review	reject	caught %

Rule × class matrix

App & attribution

collector base store redirect URL (for /c/click) partner postback URL (macros: {click_id} {install_id} {gaid} {source} {score} {verdict} {ctit}) postback secret (HMAC) app package (integrity token binding) sim proxy URL (http://user:pass@host:port or socks5://…) — sim fires through it against the public endpoint

require Play Integrity token

payout requires valid integrity

payout requires STRONG integrity

sim lab enabled

SDK keys

SDK quickstart

POST {collector}/c/install
{
  "api_key": "sk_live_…",
  "app_id": "com.partner.app",
  "install_id": "in_…",
  "gaid": "…",
  "device": { "model": "…", "fingerprint": "…", … },
  "referrer": { "present": true, "click_ts": …, "install_ts": … }
}
→ { "ok": true, "verdict": "clean", "score": 10 }