Skip to content

Behaviour & predictive

Intelligence · forecasts
RESTRICTED // INTERNAL--:--:-- PKT
13proposals awaiting decision
Live
financial fraud report — 2026-10-30Karachi · openSIM-swap wave reported across three operatorsGujranwala · openRansomware detonation on logistics sector hostMardan · openNCCIA-C-2026-100680 financial fraudKarachiNCCIA-C-2026-100512 financial fraudKarachiPhishing kit hosted on domestic VPSPeshawar · openCredential-stuffing burst against e-commerceIslamabad · open
illustrative data
Intelligence · forecasts

Getting ahead, without pretending to be certain

A forecast is useful when its uncertainty is on the same screen as its central estimate. Every band below is published with the model, not after it.

7-day outlookBacktest MAPE 5.1%IL

Volume forecast

14 days forward · 7 days back
expectedactual80% band
10113015918921810-2310-2610-2911-0111-0411-0711-10
IL Heuristic seasonal model on the local dataset. Beyond 14 days the error rises to ~16%.

Early warnings

threshold breaches

Forecast shape

next 14 days
154
10-31
167
11-01
166
11-02
164
11-03
177
11-04
172
11-05
157
11-06
164
11-07
148
11-08
149
11-09
140
11-10
131
11-11
123
11-12
Mean expected / day155
Peak day2026-11-04 · 177

When cybercrime actually arrives

weekday × hour, all intake
Peak band 20:00–00:00
0003060912151821SunMonTueWedThuFriSat

Demand is not uniform. Staffing the duty desk to this shape costs less than staffing it flat and answers more calls inside the golden hour.

Channel mix

where reports come from
101
App
100
H1991
98
Web
92
Walk-in
90
Partner referral

The app is deliberately the widest lane: it removes the walk-in journey that suppresses reporting outside the eight largest cities.

Loss cohorts

reported value
Under Rs 50k 193Rs 1M and above 100Rs 200k–1M 99Rs 50k–200k 71No loss reported 18
Reports with no loss stated4%

Language of report

Urdu vs English
English60%
Urdu (اردو)40%

The interface is bilingual by design, not by translation: Urdu carries its own type scale and line height so Nastaliq descenders never collide with the row above.

District outlook

ranked by volume
DistrictProvinceReportsHigh / criticalValue reportedTrend · 9d
LahorePunjab742437,807,065
KarachiSindh702334,812,810
IslamabadIslamabad Capital Territory431921,813,495
RawalpindiPunjab28813,539,410
FaisalabadPunjab28912,726,575
MultanPunjab281110,146,891
PeshawarKhyber Pakhtunkhwa25119,689,097
HyderabadSindh18710,589,726
BahawalpurPunjab1664,620,344
GujranwalaPunjab1458,639,786

Guardrails on prediction

what this page refuses to do
No individual risk scores. The model forecasts volume and typology, never a person. There is no score attached to any human being anywhere in this platform.
Bands, always. A central estimate without its band is a misleading number. Every forecast here carries one, and the band widens with horizon.
Advisory, not enforcement. A warning can change a public advisory. It cannot suspend an account, block a SIM, or trigger enforcement on speech. Those need a human, a warrant and a recorded basis.