Analysis
⚠ Sample Data

Auth Response Time
Distribution Analysis

Statistical investigation of payment auth latency — comparing benchmark distribution against targeted-transaction observations to determine whether elevated response times indicate a system anomaly.

Period Jan – Mar 2026 (sample)
Dataset 1,847,392 tx (benchmark) · 1,204 tx (targeted)
Question Are 48s and 83s observations anomalous?

Key Observations

Three targeted transactions under investigation — auth response times compared against the platform-wide benchmark median (14s).

OBS-A Auth
48s
Card auth · user interaction
OBS-B Auth
83s
Card auth · user interaction
OBS-C Auth
62s
Wallet · single-flow
Benchmark Median
14s
1,847,392 tx baseline

Section 1 — Auth Time Distribution

Transaction share by time bucket (0–90s). Vertical markers show where observed values fall.

Benchmark
Targeted Txns
48s (OBS-A)
62s (OBS-C)
83s (OBS-B)
Both distributions are right-skewed with a long tail beyond 40s. The targeted txns tail (60s+) accounts for ~10.6% of transactions vs the benchmark's ~1.9% — a 5.6× elevation that reflects the targeted group's higher overall median (19s vs 14s), not a system-level regression.

Section 2 — 60s+ Tail Rate by Payment Method

Percentage of transactions exceeding 60s — benchmark (bar, left axis) vs targeted txns (line, right axis).

The elevated tail rate is not isolated to a single payment method — it appears consistently across card, wallet, and bank transfer. This cross-method pattern points to a client-side environmental factor rather than a method-specific regression.

Section 3 — Log-Normal Density Fit

Fitted density curves for both groups. Vertical markers show where the three flagged observations fall.

Benchmark distribution
Targeted Txns distribution
48s (OBS-A)
62s (OBS-C)
83s (OBS-B)
The chart plots fitted log-normal density curves for both groups. The benchmark curve (μ=ln 14, σ=0.70) peaks earlier and declines steeply, reflecting a distribution concentrated around 14s. The targeted txns curve (μ=ln 19, σ=0.92) peaks later with a wider spread and longer right tail. Vertical markers show where OBS-A (48s), OBS-C (62s), and OBS-B (83s) fall — placing them at the ~84th, ~89th, and ~95th percentile of the targeted txns distribution respectively.

Section 4 — Observation Detail

Auth time breakdown by stage — user interaction vs system processing per observation.

User interaction (auth stage)
System processing
Benchmark p50 (14s)
API call completion
The auth stage (user interaction) accounts for the majority of total time across all observations. OBS-B and OBS-C show system processing under 500ms; OBS-A's API stage is ~8s (~14% of total). All auth-stage time is user interaction and outside server control.

Synthesis

What each chart section found — and what it means for the investigation.

Section 2 · Tail Rate by Method
Pattern is cross-method, not isolated
60s+ elevation appears across card, wallet, and bank transfer. A method-specific regression would show one method spiking. The uniform cross-method pattern points to a shared client-side environmental factor.
Section 3 · Density Fit
Observations within expected range
OBS-A (48s), OBS-C (62s), and OBS-B (83s) fall at the 84th, 89th, and 95th percentile of the targeted txns distribution. They are tail values — not anomalies outside the fitted model.
Section 4 · Stage Breakdown
Not a server-side issue
OBS-B and OBS-C show system processing under 500ms. OBS-A's API stage is ~8s (~14% of total). All auth-stage time is user interaction, which is outside server control.