Turning millions of sessions into proactive action
How a leading Indian private-sector bank uses Computer to watch every end-user journey — and get alerted the moment one breaks.
ICICI Prudential
A top private-sector bank, running at national scale
Tens of millions of customers transact across the bank's mobile and web apps every day — UPI payments, transfers, onboarding, servicing. Every tap is a moment the experience has to hold.
At this scale, a broken journey hides in the noise
Session data explains what happened — but only after a customer complained. The team needed to know the instant a failure pattern began, not hours later.
Reactive, not proactive
Issues surfaced through support tickets — long after the failure first appeared in the data.
Volume drowns the signal
A real but small failure cluster is invisible among millions of daily sessions with nothing watching for it.
Minutes matter
For UPI and payments, a spike untriaged for even minutes turns into thousands of failed transactions.
Every journey captured, searchable and summarised by AI
Computer records real end-user sessions across web and mobile and turns the raw stream into something a team can search, replay and reason over — with sensitive fields masked for compliance.
Session replay & journeys
Reconstruct exactly what a customer did — views, clicks, funnels — with text and labels masked.
Rich in-session signals
Custom events, API calls, network logs, round-trip time, page names and clicks — web and mobile alike.
AI summaries at scale
Agents summarise thousands of sessions so teams reason over patterns instead of scrubbing recordings.
Ask about your sessions — get a live answer, not a report
Drill into a single journey to the exact failing step
A monitor watches. An alert fires. A workflow acts.
A monitor holds a query over session events plus a threshold. When it's crossed, an alert fires — and drives a workflow that notifies the right team and opens the right issue automatically.
A lifecycle that ignores noise and confirms real problems
An alert only escalates when a condition holds. Pending is an early warning; active means it's firing; recovering is the buffer before it goes quiet — so teams act on signal, not flicker.
No query language — just ask Computer to watch it
The alert fires — and lands right in your chat
Map the user's failed API to the backend service behind it
Dynatrace via AirSync| Front-end call (session) | Back-end service (Dynatrace) | Error rate | p95 latency | |
|---|---|---|---|---|
| POST /upi/collect | → | payment-collect-service | 18.4% | 4.8s |
| GET /upi/status | → | payment-status-service | 0.2% | 210ms |
| POST /auth/login | → | identity-service | 0.4% | 180ms |
DynatraceOne memory: session data + Dynatrace, continuously synced
Front-end session data streams in directly from the PluG SDK embedded in your apps. Back-end Dynatrace APM is pulled in by an AirSync connector on a schedule. Both land in one memory — so they can be correlated in a single question.
Session data comes straight from the PluG SDK; only Dynatrace rides the connector. Both sit in one permission-aware memory — one question spans the whole path, no swivel-chairing.
The failures caught the moment they start
Each of these is a live monitor — a threshold over real session events — mapped to the journeys that matter most for a digital bank.
UPI / payment errors
Alert when "incorrect handle" and payment failures cluster above normal across payment journeys.
Network failures
Trigger on 500 / 502 responses in network logs — catch backend degradation before it spreads.
Journey drop-offs
Alert on errors and abandonment in a specific app layer so a broken step is flagged, not guessed.
Very slow APIs
Use round-trip time on web and mobile to alert when a critical API crosses a latency threshold.
A live dashboard — from a firing alert to the exact sessions
Illustrative view. One click on any alert opens the filtered session list — the full blast radius, every session where it occurred.
From "a customer told us" to "we already knew"
Alerts on top of session data change the operating model: teams see failure patterns as they form, triage from one dashboard, and reach the exact affected customers in a click.
Detect in minutes
Failure clusters trigger the moment they cross threshold — not when a ticket lands hours later.
Triage from one place
A live dashboard shows active alerts, trends and blast radius — no jumping between tools.
Act automatically
Each alert routes itself — a notification in Computer, an issue with logs to check, and the sessions to review.
How this maps to ICICI Prudential
The same monitors, translated to the moments that matter in life insurance — where a broken step means a lost policy or a stalled claim.
Buy & onboard
Alert on drop-offs in proposal, KYC and e-signature so abandoned policy journeys surface instantly.
Premium payments
Watch payment-gateway and renewal failures — catch declined premiums before a policy lapses.
Claims & servicing
Trigger on errors in claim intake, document upload and status checks — the highest-emotion moments.
Portal & app speed
Use round-trip time to alert on slow fund-value, statement and login APIs across web and mobile.
Each becomes a monitor on your session data — no rip-and-replace, built on the observability layer you already run.
Watch every journey.
Act before customers feel it.
Sessions, alerts and a live dashboard — the same proactive observability, mapped to your policy, servicing and payment journeys.
first monitors