APM for Insurance
Insurance companies have complex mixtures of legacy packaged and custom applications. Insurers migrating to packaged applications deal with high amount of change and uncertainty.
Complex core systems which are most often packaged applications are being coupled with other custom built software to build capabilities to do straight-through processing. Performance, usability and availability are key to these channels being productive. Many customers are directed through comparison sites meaning they have already gone through one hurdle to access service offerings.
- Are our critical policy and claims management systems up and running?
- 1% of the requests in the ‘submit claims’ business transaction were categorized as very slow, investigate these transaction snapshots.
- Resource consumption has spiked on a cluster supporting the payment-processing tier… where is the performance bottleneck that is causing this?
When performance degrades beyond 2 standard deviations from the calculated baseline for the ‘submit credit check’ business transaction, provision another private cloud instance of that web service tier.
Automatically restart the application server instance when the number of stalled threads exceeds a specific numerical threshold.
Remove the server from the load balancing pool when the average response time is beyond 3 standard deviations beyond the rest of the cluster.
- Create a thread dump when the number of ‘Add Claim’ business transactions perform 2 standard deviations below the baseline.
If we are planning on adding a new e-commerce insurance digital channel next year, and we are expected to growing the number of claims we process by 25% next year, how much hardware and software will we need to support that growth?
Business intelligence and performance monitoring
How many claims were processed per minute from our Tier 1 customers during the last hour and how did response time affect revenue during that period?
End user intelligence
How many customers clicked through from a comparison site in the last 12 hours who were not converted to customers? How does this compare to the average for this timeframe last week?
"Within 5-10 minutes we were able to pinpoint where the bottleneck was—at the exact line of the code."
Mark Dash, Assistant VP of Information Management and Technology, Penn Mutual