This article is a part of AMS Services’ customer “Success Story” series. If you would like to collaborate with AMS Services on a project or have a success story that you would like to share, contact us using the following link: http://amscarriercommunity.com/phoenix/ContactUs/tabid/137/Default.aspx
The Goal
The actuarial department at FCCI Insurance Group, headquartered in Sarasota, Florida has a goal: They want to understand how anticipated rate changes for the upcoming year impacts premium before filing for the new rates. This means that they need to rate policies twice: once using the current rates to establish a baseline, and at least once again with the new rates. Since the final decisions on rate changes are not certain until the rating results are reviewed, this modeling of rate changes is considered an iterative process and not likely to be final with a single pass.
The actuarial department uses rate comparison information in Microsoft ® Excel spreadsheet during their analysis. The spreadsheet groups key policy data – including the premium differences between the baseline set of rates and the proposed changes – by state and line of business into what FCCI refers to as their “Premium Comparison Report.” Ultimately, FCCI wants to model approximately 20,000 policies.
How would you approach this problem?
The Process
Most of us would immediately agree that meeting the business problem is something that should be achievable with present computer technology, given that we are in the twenty-first century! Of course, some problems are not as easy to solve as they first appear. In this case, the process imposed by the Phoenix system (designed and built “a long time ago” in computer years), proved to be a significant burden and bottleneck to FCCI.
The sole individual tasked with generating this information at FCCI was Saba Gobalasingham, CPCU and a Senior Systems Analyst. Since many of FCCI’s policies take considerable time to rate – some of their commercial policies can take up to five hours – Saba realized that he would need to use multiple workstations in order to get this job done in anything resembling a reasonable time frame.
Phoenix in its traditional form is what is referred to as a “fat client” – with all of the logic executing on each workstation. While processing power is effectively added by each additional workstation, you really need to add a warm body to perform the work when you add a workstation – an option that was not available to Saba. So, how did Saba go about rating all of those policies?
Saba “leveraged” technology by coming in on nights and weekends, logging into multiple Phoenix workstations and starting the rating process on a policy-by-policy basis. As you can imagine, the total throughput was limited to both the number of workstations available and Saba’s ability to get to each workstation to start the rating process (while keeping track of what he was doing). Clearly not a scalable or desirable solution! Saba estimates that it took over 200 hours to rate 5,000 policies.
Saba was sure there had to be a better way…
The Solution
Saba began thinking of alternatives even before he attended the 2007 Phoenix User Group conference. His initial thinking involved using the Phoenix Bulk Re-Rate product; and with this in mind Saba struck up a conversation with Paul Corriveau, the Software Architect at the Portland, Maine office of AMS Services. Paul’s immediate response to Saba’s dilemma was: “That’s easy, no problem.” Paul suggested leveraging the use of Phoenix Web Services, something being introduced at the conference.
Saba attended sessions about Phoenix Web Services, but he did not feel that he fully understood them. Nevertheless, when Saba returned from the conference he took this information back to his management, indicating that there was a viable solution available.
Further conversations between FCCI and AMS Services sorted out the particulars: AMS Services would develop a custom application specifically tailored to meet FCCI’s needs, targeting one line of business for the initial implementation. Since Phoenix Web Services was available to perform the “heavy lifting” of retrieving the data and rating it, the project’s efforts focused primarily on the needs FCCI had around policy selection criteria and the output of the information into a comma-delimited file that could be easily imported into Excel.
Saba was hopeful, but there were lingering doubts in his mind. While he felt that definite improvements possible, Phoenix Web Services were still somewhat of a mystery to him. There was always that nagging, question in the back of his mind: How much faster and easier would the final solution really be?
The Design
The application developed for FCCI incorporated the following key elements:
· The need to specify the location of the Phoenix Web Services, the location for the comma-delimited output file, and the two sets of ART trees to use (using the current rates and new rates).
· The ability to select the line of business and in-force date of the policies to filter the number of policies to be processed.
· Throughput; the application needed to invoke rating on multiple policies at one time in order to process a large amount of policies in a short period of time.
Eric Mathis and Greg Gibson were the developers assigned to the project from the AMS Services end. The process started by developing a “paper prototype” of the user interface, along with formally documenting the requirements in the form of user stories and sending this information back to FCCI for confirmation.
While FCCI reviewed the information, Eric and Greg met with Paul Corriveau to discuss the design specifics. Since speed was essential, the application incorporated a multi-threaded design, enabling it to process multiple policies at one time. Upon acceptance of the requirements, development began in earnest.
The Challenges
As with any software project, there is always the danger that something will defy the estimate, even when you have brainstormed around potential issues and risks and added some buffer time to address unanticipated problems. This project, as small-sounding as it was, leveraging already-developed and tested Phoenix Web Services for the bulk of the business processing, still encountered difficulties!
What were those problems? For a start, the version of Phoenix that FCCI was using in production was version 7.01, and Phoenix Web Services was released with version 8.0. Should this be an issue?
Not really. AMS Services – understanding the desires of customers to leverage these new Web Services – is open to installing Phoenix Web Services in 7.x environments, particularly since Phoenix Web Services are installed on a separate server and accesses the Phoenix database with its own routines. This means that Phoenix Web Services are not closely tied to the Phoenix code base (reducing the risk of problems), with one small exception: ART rating.
Phoenix Web Services uses the latest ART rating engine; Due to changes made at Phoenix version 7.02, pre-version 7.02 ART trees need to be re-compiled for use with Phoenix Web Services. The only implication here is that in pre-version 7.02 Phoenix environments where Phoenix Web Services are also being used, two sets of trees need to be maintained.
Since FCCI really wanted to perform rate comparisons to plan for future rating changes, and any changes made would be done on the “future” ART trees, this was not an issue. In fact, Phoenix Web Services are designed with the ability to configure and reference different deployments, making it very easy for the Rate Modeling Tool to use two sets of trees as FCCI desired.
In order to develop this application, FCCI provided AMS Services their latest database and ART trees at the version 7.01 level. AMS Services worked at re-compiling the ART trees for use with Phoenix Web Services, which on paper was thought to be a simple exercise. As it turned out, there were a couple of difficulties.
FCCI’s ART trees are very complex. Since only one line of business was being implemented with the initial version of the tool, considerable time and energy was spent between Cheryl McKee of AMS Services (LOB developer and tester on this project) and FCCI to determine what trees needed to be used for this line of business. The time to work through this was not anticipated by either party when the project was conceived and estimated, but was obviously necessary in order to proceed.
The next two problems arose when actual rating started. FCCI has some ART rating calls built into their trees that were designed to halt rating for various reasons. Since the goal was to obtain premium for all of the policies, these calls needed to be hunted down and commented out to prevent them from interfering with the rate modeling – another time-consuming, unplanned task. (On the plus side, some of the re-compiling time and effort was reduced through use of the ART Tree Mass Compiler tool, introduced with Phoenix version 7.02.)
The third and final difficulty with the project emerged late in the game. Despite extensive testing prior to the release of Phoenix version 8.0 (where rating was invoked multiple times in the testing of Automated Out-Of-Sequence processing), a problem surfaced within Phoenix Web Services rating.
Without going into all of the gory details, the problem was traced to an issue with compiler optimizations with the Microsoft.NET framework accessing older, legacy code like ART. This problem was escalated to Microsoft, which demanded considerable development attention to work them through defining and duplicating the issue with a test application.
Finally, despite these unexpected difficulties, everything was ready for delivery.
The End Result
The culmination of this project was an on-site visit by Eric Mathis of AMS Services to install Phoenix Web Services and the custom application, with the goal to ensure that everything worked as it did in the AMS Portland, Maine office.
Eric was delighted to fly to FCCI’s Sarasota, Florida office – and not because it was cold in Maine! (Well, maybe that was part of it.) Eric also had the following motivations:
1. He wanted to see this successfully implemented at FCCI.
2. His parents live in Sarasota, Florida and this would be a good opportunity for a quick visit.
The trip was a win-win for both Eric and FCCI. Until Eric visited FCCI, Phoenix Web Services were a distant concept to Saba. Once he worked with Eric installing them as well as seeing how the custom application applied Phoenix Web Services to his business problem, Saba became a believer. “I showed it to some of our developers – sharp developers – and they said it was a slick solution,” Saba says.
Just how much did this custom effort affect Saba’s workload? Recall that Saba spent over 200 hours working nights and weekends to manually rate and generate the comparison report for 5,000 policies. Today, with the custom tool and Phoenix Web Services the time is down to an astounding 1.5 hours to rate 5,000 policies, including the automatic generation of the comma-delimited file, ready for import into Microsoft ® Excel! Sounds more like the twenty-first century, doesn’t it?
More to Come
At the time of this writing, FCCI is working with AMS Services to scope the next phase of this project, involving other lines of business and knowledge transfer so that FCCI can become self-sufficient with the tool. Some usability enhancements have been requested to the tool itself, and the knowledge transfer will be around the use of XSLT – used to output the comma-delimited file with the desired policy data. Aside from the need to generate unique reports per line of business, the tool is otherwise self-contained.
In addition, while the original Rate Comparison workflow was deemed too difficult and time-consuming to put in the users’ hands, FCCI plans to reverse this thinking and do just that – once the final enhancements to the tool are in place. FCCI considers the tool to be both usable and fast enough that there is no reason not to let the users do their own modeling.
When asked if he would ever want to go back to his old method, Saba responded with a firm “no.” Saba went on to add, “Let someone try it with just one-hundred policies and see how they feel. If they feel it – feel the pain and see how hard it is – they won’t want to do it that way either.”
AMS Services is thrilled to help implement Phoenix Web Services for the benefit of FCCI, and is equally delighted to relieve Saba’s pain.