Tag Archives: insurance software

The Devil Is In The Details

Movies about space missions that result in catastrophe can teach us a lot about how not to manage a project (the “successful failure” of Apollo 13 comes to mind). Yet there are actual space mission catastrophes — the loss of the 1999 Mars Climate Orbiter (MCO), for example — that also offer valuable lessons in preventing fundamental mistakes.

The MCO was the major part of a $328 million NASA project intended to study the Martian atmosphere as well as act as a communications relay station for the Mars Polar Lander. Famously, after a nine-month journey to Mars, the MCO was lost on its attempt to enter planetary orbit. The spacecraft approached Mars on an incorrect trajectory and was believed to have been either destroyed or to have skipped off the atmosphere into space. The big question naturally was: What caused the loss of the spacecraft?

After months of investigation, the primary cause came down to the difference between the units of output from one software program and the units of input required by another. How, the media asked, could one part of the project produce output data in English measurements when the spacecraft navigation software was expecting to consume data in metric?

Those of us involved in expensive and high-risk projects would ask the similar question: How could this happen? What follows are a few findings from the Executive Summary of the Mars Climate Orbiter Mishap Investigation Board (MCO MIB), with lessons for us all.

  • The root cause of the loss of the MCO spacecraft was the failure to use metric units in the coding of a ground software file used in trajectory models. Specifically, thruster performance data in English units were used instead of metric units in the software application code.
  • An erroneous trajectory was subsequently computed using this incorrect data. This resulted in small errors being introduced in the trajectory estimate over the course of the nine-month journey.

That erroneous trajectory was the difference between a successful mission and failure. Lockheed Martin Astronautics, the prime contractor for the Mars craft, claimed some responsibility, stating that it was up to its company’s engineers to assure that the metric systems used in one computer program were compatible with the English system used in another program. The simple conversion check was not done. “It was, frankly, just overlooked,” said their spokesman.

Just overlooked? Those of us in project management know that large-scale projects require the ability to see not only the big picture — the goals and objectives of the project — but also the details.

While not as prominent as space exploration, insurance software development also has millions of dollars at stake. Insurance products can be very complex, and the interactions required in business systems along with the calculations involved are all critical to producing accurate results.

Errors in the way in which calculations are derived can produce problems ranging from failure to comply with the company’s obligations under its filings to loss of revenue. Even apparently simple matters such as whether to round up or down on a calculation can have profound impacts on a company’s bottom line.

Although the failure to address the difference between English and metric measurements was identified as the root cause of the problem with the MCO, the real issue at hand is what caused that failure. How was it missed?

Taking a project management perspective requires asking the question, “Why?” Why was a key element overlooked? What led an experienced team to miss a crucial detail?

In the search for answers, it’s interesting to look deeper inside the report by the Mars Climate Orbiter Mishap Investigation Board (MCO MIB). In addition to the root cause of failure to use standard units of measurement across the entire project, the report found a series of other issues that also contributed to the catastrophe. The following are other lessons of the MCO mission and how they can be applied more widely to project management.

  • Lack of shared knowledge. The operations navigation team was not familiar enough with the attitude controls systems on the spacecraft and did not fully understand the significance of errors in orbit determination. This made it more difficult for the team to diagnose the actual problem they were facing.

    It is likewise common for insurance software projects to have mutually dependent complex areas — for example, between the policy administration system and the billing system. If one team does not fully understand the needs of the other, there can be costly gaps in understanding.

    The MCO MIB recommended comprehensive training on the attitude systems, face-to-face meetings between the development and operations team, and attitude control experts being brought onto the operation’s navigation team. Similarly, face-to-face meetings between the policy experts and the billing experts, between the business side and the technology side, will go a long way toward a successful project. In the world of e-mail and instant messaging, I think all of us spend less face-to-face time. Nonverbal communication is 60% of our communication and is often very helpful; there’s zero face time when we rely on electronic communication.

  • Lack of contingency planning. The team did not take advantage of an existing Trajectory Correction Maneuver (TCM) that might have saved the spacecraft, since they were not prepared for it. The MCO MIB recommended that there be proper contingency planning for the use of the TCM, along with training on execution and specific criteria for making the decision to employ the TCM.

    The need for contingencies in insurance software development is important too. Strong project management will consider project risks and therefore contingencies. And contingency plans are important at every stage — development, implementation, and once the system is live. Issues must be dealt with rapidly and effectively since they have an impact on the entire business. Regular reviews of the contingency plans are also useful.

  • Inadequate handoffs between teams. Poor transition of the systems’ engineering process from development to operations meant that the navigation team was not fully aware of some important spacecraft design characteristics.

    In complex insurance software projects, there are frequent handoffs to other teams, and the transition of knowledge is a critical piece of this process. These large, complex projects should have a whole team dedicated to ensuring knowledge transfer occurs. No matter how good the specifications, once again, it’s vital to get face to face.

  • Poor communication among project teams. The report stated there was poor communication across the entire project. This lack of communication between project elements included the isolation of the operations navigation team (including lack of peer review), insufficient knowledge transfer, and failures to adequately resolve problems using cross-team discussion. As the report further notes:

    “When conflicts in the data were uncovered, the team relied on e mail to solve problems instead of formal problem resolution processes. Failing to adequately employ the problem-tracking system contributed to this problem slipping through the cracks.”

    This area had one of the largest set of recommendations from the MCO MIB, including formal and informal face-to-face meetings, internal communication forums, independent peer review, elevation of issues, and a mission systems engineer (aka really strong program or project manager) to bridge all key areas. Needless to say, this kind of communication is a critical part of any insurance software project, and these lessons are easily applied. Zealously hold project reviews (walk-throughs). Do them early and often. The time spent will pay you back with success.

  • The Operations Navigation Team was inadequately staffed. The project team was running three missions simultaneously — all of them part of the overall Mars project — and this diluted their attention to any specific part of the project. The result was an inability of the team to effectively monitor everything that required their attention.

    Sound familiar? We just experienced this on a software implementation project where the software vendor outsold its capacity to be successful. Projects are expected to run lean because of cost considerations, but it’s always important to ensure that staff is not stretched to the point of compromising the project.

  • There was a failure in the verification and validation process, including the application of the software standards that were supposed to apply. As the MCO MIB noted:

    “The Software Interface Specification (SIS) was developed but not properly used in the small forces ground software development and testing. End-to-end testing to validate the small forces ground software performance and its applicability to the specification did not appear to be accomplished.”

Every project manager will recognize the need to stick to protocol and agreed-upon processes during a software project. Ensuring that project team members know the project/system specifications and standards is essential to successful project delivery.

And so, the devil is in the details. My career in and around insurance technology has spanned three decades now. While I have learned much, two things are abundantly clear:

  • There is no substitute for really good project management.
  • There is no substitute for great business analysts.
  • There is no substitute for great communication.

Okay, make that three things! It’s bonus day.

The full MCO report is available here.

Insurance And Manufacturing: Lessons In Software, Systems, And Supply Chains

Recently, my boss Steve and I were talking about his early career days with one of those Big 8, then Big 6, then Big 5, then Big 4 intergalactic consulting firms. Steve came out of college with an engineering degree, so it was natural to start in the manufacturing industry. Learning about bills of material, routings, design engineering, CAD/CAM … “Ah yes,” he recalled, “Those were heady days.” And all those vendor-packaged manufacturing ERP systems that were starting to take the market by storm.

Eventually Steve found his way into the insurance industry, and thus began our discussion. One of the first things that struck Steve was the lack of standard software packages in the insurance industry. I don’t mean the lack of software vendors — there are plenty of those. Seemingly, though, each software solution was a one-off. Or custom. Or some hybrid combination. “Why?” we wondered.

The reasons, as we now know, were primarily reflected in an overall industry mindset:

  • A “but we are unique!” attitude was pervasive. Companies were convinced that if they all used the same software, there would be little to differentiate themselves from one another.
  • There was also an accepted industrywide, one-off approach. Conversations went something like this: “XYZ is our vendor. We really don’t like them. Taking new versions just about kills us. We don’t know why we even pay for maintenance, but we do.”

But the chief reason for a lack of standard software was the inability to separate product from process. What does this mean?

Well, you can certainly envision that your auto product in Minnesota is handled differently than your homeowners’ product in California. I’m not referring to just the obvious elements (limits, deductibles, rating attributes), but also the steps required for underwriting, renewal, and cancellation. Separation of product from process must go beyond the obvious rate/rule/form variations to also encompass internal business and external compliance process variations.

But there’s still plenty of processing — the heavy lifting of transaction processing — that’s the same and does not vary. For example, out-of-sequence endorsement processing is not something that makes a company unique and therefore would not require a custom solution.

Where the rubber meets the road, and where vendor packages have really improved their architecture over the last several years, is by providing the capability in their policy admin systems for companies to “drop” very specific product information, along with associated variations, into a very generic transaction system.

Once product “components” (digitized) are separated from the insurance processing engine, and once companies have a formal way to define them (standard language), they can truly start making their products “unique” with reuse and mass customization. Much like those manufacturing bills of material and routings looked to Steve way back when.

This separation of policy from product has been a key breakthrough in insurance software. So what is an insurance product, at least in respect to systems automation?

From Muddled To Modeled
The typical scenario to avoid goes something like this:

  • The business people pore over their filings and manuals and say, “This is the product we sell and issue.”
  • The IT people pore over program code and say, “That’s the product we have automated.”
  • The business people write a lot of text in their word processing documents. They find a business analyst to translate it into something more structured, but still text.
  • The business analyst finds a designer to make the leap from business text to IT data structures and object diagrams.
  • The designer then finds a programmer to turn that into code.

One version of the truth? More like two ships passing, and it’s more common than you may think. How can organizations expect success when the product development process is not aligned? Without alignment, how can organizations expect market and compliance responsiveness?

What’s the alternative? It revolves around an insurance “product model.” Much like general, industry-standard data models and object models, a product model uses a precise set of symbols and language to define insurance product rates, rules, and forms — the static or structural parts of an insurance product. In addition, the product model must also define the actions that are allowed to be taken with the policy during the life of the contract — the dynamic or behavioral aspect of the product model. So for example, on a commercial auto product in California, the model will direct the user to attach a particular form (structure) for new business issuance only (actions).

Anyone familiar with object and data modeling knows there are well-defined standards for these all-purpose models. For insurance product modeling, at least currently, such standards are more proprietary, such as IBM’s and Camilion’s models, and of course there are others. It is interesting to note that ACORD now has under its auspices the Product Schema as the result of IBM’s donation of aspects of IAA. Might this lead to more industry standardization?

With product modeling as an enabler, there’s yet another key element to address. Yes, that would be the product modelers — the people responsible for making it work. Product modeling gives us the lexicon or taxonomy to do product development work, but who should perform that work? IT designers with sound business knowledge? Business people with analytical skills? Yes and yes. We must finally drop the history of disconnects where one side of the house fails to understand the other.

With a foundation of product modeling and product modelers in place, we can move to a more agile or lean product life cycle management approach — cross-functional teams versus narrow, specialized skills; ongoing team continuity versus ad hoc departmental members; frequent, incremental product improvements versus slow, infrequent, big product replacements.

It all sounds good, but what about the product source supplier — the bureaus?

Supply Chain: The Kinks In Your Links
Here is where the comparison between insurance and manufacturing takes a sharp turn. In their pursuit of quality and just-in-time delivery, manufacturers can make demands on their supply chain vendors. Insurance companies, on the other hand, are at the mercy of the bureaus. ISO, NCCI, and AAIS all develop rates, rules, and forms, of course. They then deliver these updates to their member subscribers via paper manuals or electronically via text.

From there the fun really begins. Insurance companies must log the info, determine which of their products and territories are impacted, compare the updates to what they already have implemented and filed, conduct marketing and business reviews, and hopefully and eventually, implement at least some of those updates.

Recent studies by Novarica and SMA indicate there are approximately 3,000 to 4,000 changes per year in commercial lines alone. The labor cost to implement just one ISO circular with a form change and a rate change is estimated to be $135,000, with the majority of costs in the analysis and system update steps.

There has got to be a better way …

ISO at least has taken a step in right direction with the availability of its Electronic Rating Content. In either Excel or XML format, ISO interprets its own content to specify such constructs as premium calculations (e.g., defined order of calculation, rounding rules), form attachment logic (for conditional forms), and stat code assignment logic (to support the full plan).

A step in the right direction, no doubt. But what if ISO used a standard mechanism and format to do this? ACORD now has under its control the ACORD Product Schema. This is part of IBM’s fairly recent IAA donation. It provides us a standard way to represent the insurance product and a standard way to integrate with policy admin systems. What if ISO and the other key providers in the product supply chain started it all off this way?

Dream on, you say? While you may not have the clout to demand that the bureaus change today, you do pay membership fees, and collectively the members have a voice in encouraging ongoing improvements in the insurance “supply chain.”

In the meantime, the goal to be lean and agile with product life cycle management continues. We must respond quickly and cost-effectively to market opportunities, policyholder feedback, and regulatory requirements. That all starts at the product source … but it doesn’t end there. So while the supply chain improves its quality and delivery, insurance companies will need to gain efficiencies throughout every corner of their organizations in order to achieve those lean goals.

In writing this article, David collaborated with his boss Steve Kronsnoble. Steve is a senior manager at Wipfli and an expert in the development, integration, and management of information technology. He has more than 25 years of systems implementation experience with both custom-developed and packaged software using a variety of underlying technologies. Prior to Wipfli, Steve worked for a major insurance company and leverages that experience to better serve his clients.