Frequently Asked Questions

General | Claim Scoring | Pricing | Reserving | Fraud

General Questions

What methods or techniques can Claim Analytics provide to us that we don't receive from our own actuaries?

In-depth expertise in advanced predictive modelling.

Modern computing allows us to analyze data in new and extremely powerful ways. Claim Analytics harnesses the powers of modern computing — powerful processors, online data storage, and powerful predictive tools — to further reveal and quantify insurance opportunities.

Our data is confidential - how can we be comfortable about sharing it?

Understood! We believe strongly that raw data belongs to our client – not us. A company’s data is used for the benefit of, and only for the benefit of, that company. We do not share any information about one company’s data with any other company (unless approval has been obtained, as in the case of a joint project) and we are happy to sign non-disclosure agreements concerning data.

We're worried about privacy. How does Claim Analytics protect the privacy of our policy holders and claimants?

No data that could uniquely identify an individual (i.e. name, social security number) is requested or used in our analysis.



Disability Claim Scoring Questions

What is a predictive scoring model?

A predictive scoring model uses data mining techniques to predict likelihood of return to work in a given timeframe.

Its predictions are given as scores from 1 to 10. The higher a score a claimant receives, the greater the likelihood of return to work.

What do you mean by claim scoring?

Claim scoring is a tool to help disability claim experts in their day-to-day work. Claim scoring provides a fast, objective, consistent way of ranking the quality of each disability claim.



Claims are scored from one to ten. The higher the score, the greater the likelihood of return to work. The scores are calibrated: a score of 1 indicates a 0-10% likelihood of RTW, 2 indicates 11-20%, and so on.

Why should we score our disability claims?

Claims experts are confronted daily by a myriad of claims that they must work through one by one, evaluating and ranking as they go. Claim scoring eliminates a significant amount of this work.

Claim scoring provides an accurate, objective starting point for each new claim. Claims experts and clinical staff can use the scores - and the time freed up by expedited evaluation - to help focus their efforts where they will be most productive.

Is claim scoring of any benefit to experienced claims people?

Yes. In fact, we find that the benefits of claims scoring are best realized in sophisticated claims departments.

Claim scoring, because it precisely and objectively ranks each disability claim, can be used by claims managers to optimize the allocation of resources, and to decide which claimants will benefit most from the various case management options available.

Our practices are unique. How do we know the scoring model will work for our company?

Every Claim Analytics scoring model is tailored to the client - built from the client’s own claims database. Its scores predict recovery based on that client’s unique claim practices.

We test each model upon completion, using a set of historic data that is new to the model.

In the chart below, a typical example of one such testing, note that actual results closely track the model’s predictions. Example: In the dataset below, actual aggregate recovery for 6's was 57% - closely matched by the model's predicted aggregate recovery for 6's: 50-60%.

How many claims are required to build a statistically robust scoring model?

A statistically robust scoring model can be built with as few as 500 historical claims.

Will the implementation of claim scoring tie up staff in new learning or new operational activities?

No. The Claim Analytics model is built and maintained offsite. Our clients will attest to the fact that the process is non-intrusive to their operations.

Scoring of new claims is only a matter of sending data on a weekly or monthly basis.

What are the systems and data requirements for building the model?

There are no systems requirements, as the model is built and run offsite. Scores are sent via an Excel file.

There are also no specific data requirements. Each model is tailored to the claims data that a client can readily provide. A suggested list of data fields can be found here.

Is the service expensive and will the benefits offset the cost?

The Claim Analytics service is all about saving money. The benefits of implementing the system far exceed the cost of the services.

Significant financial benefits are realized in two areas:

  1. Expense savings, through optimized use of rehab, medical and investigation services
  2. Improved claims experience.

Is there a trial period?

Yes.



We are so confident in the value of our claim scoring system that we build the model for our clients at our own cost, and then allow a trial period of six months at 50% of our normal monthly charge. There is no obligation to the client to continue after the trial period.

How can a model developed by actuaries help disability claims people in their work?

Every scoring model is based upon, and “learns” from, a company’s own historical claims database. By identifying complex patterns and relationships within that database, the model is able to provide information (the scores) that can be used as a decision-making tool in the claims management process.

Pricing Questions

What does the Claim Analytics pricing approach offer beyond current methodologies?

Claim Analytics provides expertise in cutting-edge predictive modeling and pattern quantification techniques. The technology we employ is demonstrably superior to classic actuarial tools in quantifying the relationship between claim drivers and claim costs. This technology is especially strong at analyzing several factors in simultaneity ( age, gender, EP, claim size, industry, occupation, salary, benefit level, change in definition period, region and many more).

In addition, our pricing capabilities are enhanced by our reserving capabilities. Our claimant-specific termination rates result in reserves that are significantly more accurate at the individual claimant level. This reserving approach allows us to more accurately allocate claim costs, and thus better quantify the relationship between claim drivers and claim costs.


We don’t feel comfortable using a black box for rate setting.

Neither would we. Our analysis enables our clients to isolate and quantify the influence of each claim driver on claim costs. We provide our clients with superior knowledge of the true cost of each claim driver. Our clients then use this knowledge to finalize their rate factors.

How will working with Claim Analytics affect our pricing process?

The Claim Analytics pricing methodology results in little or no impact on processes.

To start, clients provide us with historic census and claims data – exactly the same data as they normally use for pricing. We then perform a predictive analysis and provide a report in a format that meets client needs.

Reserving Questions

How does the Claim Analytics LTD claim reserving methodology differ from others?

Current reserving methodologies determine termination rates using simple statistics and only a limited set of claimant characteristics. The Claim Analytic's approach to calculating LTD claim reserves is to determine claimant-specific termination rates.

The Claim Analytics approach is to apply advanced predictive modeling techniques to determine very precise, claimant-specific, termination rates based on all the key drivers of termination, such as: age, gender, elimination period, primary and secondary diagnoses (not just diagnostic category), occupation, industry, income, benefit level, change in definition date, geographic region and number of previous claims by the same claimant.

What are the advantages of the Claim Analytics LTD claim reserving methodology?

The Claim Analytics reserving methodology results in:

Fraud Detection Questions

How does the Claim Analytics fraud detection approach differ from current methodologies?

The most common approach to fraud detection is to apply rules-based technology. Claims are screened for procedures that match known types of fraud.

This approach has two weaknesses.

  1. A rules-based approach cannot screen for new, unknown, types of fraudulent behavior.
  2. A rules-based approach does not recognize that each practitioner/claimant has a history of claims and that some practitioners/claimants have demonstrated a greater or lesser degree of atypical activity.

The Claim Analytics approach to screening for fraud is to analyze all claims paid by the insurer over the past 12 or more months. We use advanced pattern-detection technology to:

  1. Categorize each claim as typical or atypical,
  2. Roll this information back up to the practitioner level, and thereby
  3. Identify practitioners with atypical claims patterns.

Our approach allows the insurer to identify and concentrate on those practitioners with the most suspicious claims activities – regardless of whether or not their behaviors match already known fraud schemes.

What are the data requirements to work with Claim Analytics?

Claim Analytics requires an extract of recent claims paid, typically 6 to 24 months of data. The primary data fields we require are: claim #, date service performed, procedure/product code, practitioner ID code, claimant ID#, age of claimant, type of work performed, total amount billed and total amount paid.

Claim Analytics was founded in 2000 by Barry Senensky and Jonathan Polon, both Canadian life insurance actuaries. Barry Senensky and Jonathan Polon had worked for the same major life insurance firm for many years, and knew that data mining techniques and predictive modeling were under-utilized by insurers. They were interested in starting their own business in the life insurance field. Their first predictive model they were commissioned to build was a claim scoring model, that scored ltd [long term disability] [LTD] claims on likelihood of return to work (rtw).

Jonathan Polon had recently led a team of model-builders in creating a highly successful predictive model to detect credit card fraud. He and Barry Senensky were convinced that many more applications for pattern detection and predictive models existed in insurance.

They chose disability claim scoring as their first project for a number of reasons. First, because almost all firms providing group long term disability [LTD] insurance held a considerable database of stored claims information. Finding an adequate amount of data to build a claims scoring model would not be a problem.

A second reason was the costliness of long term disability claims in the field of insurance. Barry and Jonathan knew that, with the scale of investment involved, even relatively minor steps forward in the management of group disability claims management would lead to considerable savings. Claim scoring was a powerful tool in bringing greater efficiencies and reduced costs to disability claims management. Predictive modeling, particularly claims scoring, offered huge potential for software support of claims adjudicators, claims processors, and claims managers.

Claim scoring proved to hold even greater promise than Barry and Jonathan had envisaged. In addition to predicting, with startling accuracy and precision, the likelihood of a disability claimant returning to work, claims scoring became a tool for ongoing improvement of claims management.

Claim Analytics now offers pattern detection and predictive models to major insurance firms to assist in objectivity and accuracy of pricing, reserving, and fraud detection, as well as custom models.

[The proper name of the company is "Claim Analytics," not "Claims Analytics."] [The proper name of the company is "claim analytics," not "claims analytics."]