Deploying prices with errors can have significant implications for insurers and pricing leaders. Many insurers have gone live with problems in their rating, but this is hardly disclosed nor broadcast to the world. Many times, problems are found quickly. Sometimes they are not found for some time. In the past, prices were relatively simple and could be checked by looking at a small sample of quotes. This won’t work anymore because the pricing process is necessarily complicated. Insurers who can consistently set live rates without errors will win at pricing. According to Sun Tzu, “The greatest general is he who makes the fewest mistakes."
Rates need to be thoroughly tested before deployment. The main challenge in testing prices is the necessary complexity of the rating engine. Pricing models often involve multiple rating factors over multiple layers that determine the final price. While there I only one end price there is a myriad of outputs for customer comms and MI. Running complete batch tests from scratch, that is getting thousands of quotes and checking they produce the right output, is onerous and often impractical. This needs a separate version of the entire rating from which the test cases are taken. It leads to a classic problem of how you check the test batches. Additionally, this way means outputting large volumes of results. This can be reviewed to identify any pricing errors; however, manual review is time-consuming and itself also prone to human error. Insurance pricing leaders are advised to be cautious when relying on manual reviews to ensure pricing accuracy.
Effective Price Checking To increase the effectiveness of price checking and minimize errors, the following suggestions can be followed:
Designing rating to be modular so that unit testing small pieces is possible.
Regression testing a large batch of test quotes to see how the output from the new models differs to the previous ones. Using this to create diagnostics like average change and checking this for all outputs.
Setting tolerances and expectations for the outputs and layers. Then Automatically flagging when outputs are outside these tolerances.
Post-live automatically comparing what is delivered in live with what was expected before deployment.
Post-live checking that values remain within tolerance and automatic flagging if they do not.
Real-time monitoring of pricing values compared to expectations and tolerances.
Top Tips to Consider During Price Checking
Do not rely on people alone to review your prices.
Remember that even items that do not affect the end price may still cause problems such as those printed on customer communications or included in MI.
The ultimate in checking is to know independently of the rating engine what every premium should be – but this is far from practical.
Set common-sense tolerances in your system which flag if they are broken. For example, layers or parts of the premium exceeding sensible limits. Or parts of the calculation delivering negative values.
Always have the ability to switch easily between rating series if one is found to have a problem.
Carry out the following exercise to determine what you’re doing right and what needs to be changed:
Collate the impact on GWP and profitability of any rating errors that have occurred in the past five years.
Record rating errors and note if they impacted customers
Keep information on customer feedback and complaints related to errors.
Manually checking prices is a start but it will not prevent mistakes. The ultimate in checking prices is to know independently of the rating engine what every premium should be – but this is not often practical. Price checking needs to be thorough, automated, and effective. However much it costs to check your rates are correct is nothing compared to the potential impact of incorrect prices. Insurance pricing leaders need to implement testing regimes that can identify pricing errors before prices are deployed. They also need to have a continual process to check rates in real-time.
Meet me to develop your checking and testing strategy, find out what you need to be doing, get knowledge about what your peers are doing, learn how to prevent rating problems for good, and win at pricing.