What are the main problems insurers face when it comes to delegated authority data?
With delegated authority in insurance, the underwriter has less control over the data received from the original insured. They rely on the coverholder, or in the case of a claim, the Third Party Administrator (TPA) to collect quality information in a format that can be easily read by the underwriters’ own systems.
Today, this information is usually provided in what are known as Bordereaux – detailed reports or summaries of risks, premium and claims, usually in the form of Excel spreadsheets. However, much of the data entering the specialty insurance market, also including SOVs (Statement of Values) for example, comes in different formats with no standard schema.
These different data formats mean that, in the raw state, there are many limitations on how the underwriter can use the data. In most cases, someone needs to manually clean up and reformat the spreadsheets in order to extract meaningful insights and ingest the data for analysis. It might then be used to compare key insights across different coverholders for example. This is very time-consuming and expensive to do with people working on these tasks either in-house, or having to pay a third party to do this data cleansing.
What are your clients asking for?
At Scrub AI, we have developed AI-based tools to instantly clean up complex data regardless of the original format, leading to significant cost savings. Our clients benefitting from this, are realising new opportunities to reliably analyse their data in downstream systems, where they were having to join the dots and use unreliable data previously. With clean, validated delegated data automatically pushed to downstream systems and data warehouses, clients can accurately track:
- How many claims were made during a specific period, and how many remain open?
- How has exposure information changed year over year?
- Are coverholders paying in a timely manner?
- Have coverholders adhered to all the terms in the bordereaux?
- Should insurers adjust terms and conditions based on last year’s performance?
Our clients were otherwise unable to effectively and efficiently ingest this data since no one else was providing these capabilities, so we leveraged the Machine Learning models we had already built and trained for SOV data ingestion and built a tool ourselves.
Why else do your clients come to Scrub?
Our unique selling point has always been the ability to clean, validate and transform data so that it can be ingested into any database or platform. That’s where we’ve started. This means we have created a credible and reliable set of machine learning models for data cleaning. This can now be brought to life with the additional functionality we are offering, that is very fast and provides important business insights to our clients.
We find that insurers often do not have a platform to examine their own data or what they receive from coverholders. They also lack a consistent, efficient way to create reports or the internal resources to do the work.
We offer the ability to check and report on every row of the data from every coverholder and TPA.
What are the products you offer?
The whole package is our “Delegated Authority (DA) platform”. It is offered in two main parts.
The first is the “Data Cleansing” platform which cleans, validates and transforms data to a standardised output, such as v5.2. The validation checks include ensuring the data is complete and accurate, and intelligently suggesting corrections or data to fill in blanks. For example, flagging invalid currencies, dates and data types. Validation checks also include flagging where data in the bordereau falls outside of the binder terms and conditions, for example, flagging where a claim has been paid or premium received for a location outside of the territorial limits of the binder.
The second part of the product is the “ScrubHub”. This is a data warehouse where clients can store their cleaned data. It also comes with additional validation checks that check the current bordereaux against historical bordereaux in the warehouse and flag irregularities, such as where an endorsement or cancellation has been received against a risk that does not exist. Later this year a reporting portal will also be released allowing clients to build automated, custom reports on any of this clean data.
How much is the lack of standardisation costing the industry today?
We believe that the industry is wasting between $2 – $5 billion USD a year on unnecessary and repeated data processing.
Can you provide more detail about how this solution addresses this problem?
Our USP is focusing on cleaning and transforming data. We clean every row of every bordereau using an AI-first approach. We use specially trained machine learning models, developed from our existing products and apply these highly accurate models to our new product.
Regardless of how messy the data is or what format it is in, we are able to ingest it. Additionally, our AI learns how an individual user cleans data, as well as the particular characteristics of each coverholder that submits data. This improves efficiency over time, making the system smarter as its usage grows.
What are examples of where you are seeing data being represented in different ways?
One example is our address module which can process addresses in any format they are received. The module reads and interprets the address, places it into the correct columns, and validates it for accuracy. Additionally, we can handle any date format. Another example is our ability to read free text fields to identify details such as the cause of a loss.
Where has the London insurance market got to with defining and using common data standards? Do you still have to allow your clients to create customised outputs and reporting capabilities?
The London Market tried to standardise the way that bordereaux were sent and received by introducing Standard Version 5.2 as an output format. That has not become the de facto standard for sending and sharing delegated data. So, our platform is designed to ingest any format and output any format required, while also supporting existing standards.
Our clients have various needs, from proprietary catastrophe models to specific internal conduct reports. Some companies have their own internal database and may need a different output format. We help them manage all these different types of data easily.
Another feature our clients love is that our tools learn the coverholder mappings. Since coverholders typically submit bordereaux monthly, our platform identifies whether the data is in the same format as previous months, or if a new column has been added that needs to be looked at.
What flexibility do your clients have on how they use and access your tools?
Clients can opt to use just the data cleansing part of the platform and send their data via API to a different downstream system; or they can opt for the ScrubHub; or they can opt for both and send the clean data to both downstream systems and the ScrubHub.
Because we take an AI-first approach, the process is extremely fast and scalable. We are not relying on large teams of people sitting in a room cleaning data manually. It’s the technology that’s doing the cleaning. Our clients tell us that they like the fact that it is very flexible, allowing them to build their own output formats, while also appreciating that the User Interface (UI) is highly adaptable. Unlike some other proprietary tools that companies have been using, our clients are not tied to predefined mappings and standards set by others.
Another feature people appreciate is that we are also able to clean, validate and transform risk bordereaux for the catastrophe modelling process. This can be done by using our Underwriting Platform which supports a range of applications.
How will your solution help the insurance industry in the future?
Our goal is to give people confidence and trust in their bordereaux data. Until now, it has not been possible to accurately aggregate data, and with the tools currently available, it is often impossible to reliably report on it. We are unlocking the ability to trust and accurately report on data, empowering users regardless of the tools they use alongside our products.
Delegated business makes up 40% of Lloyd’s premiums, which is a massive amount, and insurers are giving their underwriting pen away to others. Therefore, they need to be able to ensure these companies are conducting themselves appropriately and preserving their reputation. For example, is a claim being paid on time? Are premiums being received on time? We are ensuring these insights can be achieved accurately and efficiently.
Who are you looking to connect with?
We are looking to connect with insurance carriers, MGAs and brokers who manage delegated authority bordereaux data and are looking to improve its accuracy and consistency. Whether it’s cleaning risk, claims and premium bordereaux or ensuring data is structured correctly for your binder management systems, we can help streamline the process.
If this is relevant to you, feel free to reach out via LinkedIn, our website, or email us at [email protected]
