Published December, 2021

Liberty Mutual Surety, a global business with bond capabilities in more than 60 countries, relies on access to timely and accurate financial statement data on public and private companies, mostly to analyze corporate financial health for bonding and to conduct benchmarking and trend analysis.  The financial data is also used by the claims department when evaluating a company’s viability and ability to fulfil their obligations.

The data that is created and stored in Liberty’s internal systems is normalized to meet the requirements of the various employees within the company that use it. “Normalization” is the process of structuring as-reported data in accordance with a set of norms, in this case norms set by Liberty, in order to reduce data redundancy and improve data integrity, so that all the data looks and reads the same way across records in a relational database. Normalization can also be described as aggregating facts reported by various companies into a single line item so that all the companies can be compared.

For example, a data user may just be interested in comparing the top-level line item “Accounts Payable and Accrued Liabilities, Current” for companies 1, 2, and 3. However, as shown in the figure below on the left, Company 1 reports a value called “Accounts Payable, Current.” Company 2 reports values for both “Accrued Liabilities, Current” and “Taxes Payable, Current.” Company 3 reports a value for “Accrued Liabilities, Current.” Because all of these values can be categorized as “Accounts Payable and Accrued Liabilities, Current,” normalization of this data would report each of these facts as the top level concept “Accounts Payable and Accrued Liabilities, Current” for companies 1, 2, and 3 as shown on the right.

Liberty’s custom normalization helps their internal teams hone in on areas of importance to Liberty’s underwriting process, Liberty’s normalization may be very different from a commercial data provider that collects financial data into proprietary databases and resells to investors. It may be different even from another surety that considers the data through their own tailored lens.

Historically, Liberty’s normalization process has been time-consuming and labor-intensive, involving manual entry of as-reported data, multiple checkpoints for accuracy, and the use of a system to aggregate data into the required standardized structure. The surety team at Liberty knew that there was a better way to manage this data by using standardized, highly granular eXtensible Business Reporting Language (XBRL) data to automate the process of extracting and normalizing the data needed.

Today’s process

Liberty Mutual Surety’s internally developed system – WorkBench – provides an electronic file and analysis system for everything pertaining to surety accounts, agents, principals, and brokers, globally.

To bring financial statement data into the WorkBench, it is typically manually rekeyed from a PDF, paper copy, or an online statement. In some non-U.S. markets, Liberty is able to download financial statements from various sources, but it is different in each country. Manual data entry and normalization takes 60 minutes on average. The as-reported data is then reviewed to resolve any keying errors.

The underwriter then analyzes the data, which may involve reading statement footnotes, reviewing other documents, or getting further explanation from the company account themselves. The complexity of the financials impacts the amount of time needed for the analysis which will involve making adjustments and eliminations, and checking various ratios, some of which is calculated in their internal system. This step can take from 30 minutes to several hours for each company. Additional reviews may be conducted. At every level of review, the data and analysis are checked for accuracy.

Automating Custom Data Collection

While “custom normalization or standardization” may sound like an oxymoron, it is not uncommon. Companies like Liberty have a particular way they need to view and analyze corporate data. Luckily, the highly granular nature of the XBRL standard makes it possible to break down financial facts to an atomic level, company by company. U.S. public companies have been reporting in this highly detailed fashion since 2009 (and foreign private issuers since 2017), so there is a lot of very structured, clearly defined data available. The granularity of the data means it can be machine-read and aggregated into a defined structure (normalized). Liberty is interested in collecting data from both private companies and public companies. Private company data is not yet available in a structured format in the U.S. today, but Liberty knew that capturing public company data could help them solve a key part of their data collection challenge.

By leveraging the open-source XULE Processing Language (XULE – https://xbrl.us/what-is-xule), and with some assistance from XBRL US, Liberty built an automated framework to extract company data exactly the way they need to view it. They started with companies that fall into the commercial and industrial (C&I) industry classification. C&I companies have a similar structure to their financial statements. Companies like real estate investment trusts (REIT), for example, have a different way of reporting and would need an adjusted framework.

Building the framework

Liberty collects data in their custom normalized structure for the three primary financial statements: balance sheet, income statement, and cash flow statement. Using XULE, they first mapped each Liberty line item, for example, Investing Cash Flows, Sale of Plant and Equipment, to the appropriate concepts in the 2021 US GAAP Financial Reporting Taxonomy. The figure below shows the concepts in the taxonomy that may be included in facts reported with the US GAAP concept “Proceeds from Sale of Property, Plant and Equipment” which equates to the line item in Liberty’s WorkBench, called Investing Cash Flows, Sale of Plant and Equipment.

One company may report only one of these facts, for example, “Proceeds from Sale of Buildings.” A second company may report both “Proceeds from Sale of Buildings” and “Proceeds from Sale of Water Systems.” A third company may report all seven of these items. In every case, XULE sums the facts reported by each company to create a single fact for each company called Investing Cash Flows, Sale of Plant and Equipment to fit the Liberty normalized framework. This exercise was performed on every line item that Liberty needed to capture from the balance sheet, income statement, and cash flow statement.

Second, once the initial mapping was complete, XULE was run in an automated fashion against every C&I company that reported their financials in 2021. This effectively produced new financial statements for each C&I company, structured in the Liberty normalization framework with data reported for the US GAAP concepts identified in the first step, summed into the appropriate Liberty line item. 

Finally, a manual review was conducted of the automated categorization performed by XULE to make further refinements. For example, the US GAAP Taxonomy allows companies to report “custom extensions” on their financials, creating their own concepts when they do not exist in the official taxonomy. During the manual review, these extensions were analyzed to determine where best to place them in the normalized structure. Company extension items were mapped to a US GAAP Taxonomy element so that they could be aggregated to top line categories automatically.

With this framework, Liberty pulls data into WorkBench, structured exactly the way they need it. Most importantly, this normalization process only needs to be conducted once, and the framework is used every quarter to pull in public company data. An annual maintenance process will review the framework in light of a new release of the US GAAP Taxonomy which the Financial Accounting Standards Board (FASB) typically produces to accommodate changes in accounting standards.

Collecting normalized data each quarter

With the framework complete, Liberty will collect financial data directly from the Securities and Exchange Commission (SEC). To identify when new filings are available in the SEC EDGAR system, Liberty will use the open-source XBRL API which will confirm new filing availability using the XBRL US database. Once identified, Liberty will pull the filing from the SEC EDGAR system and process it through the business rules established by XULE. Because Liberty will be using the SEC data directly, it will enable them to automatically ingest and categorize each financial statement into Liberty’s WorkBench within minutes of the public company reporting to the SEC in the manner Liberty analysts and underwriters need it, without manual processing.

Costs versus benefits of the automated system

The cost of establishing the normalization structure for Liberty involved time spent by underwriters, analysts, and IT staff, as well as project management, to keep the program on track. Analysts at Liberty reviewed and refined the mappings developed through the XULE processing to ensure that the categorizations were structured as needed, which required a modest amount of time on the part of the analyst team. IT staff were also involved in making adjustments to WorkBench and to the database to accommodate the additional data being brought in, update the mapping files, and perform some user-interface development needed. A proof of concept was conducted before the implementation began to confirm the viability of the project.

The Liberty team engaged XBRL US to prepare sample mappings for review, develop the XULE processing necessary to categorize the data, and refine the normalization structure.

The gains expected from Liberty’s switch to automated data collection, and away from manual data entry, will quickly outweigh the costs of standing up the new normalization framework. Liberty estimates a 150% return on investment in the first year alone. The most important benefit is the ability to redirect employee time towards higher value work at every level, for example, in underwriting, analysis, and identification of risk, rather than data entry. While the new method will save time and money (650% ROI over 5 yrs). Liberty’s most important asset is its skilled employees. This reset will increase staff productivity and allow them to focus on areas where they can provide the highest value.

Future plans

The Liberty team sees this as the first step. Looking ahead, they are interested in pulling in data for other types of public companies, such as financial services, real estate and insurance companies, which will require a somewhat different normalization framework. They also intend to begin ingesting financial statement footnote data, and to use artificial intelligence tools on the data to assist underwriters in identifying key risks with greater ease.

Liberty plans to automate data collection from foreign private issuers as well — companies that list on a U.S. exchange but are based outside of the U.S. Foreign private issuers have been filing their IFRS financials in XBRL format to the SEC since 2017. In addition to foreign private issuers, many other non-U.S. companies report in XBRL format to their local government regulators. In fact, the European Securities Markets Authorities (ESMA) has mandated that any company reporting to a European Union regulator must provide their financials in XBRL format. The ESMA roll-out has already begun and will pick up speed in 2022.

Liberty is also keenly interested in capturing private company financial data from contractors which of course, is key to the surety business. Private company data in non-US markets is often reported to tax regulators in XBRL format, and this is content that Liberty can begin collecting now. While today, U.S. private company data is not available in XBRL, Liberty can leverage the normalization framework for public companies to “XBRL-format” private company financials that they collect, in the same manner, for greater efficiency and ease of comparison.

The ability to mine the highly structured, machine-readable, and more timely data will also allow Liberty to better benchmark and prospect by analyzing data not directly related to existing accounts but that supports their work.

This initiative for Liberty is an extension of the work being conducted by the XBRL US Surety Working Group, which has been advocating to automate the process of collecting data reported by contractors, including the Work-in-Process (WIP) report and financial statements. Automation will improve efficiencies and allow sureties to better meet the needs of bond agents and contractors. That working group has developed financial data standards representing the WIP, and continues to engage with the contractor and regulator communities to encourage the adoption of data standards.

Liberty Mutual Surety, as a founding member of the XBRL US Surety Working Group, is leading the way in bringing automation into the surety industry.