Legal Utopia – a LawTech start-up making access to law and lawyers easier – collaborated with The Data Lab at The University of Edinburgh and The University of Aberdeen to use artificial intelligence technology to make comparing law firms easier for first time users of legal services.
The study found that, on existing data, the artificial intelligence results were unsatisfactory due to a severely imbalanced data landscape and, as a result, a prototype of the solution could not be released; however, the results from the Solicitors Regulation Authority produced data insights indicating a reduction in the total number of law firms operating between February 2021 and July 2021.
The reduction in law firms and increase in user data from Legal Utopia’s mobile app is likely to make a comparison solution possible for user testing and market release soon.
The project was commissioned, on application, by Legal Utopia and generously funded by The Data Lab with academic collaboration from the School of Natural and Computing Sciences, The University of Aberdeen.
The aims of the project were to calculate the weights of quality indicators (QIs) designated by Legal Utopia for scoring and evaluating all SRA-regulated legal services providers (LSPs) in the UK. Quality indicators included location, legal coverage, regulation history, and website availability as features that influence consumers into contacting a law firm.
The research included data from the regulator of solicitors – Solicitors Regulation Authority – which included a 6-month manual review by legal researchers at Legal Utopia of the regulator’s data. This included a comparison review of the data from the Solicitors Regulation Authority comparing their data from February 2021 and July 2021.
This was swiftly followed by a 3-month artificial intelligence modelling exercise to test different ways of training algorithms to predict comparable law firms based on prior consumers searching for law firms.
Law Firm Comparison AI
The study sought to employ classification machine learning algorithms to law firms regulated by the Solicitors Regulation Authority. The aim was to determine if one firm over, or in addition to, another could be “comparable” based on a consumer’s search preferences and behaviour on Legal Utopia’s Find-A-Lawyer service.
The historical user data of searching behaviour from the existing Find-A-Lawyer service distinguished cases where users contacted a law firm via email, phone, or their website. The artificial intelligence algorithms considered a range of information features of these “Contacted Firms” and those of “Not-Contacted Firms”, as well as the users’ searching behaviours and service usage data.
It then sought to use a computer-generated mathematical model to compute the relative influential weight of different information features including a law firm’s website availability, number of practice areas covered, location, and number of regulatory authorisation years.
The mathematical model determined that the Authorisation Duration of a law firm ranked as the most influential feature, this was followed by the number of legal practice areas covered by a firm, then the firm’s website availability as third most influential with the office location of the firm ranking the lowest of the four features in influencing users to contact a law firm.
The ranking influence of this information was used to “score” law firms and then use an AI algorithm to predict which law firms to present to a user of the Find-A-Lawyer service. This algorithm uses the “score” following the user’s filtering and searching preferences to then identify a refined selection of law firms to choose from.
The outcome would remove the confusion and frustration of identifying relevant law firms by automating the searching process of the 9,900+ law firms across the United Kingdom easing the friction of finding the right firm to contact.
A significant data review exercise was also undertaken to facilitate the AI modelling exercise, as well as to generate insights to help understand the existing data landscape and any market landscape changes occurring between February 2021 and July 2021.
The vast majority of data review was focused on the SRA Register’s data, however, a review of the BSB’s public access barristers’ register was also completed; although outside of scope of the AI modelling exercise. This was conducted to understand the availability and quality of data points that could be individually linked to each regulated LSP for the purposes of quantifying these data points, as well as to generate the below insights on the data landscape and market landscape.
- The data held by the SRA of all entities on the Register had approximately 350 entities with duplicate trading names across multiple entities with just over 300 having 2 duplicates and a maximum of 19 duplicates to any one trading name;
- The data held by the SRA and produced by its API had no entity geo-location data, no Google Business ID, no website security or status ID, no trading history value, no pricing value, no turnover values, no cybersecurity or insurance credentials values, and no SRA badge compliance status. There was no attribution of individual solicitors or lawyers that worked for or with any SRA entity;
- The data held by the SRA and produced by its API had no standardisation of the “Work Area”, “Freelance Basis”, or “Reserved Activities” as data points. All data points recorded on the SRA Register API data were also only mandated to be updated annually and only by actively regulated and authorised entities;
- The data held by the SRA of all entities, only 52% had a URL website link available within the dataset and 48% without, this does not, however, factor website quality, security, loading speed, design, or working status of which, on manual review, scored poorly in all fields;
- The study sought to quantify the importance of 10 quality indicators (Authorisation Duration, Cybersecurity Credentials, No. of Practice Fields, Customer Reviews, Cost, Location, Regulatory History, Website, No. Payment Options, and Insurance Coverage) identified by Legal Utopia, however, due to the lack of regulator engagement and poor data landscape, the study was conducted against only four quality indicators; and
- The study focused on the following as “quality indicators”: Authorisation Duration, Number of Practice Field(s), Location, and Website Availability.
- The SRA maintains a database of 23,434 entities, comprising the total number of rows of information accessible from the SRA API, of which, in February 2021, only 10,003 were actively regulated and authorised by the SRA;
- The SRA also maintains branch/office-based data on the 23,434 entities on their register, this includes, as February 2021, 30,849 rows of information accessible from the SRA API;
- The SRA entities that are actively regulated and authorised by the SRA reduced at a rate of 22.5 firms per month between February 2021 and July 2021 from 10,003 to 9,913;
- The SRA entities that are actively regulated and authorised by the SRA had almost 3,500 entities with 10-years authorisation history (number of years authorised by the SRA) as of February 2021, but this reduced by 300 firms by July 2021;
- The number of new actively regulated and authorised entities by the SRA has increased from fewer than 100 firms to almost 300 between February 2021 and July 2021;
- The actively regulated entities by the SRA have an unbalanced authorisation type with approximately 7,000 firms authorised as a ‘Recognised Body’ (with or without conditions) but only 2,000 as ‘Recognised Sole Practice’ (with or without conditions), and 1,000 as a ‘Licensed Body’. This went relatively unchanged between February 2021 and July 2021;
- The data held by the SRA, as of February 2021, showed that approximately 700 SRA entities have 2 branches/office locations with the highest number being 11 offices to a single entity, however, by July 2021 this when largely unchanged other than the prevalence of those with 11 offices reduced negligible; and
- Of all firms with more than 2 offices, none had 8 or 10 offices but the maximum number of offices to one entity to was 11.
You can access a full copy of the study from our website at here