Imperial Innovations — Update 6 September 2016

We last wrote on IVO’s ICT & digital portfolio in March 2015. Since then, IVO has been actively investing in this area – it has added six companies to the portfolio for a total investment of £12m as well as making follow-on investments in existing portfolio companies.

The table below shows the movements in the overall IVO portfolio from the end of FY14 to the end of H116. Where disclosed, we also show the percentage of the overall IVO portfolio made up by ICT & digital companies and their contribution to the top 20 or top 10 companies in the portfolio.

In January, IVO announced it had signed an agreement with the UCL Technology Fund. This provides formal access to IP developed by UCL researchers and is the first investment fund created by UCL to commercialise its research. The fund is expected to invest £50m over the next five years to support ideas from academics in life sciences and physical sciences. It will be used to fund early-stage proof of concept, licensing opportunities and new spin-out companies. As a limited partner in the fund, IVO has committed £24.75m over a five-year period; this will be matched by the European Investment Fund (EIF).

IVO added one company to the ICT & digital portfolio in each of FY14 and FY15, but during the course of H116 the pace of investment accelerated, with five new companies added for a total investment of £9.0m. We summarise the investments in Exhibit 2.

We summarise the follow-on funding rounds in which IVO participated over the last 18 months in Exhibit 3.

In the next section, we revisit Featurespace, one of the ICT & digital portfolio’s larger investments.

Featurespace uses proprietary software to analyse the behaviour of people in a commercial and social environment. It can be used in a wide variety of applications but is focused on predicting, identifying and preventing fraud in the financial services and gaming industries. In a case study from Betfair, use of Featurespace technology resulted in a 50% reduction in the number of staff conducting manual assessments and a 77% reduction in false positives. The combination of better fraud detection, lower costs to service customers and improved revenues due to fewer false positives makes the company’s software an attractive alternative to widely used rules-based solutions. Featurespace has c 30 customers and is on track to grow revenue by more than 150% this year.

Featurespace was formed in 2005 by Dave Excell (CTO) and Cambridge professor Bill Fitzgerald as a university-based consultancy. They were approached by Betfair to solve a problem it was having with identifying fraudulent transactions. The deal was agreed in 2008 and Featurespace built its first on-premise Fraud Manager system. Using their novel Bayesian inference methodology, the Featurespace co-founders correctly identified 90% of fraudulent transactions in a sample of data provided by Betfair. As a result, Betfair became Featurespace’s first customer. It has since expanded to c 30 customers over a variety of segments, and anticipates more than 150% growth in revenues and 210% growth in bookings this year.

The core software engine is branded ARIC (adaptive real-time individual change identification) and has been developed into three products: ARIC Fraud Manager, ARIC Responsible Gambling and ARIC Accept. Each of these products has significant opportunity but Fraud Manager is the current focus. Revenue is based on a standard software licensing model with additional support and maintenance revenue. There are currently 52 members of staff, with engineers and data scientists making up more than half of the total.

Martina King (CEO) was previously managing director of Capital Radio and Yahoo Europe and is non-executive director of Cineworld and Debenhams. Gordon Hurst, previously the FD of Capita, is Chairman and Dr Mike Lynch (founder of Autonomy) is also a non-executive director.

The ARIC core engine uses a combination of Bayesian inference, signal processing and machine learning to derive meaning from a variety of data sources. Each install is capable of processing over 120m events each day in real time, with a millisecond response, and adapts over time. It means that rather than categorising clients based on a manually specified set of criteria or rules, the engine learns the typical pattern of behaviour of a good customer and with its anomaly detection algorithms can detect fraudulent transactions automatically. This principle can be applied to multiple areas but fraud management in financial services and gambling are the main areas of focus. The software is designed so that other sources of risk management data (eg third-party ID checks) can be fed into the system to increase accuracy. The software can be implemented on-premise, in the cloud or hosted by Featurespace. The set-up time can vary significantly by client and can range from weeks to months, depending on how quickly the customer can develop processes around the new technology. Most customers choose to implement the software on-premise.

The cost of fraud in the financial services industries has several sources: running fraud detection and prevention services, reimbursing customers who are the victims of fraud and the revenues lost from false positives. Rules-based analysis is the most common method of fraud detection, but it tends to result in a high level of false positives, ie, the system prevents legitimate transactions being processed as it concludes that the transaction is fraudulent.

In the financial services market, Featurespace attempts to identify fraudulent transactions passing through a bank or payment processor’s systems by learning the behaviours of each customer and all customers simultaneously in real time and spotting unusual activities. Getting the balance right between false positives and false negatives is crucial in terms of cost, reputation and customer satisfaction. According to management, Featurespace’s adaptive behavioural analytics provides a more optimal trade-off than other currently available systems so there is a clear positive return on investment that Featurespace should be able to use to win market share.

Research carried out by Oakhall Advisors1 suggests that the annual cost of card fraud globally is c $16bn (ie the value of fraud losses). In addition, false positives drive at least another $14bn in costs: c $4bn in lost card fee income and more than $10bn of costs incurred in dealing with customer queries/complaints arising from having transactions blocked. Data provided by Featurespace to Oakhall shows that its software is capable of reducing fraudulent transactions by 25% (ie potentially worth $4bn across the industry), while at the same time reducing the incidence of false positives by 70% (worth nearly $3bn) and reducing customer-service related costs by 50% (worth more than $5bn). Overall, use of Featurepace across the card industry could cut fraud-related costs by as much as 40%.

Rules-based analysis is good at detecting fraud from known methods. However, when a new way of committing fraud emerges, it will take time for this to be identified and for the rules to be updated. Featurespace’s software is more quickly able to identify new sources of fraud by understanding normal behaviour and thereby spotting anomalies.

The main competitor in financial fraud prevention is $3.7bn market cap, US listed, Fair Isaac Corp (FICO), which provides a range of services including credit rating, fraud protection, data management and marketing & customer engagement services. IBM and SAS also offer fraud solutions to this industry.

Featurespace has built and deployed a fraud detection system for Vocalink-owned mobile payment platform, Zapp. This year, Featurespace signed a contract with Total System Services (TSYS). TSYS is a US-listed payments company with annual revenues of $2.8bn from payment processing (for c 650,000 merchants) and card issuing services (for more than 400 customers). TSYS will use Featurespace’s ARIC Fraud Manager software to help its clients reduce fraud. We view TSYS as a crucial reference customer – success at TSYS should pave the way for new financial services customers. More recently, Featurespace has signed up a top 20 North American bank to use its fraud solution.

Other customers include a short-term loan provider in the UK (where Featurespace has enabled that client to lower its risk while improving performance), a global credit card issuer with more than one million UK customers, and a UK credit reference agency (where Featurespace software is used to identify criminal behaviour at the early application stage).

Fraud is also a big problem for gaming companies and Featurespace can help by identifying the key characteristics of fraud and predicting when it might occur before it actually happens. It can also help reduce customer churn and encourage responsible gambling by providing a player health check across all players on a platform. Fraud identification in gaming companies has the same false positive versus false negative trade-off as the banking industry and therefore the analytics tool that can optimise this trade-off can generate substantial value and command a pricing premium.

The Betfair case study is a good example of how Featurespace can generate value for gaming companies. Before using Featurespace, Betfair used a manually defined rule-based system and score cards to identify customers that might commit fraud. This, however, meant that unknown fraud types would not be identified. Featurespace used data such as bets placed, cash deposited and withdrawn from accounts and the sporting calendar to independently and automatically identify potentially fraudulent clients. This resulted in a 54% reduction in chargebacks, a 50% reduction in manual review by analysts, and a 77% reduction in false positives. Betfair now claims to operate with one of the lowest fraud rates in the gaming industry and can scale its operations without having to expand its fraud analyst team significantly.

Last year, William Hill signed up to use Featurespace to interpret data and achieve greater insight. and Camelot signed up for a six-month trial to identify problem gamblers. In July, OpenBet signed up to use Featurespace for fraud, risk and compliance management. Also in July, Kyte Consulting became a reselling partner, with a focus on gaming operations in Malta, Europe, the Middle East and Africa.

In the insurance sector, Featurespace not only helps prevent fraud but also optimises insurance pricing by analysing and attempting to predict customer behaviour. Pricing of insurance cover is critical as if it is set too low the insurer will make a loss and too high they will lose customers in what is a very competitive market. The insurer that can therefore correctly analyse the likelihood of claim will be able to offer the most competitive prices. Therefore, similar to banking, there is a clear positive return to the customer. The software is also able to help in the fight to identify “ghost brokers”. These are brokers who provide policies that are bought from legitimate insurance companies using false information and then doctored before being sold on to customers, or provide fake policy documents that are designed to look like they have been issued by legitimate insurance brokers.

Competitors in the insurance space include some small players such as Earnix (UK-based private company) and larger consultancies such as Infosys, SAS and IBM.

We summarise below the key milestones achieved over the last two years and assess the growth prospects of the company.

Milestones completed

With a large financial services industry, North America is an attractive target market for Featurespace – in 2015, the company decided to expand into the US and opened an office in North Carolina. The recently signed TSYS and North American bank contracts highlight the strong start the company has made with its US expansion strategy and are a strong endorsement of Featurespace’s technology. Successful roll-out of these contracts should pave the way for additional new business from the financial services sector, in addition to the company’s growing penetration of the gaming market. Based on bookings received to date and the quantum of business in the pipeline, the company has growing visibility on revenues for 2016 and beyond.