Why we invested in Diffblue?
News, by Kibriya Rahman
Why We Invested in Diffblue
The Problem: Code Testing and Code Health
Software was predicted to eat the world, and suffice to say, it has. We’ve seen software become critical to company operations, products and services in every industry; and we are still seeing new developments and further complexity. This has led to huge volumes of code being written. Whilst there is no accurate estimate for the total number of lines of code in the world (it is likely to be in the trillions), it will only continue to grow.
But all this code needs regular testing for reasons of efficiency, agility, and quality control. It helps to find and fix problems in the code early, resulting in the delivery of better-quality software more quickly. Testing is required for code maintenance, and it is also important with the increasing shift to agile development where development and testing happen concurrently.
However, there are already not enough software developers available to write all of the necessary code (the global shortage will reach 4m by 2025), and testing code is time-consuming and manual. Currently, businesses are forced to trade off the amount of code they build with the amount of testing they do. And where code quality is prioritised, this can mean that developers can spend up to half their time on testing! Not the best use of this scarce, highly valuable resource – but necessary where code is safety or business-critical.
This can also lead to code targets not being met and quality being sacrificed. For example, 43% of developers who work in financial services in the US and UK have found it difficult to meet the unit test coverage targets; 90% of software developers in financial services believe better testing is needed to improve software quality; and 49% stated that their company suffers from increased costs for fixing errors as a result of poor software quality.
The Solution: Diffblue – Autonomous AI-for-Code Software
Diffblue is the leading pioneer of software creation through the power of AI. Founded by researchers from the University of Oxford, its flagship product ‘Diffblue Cover’ uses AI for code to solve the problem of effective unit testing (unit testing is one of the fastest ways to detect code errors early on). Capable of writing unit tests 250x faster than a human developer, it helps software teams improve code quality, expand test coverage and increase productivity. Therefore, they can ship software faster, more frequently and with fewer defects.
Diffblue Cover analyses Java, one of the most popular languages for writing code, and writes unit regression tests that reflect the current behaviour of the code. Diffblue Cover can write a unit test in about 2.5 seconds, making it an ideal partner for developers looking to get meaningful unit test coverage quickly. They can focus on writing the key unit tests that verify the intent of their code, while leadership can see the overall picture and address risks and gaps in coverage.
Their customer base already includes leading names across Financial Services, Telecommunications and the Information Services (incl. J.P.Morgan, Citi and S&P Global); who have seen their testing processes fully automated and reduced from months to hours, with accuracy rates of up to 100%.
Why Diffblue: What We Are Excited About
- The market is large and growing. The software testing market is currently worth $40bn and is forecasted to reach $60bn by 2027. Growth is being driven by underlying secular trends in software development, such as digital transformation, cloud computing, cybersecurity, automation and artificial intelligence/machine learning. Enterprises want to make their software development more efficient, agile, and quality-assured; as well as help developers to focus more on developing software to keep them engaged, motivated and creating value.
- The technology is unique and differentiated. The underlying AI approach behind Diffblue Cover is called reinforcement learning (the same machine learning strategy that powered AlphaGo, DeepMind’s software program that beat the world champion player of Go!). This is unique in the market and superior in many ways: its unsupervised learning model requires no pre-training; it has accuracy rates of up to 100%; it requires low CPU and memory requirements; and it is fully automated and autonomous.
- Customers love the product. Diffblue is able to automate some of the most manual work for developers: Diffblue Cover can write a year’s worth of unit tests in 8 hours, at the click of a button. Hence the developer community rates the product very highly. Development teams find it improves the velocity and quality of the software delivered, and frees up time to focus on creating new features. Meanwhile, DevOps teams find it helps catch errors and issues earlier, and improves deployment frequency, lead time, and mean time to repair.
- A fantastic team and company culture. Diffblue is led by CEO Mathew Lodge, who has over 25 years of cross-functional experience building and growing software companies across the U.S. and European markets. The rest of the team have unrivalled domain expertise, product knowledge and customer centricity; which were repeatedly praised by clients in our references. Find out more about the company culture and open roles here.
Our investment in Diffblue reflects AlbionVC’s focus on building category-leading software companies that are leveraging new technologies and machine learning to disrupt sectors crying out for change. It follows our other investments into AI/ML companies, including Black Swan Data, Credit Kudos, Elliptic, Hazy, Imandra, Phrasee, Quantexa, Seldon and Speechmatics.