CbusineZ invests in machine learning to improve ambulance response times
YieldDD was asked to take an in-depth look at the Stokhos software.
A Dutch ambulance is supposed to arrive at the scene of an emergency within 15 minutes after an emergency call. Around 35,000 times a year however, or in 7% of all instances, ambulances arrive too late and may not be able to provide proper care in time. This can lead to severe adverse health effects for the patients in question. Stokhos, a start-up company with roots in Delft University of Technology, has developed a particularly smart way to improve ambulance response times and was therefore an attractive opportunity for CZ's venture capital arm CbusineZ. As part of the investment process, CbusineZ asked YieldDD to take an in-depth look at the Stokhos software.
Calculating optimum ambulance coverage
A straightforward way to reduce ambulance arrival times would be to deploy more ambulances, but because of budgetary constraints and a scarcity of qualified medical personnel this is hardly feasible. The key then is to deploy the existing fleet of ambulances more efficiently. Stokhos developed a machine learning algorithm that, based on historical emergency data, current traffic conditions, the whereabouts of other ambulances in the vicinity and a range of other parameters, continuously calculates the optimum coverage of ambulance within a region.
Counterintuitive but very promising
Instead of ambulances waiting for the next emergency call at a fixed location, i.e. the hospital, the algorithm now proactively dispatches them to locations closer to where a next emergency is likely to occur. Although driving an ambulance to a seemingly random location where nothing has happened may initially be counterintuitive for the dispatchers and ambulance crews involved, the first results are very promising. The number of late arrivals can already be reduced to 5% (down from 7%) and possibly even further to around 3 or 4%.
Good programmers and good software engineers
Developing a smart algorithm that achieves such impressive results is one thing. Embedding it in a software environment that turns it into a practical solution ready for day-to-day use and that can be scaled-up and also cover additional regions, that is secure from hackers and free from potential IP infringements, is quite another. "Data science and machine learning specialists may be good programmers, but they're not necessarily good software engineers too", as YieldDD's Marco van Os puts it. But in this case the Stokhos solution not only performed well in terms of functionality, is also proved to be well designed from a business and investment perspective. So as far as software quality was concerned, YieldDD could give the green light to CbusineZ to go ahead with their investment plans.
Working with YieldDD on this deal showed us how valuable software due diligence can be, not only for target companies but for all our portfolio companies.
Capturing enormous value
In the process, CbusineZ came to recognize the added value of software due diligence. Together with YieldDD they developed an additional software and IT module for the evaluation method they use for the annual assessment of all their portfolio companies. "This is exactly the kind of relationship we like to develop with our clients", says Marco van Os. "That is because we're convinced we can help them capture enormous value by doing software due diligence on a regular basis, instead of as a primarily deal-driven activity."