I spent an informative hour with Veriluma’s chief executive, Elizabeth Whitelock, its data scientist Stuart Nettleton and Damon Jalili, an independent government relations specialist and formerly head of government relations for Kreab Gavin Anderson.
The premise was simple. The “old” methods of data analysis, namely descriptive (past performance) and predictive (what could happen) are data-based whereas prescriptive uses all sorts of data, fed into a user defined model (that is the crux to its accuracy) to anticipate the likelihood of what will happen. In developing the model, inherent biases and gut feels are identified and, if necessary, ignored.
Prescriptive has been called the final frontier of business analytics and works from the “model” backwards. What question are you trying to answer, what is the information or indicators you need to know to answer it, and what are the “weird, unknowable, left field things that could skew the results? Veriluma has a patented collaborative decision-making “engine” and application programming interface (API) that allows users to run up a model, refine and use it very quickly on x86 notebooks or in the AWS cloud.
Prescriptive can use all inputs and can use separate input loops comprising data to feed results into other loops – relationships, etc.
Veriluma started in a Commonwealth Research Centre for the Australian Defence Intelligence Organisation, working on defence models. Defence situations can lack certainty and reliable information and may depend on input that is subjective or even inaccurate. Regardless, in these situations, decision makers still demand fast insight that is accurate and actionable.
After the government got what it wanted it put the technology up for commercialisation and Veriluma was born in 2010. It was floated on the ASX in late September 2016. Its software has wide applications across multiple industries including banking and financial services, government, defence and national security, insurance, legal, health, pharmaceutical and resources.
During the Q&A session, we discussed many uses – Fintech to assess credit risk, financial planning, mergers and acquisitions, policy regulators and more. At that stage, it became clear that it is not really about the data but about what question you need to be answered.
Nettleton was enthusiastic, “Using predictive analysis should move you from either being right or wrong – 50/50 – to mostly right all of the time. The world is not perfect, and the starting at the question and working backwards tells you more than starting at the machine learning insights and trying to figure out what to do. It considers all factors without emotion, bias, and prejudice.”
Jalili stated that as data increases, decision makers are increasingly swamped by it and often paralysed. Prescriptive says what is the question you want to be solved – not here is the data, analyse it for insights – insights are fine, but they are not the best for making decisions.
In a government sense, prescriptive mandates that all biases are exposed and there is transparency in the decision.
Whitelock spoke of the expansion plans, as far as a listed company can. “As a small company, we sold directly to the end user. Now we are looking to find, train and retain a partner network that can apply their expertise to specific geographical or vertical markets. You don’t need to be a data scientist to use this."