Report from one of the sessions at the RSS 2017 Conference. More reports of conference sessions are listed here.
Wednesday morning’s session at the RSS conference was jointly organised by Babcock International Group and the RSS Business & Industrial Section (BIS) and chaired by Neil Spencer from BIS. Three industry-focused presentations and case studies explored the breadth and depth of applied statistics used across engineering, commerce and technology. Aimed at those with an interest in the practical uses of statistics, the session was well-attended, lively and engaging, with thoughtful questions being posed by the audience.
Dr Sophie Carr of Bays Consulting started us on a journey around the British Isles with no traditional navigation tools available, using Bayesian approaches to fuse data sources on the marine environment. Working out where you are, where you're heading and how you get to where you want to be is not always easy when data is sparse and incomplete. Bayesian approaches offer a way of efficiently and effectively combining multiple data sources whilst also being clear about the level of error and uncertainty incorporated with the overall result. A case study showed how an engineering system could be built around tight constraints and incorporate efficient algorithms to enable rapid updating and presentation of results.
Misti Paul from Advanced Analytics gave a presentation focused on conjoint analysis with its role in new product development, pricing and asset management. She showed how experimental design and analysis allowed businesses to not only determine the value of current assets but to optimise new product development, combine product offerings and evaluate current products. She also showed the impact these tools can have on focusing customer preferences to efficiently maximise the value in collected data, through survey questions and generation of utility scores. This will allow companies to not only review market pricing, customer value and product offerings but to avoid cannibalisation of existing offers.
Finishing off the session, Dr Nira Chamberlain from Babcock gave a presentation on scientific simulation of key performance indicator (KPI) objectives covering maintenance, care and availability for complex physical assets across their lifecycles. Often, attention is focused on the building cost of complex assets such as a new power station, aircraft carrier or railway, when the through-life performance is equally if not more important. Where assets are being supplied to a customer, KPI targets may be both non-negotiable and not based on any scientific analysis, with a dearth of data behind decisions of huge impact. Two case studies demonstrated the power of simulation to optimise support strategies to achieve these KPIs. For a global communications system, the development of a dynamic travelling repairman had reduced the cost of achieving the KPI by 50%. A fleet management the importance of simulating the dominoes effect - how even a small delay can impact a customer KPI target even when there is only a single factor outwith the control of the provider.
What is clear from all three industry presentations is the inevitable meeting of theory and real data. How can statistics and applied mathematics help decision makers understand the results being obtained and what constitutes 'good' performance? What happens when data is incomplete or not available? In both cases, clear communication between all involved is essential. However, the burden lies with the analysts to clearly show when measures of effectiveness and KPIs may need revising. The ability to explain the underpinning method and impact of the results is as crucial as the use of an appropriate technique.
This session was titled Exploiting data to manage complex assets.