The optimisation of rosters to meet public demand Presentation to the Criminal Justice SIG 18 November 2015 David Wrigley david.wrigley@orvis-consulting.co.uk
Structure Scope of work Summary Results Benefits Lessons learned Questions
The Client Public sector organisation operating 24 / 7 Multiple sites Workload is driven by public activity outside of client control, varying moment by moment Staff come from several groups, each with own skills base and contractual terms Moving towards unification of contracts and cross-training
Study requirement Recommending approaches/methods for optimising the rosters … … taking account of staff wellbeing, operational efficiency, flexibility/robustness to disruption, customer service Putting forward options for software tools to aid rostering which may include automating part, or all of the rostering process Demonstrating compliance with working rules and regulations ….. including types of contract and availability for duties
Study tasks Scoping study Software trial May 2014 – October 2014 December 2014 – September 2015
Scoping study Three major sites selected as case studies Interviews with staff, managers, rostering teams, unions Developed measures of merit of a roster Brief review of off-the-shelf software Review of roster optimisation methods Workshop with senior managers and rostering staff
Summary findings from Scoping Study Some staff are exhausted by current rosters Proposed set of Measures of Merit of a roster Baseline performance Meta heuristics is the most appropriate optimisation technique Software is available off-the-shelf Use MoM as benefit function for optimisation Demonstration of example software tool Outline for new rostering approach
Software Trial One site Using existing time management software Catalogue of rostering rules, principles and constraints Data collection and cleansing Exploration of software and dry runs Trials against live rostering using new process Use MoM derived in Scoping Study phase to compare rosters
Working with a Black Box Cannot be certain what it is actually doing Technical contact with supplier … Dry runs ….mislead myself about its capabilities and workings A lot of time spent trying to understand it and control it “It does what it says on the tin”
Rostering targets Manual rostering against minimum staffing target per shift
What we were trying to achieve Effectiveness: reduce under-coverage Efficiency: reduce over-coverage
Metrics Efficiency – over coverage Effectiveness – under coverage Work life balance Compliance with HSE Guidelines for rosters including night shifts Plus an overall Figure of Merit combining these
Rostering against minimum staffing target per shift Rosters assessed against minimum staffing requirement per shift
Against time varying staffing requirement Rosters assessed against time varying staffing requirement Manual rostering against minimum staffing target per shift Automated rostering against time varying staff requirement
Automated shift design Shifts generated automatically given wide degree of flexibility in shift length and start time
Adjusted standard shifts Rosters assessed against time varying staffing requirement
Potential benefits Improved efficiency of staff allocation Reduced staff effort to construct rosters Improved staff well being through improved work life balance Faster response to changing circumstances
Rostering is difficult Once you build in a complex set of contractual rules, the working time directive, legislation, trade union agreements and various formal and informal arrangement …. …it is often difficult to find a staff member to fill a shift Automation is the best way to tackle this … … but there is no guarantee of a perfect solution
Automation is possible and beneficial They said it couldn’t be done We tried it and it worked And we got better results against the selected metrics on all counts This was the simple problem: one duty at one location Confident that it will work even better for multiple duties at multiple locations
What would I do differently … Ask the obvious (?) questions, even though we all think we know the answer Push the question of data transfer more strongly and get a definitive answer before starting work Which means: doing the risk analysis more rigorously And: try a different software approach as a comparator
Questions?