Move beyond ERP and use data to get supplies delivered where they're most needed
Public sector services and humanitarian organisations must take inspiration from forecasting methods used by big business, says Dr Bahman Rostami-Tabar, assistant professor in management science at Cardiff Business School.
Forecasting for social good is the phrase coined by Rostami-Tabar, forecasting expert who is keen to promote forecasting-led initiatives that “solve real world problems, but are not primarily driven by profit”, particularly important when life or death decisions hang in the balance. Rostami-Tabar has been working with UK healthcare providers, and also the International Committee of the Red Cross to understand the unique challenges they face, and gives the example of storing medical goods.
There are warehouses in the 80+ countries in which the Red Cross has delegations, which teams have to keep adequately stocked with medical and non-medical items – some of which have a short shelf-life, but are nevertheless vital in medical emergencies.
Overstocked perishable items that reach the end of their shelf-life have to be given away or incinerated. But, aside from the waste of valuable resources within stretched budgets, there is a greater danger of running out of those items – particularly when there are long lead times to delivery – could be catastrophic.
“Worst case scenario, someone may die,” Rostami-Tabar says. Introducing more robust forecasting methods should help organisations such as the Red Cross, as they need a more sophisticated strategy than commercial enterprise resource planning (ERP) software currently offers. ERP, he argues, does not take into account intermittent demand that each warehouse will face: it is unable to predict war, famine or other humanitarian crises in each of the locations. “ERP forecasting tools tend to forecast on the basis of demand in the previous time period,” he says.
“Having the forecast and applying the social good element means going beyond the data. It also relies on the experience of the people involved,” he adds.
Rostami-Tabar sets out three elements for forecasting for social good. In addition to assessing the historical demand for an item, formulae should also include probabilities calculations for events such as epidemics, or sudden population changes. The third element is “judgemental adjustment”, where people use professional expertise to adjust figures based on information outside.
While judgemental adjustment is common practice for many organisations, Rostami-Tabar suggests people make a critical omission when they do this. “They need to provide the reason they are adjusting the number – and record it. They will need to come back to it later to improve their decision making,” he adds.
hidden data can unlock value
In the UK, Rostami-Tabar is working with health organisations, including the Welsh Ambulance Service. He says the service has a rich amount of, as yet, under-utilised data, going back eight years, including types of calls and location of incidents. One of his projects is looking at demand for the service over the winter months, so that it can better match demand with capacity availability, organising its resources such as vehicles and staff rotas accordingly.
Rostami-Tabar also conducted research in 2018 to predict the impact of public holidays and festive days on the number of daily A&E visits in University Hospital of Wales in Cardiff. As a result, he has developed a robust model, applicable to hospitals and health services, that more accurately forecasts the impact of festivals (such as Bonfire Night) and major sporting events on A&E demand, helping departments allocate resources more effectively.
Much of the research in forecasting for social good relies on secondary data – information that has already been collected, and held in a database on an organisation’s IT system.
Stick to the important points
Rostami-Tabar stresses that researchers working in this field need to be selective about the data they use. They must work collaboratively with an expert from the organisation to make sure only relevant information is extracted. “The type of data we use depends on the question we’re trying to answer, or the decision we want to support,” he says, “organisations record a huge number of variables, but only a few are relevant to solve a particular problem.”
Jane Lynch, senior lecturer for logistics and operations management at Cardiff Business School, warns that loose categorisation often affects the quality of data many organisations hold. “Often a variety of terms are being used for the same thing. It’s not just a case of making sure the number is accurate,” she says.
Whilst it’s important that organisations effectively identify and utilise their secondary data, Rostami-Tabar adds that forecasting can be enhanced by primary data, such as information gained through interviews, opinion polls, focus groups or questionnaires. These are techniques increasingly used by retailers, which Rostami-Tabar says have much more advanced forecasting models.
But social forecasting by itself is not useful, stresses Rostami-Tabar. “It’s important to link the forecast to decisions, seeing how making a more accurate forecast can improve things such as waiting times, or to reduce the pressure on nursing agencies,” he says.
“There must be a collaborative effort from experts in the organisation. If they are not involved in the whole process, it is almost impossible to feed findings into practical decisions.”
Lynch adds: “From a procurement perspective, it’s important to think, what does forecasting give us? Are we throwing money at the wrong areas? It helps us categorise on spend, and think about the impact of that spend.”
Rostami-Tabar has observed that some UK hospitals lack a coherent centralised procurement strategy. For example, critical items are ordered by housekeepers of wards, and orders for the same items can be duplicated by different housekeepers coming on and off shift. “It’s a method that leads to a lot of errors and waste,” he says.
Lynch agrees that a more centralised procurement strategy would be leaner, but adds that regional and departmental staff should also have input from their expert knowledge and experience.
And while Brexit is hard to forecast, social forecasting can help. “The best thing to do is to have a contingency plan. Collaboration is also vital – we’re in unprecedented territory,” she says.
Forecasting for social good
• Complicated does not always mean better – choose the simplest forecasting model that is capable of solving your problem.
• Make sure your software is equipped with an up-to-date forecasting toolbox.
• There is no magic single forecasting model that works for all possible situations. You may need to tune your forecasting method to deal with different scenarios.
• Categorise and prioritise your items based on criteria that are important for your organisation: for those items that are not critical, you can automate your forecasting procedure, whereas in the case of critical items, you should continuously monitor the accuracy of your forecasts.
• Judge the accuracy of forecasting models using criteria that matter to your organisation, and are not based solely on cost or profit.
The right data
• Only capture the data that you need to solve the problem or support the decision.
• Make the collection of accurate data a strategic priority for your organisation.
• Incorporate an automatic procedure to control the quality of data. Communicate with the data provider in case of suspected quality issues.
• Set standards for recording data, ensuring buy-in from all parts of the organisation. For example, agreeing which fields are mandatory.