Big data, big opportunities

What does the future of modelling hold? Can big data support basic decision-making? How can we model connected travellers? How will human roles evolve? Systras leading experts consider key data and modelling issues

We live in an increasingly multi-modal world. Travel patterns become more complex and less predictable year on year. Big data provides more information about travel patterns and network conditions in more detail and with more accuracy than ever before, and clearly offers the opportunity to provide more powerful and more predictive models

We live in an increasingly multi-modal world. Travel patterns become more complex and less predictable year on year. Big data provides more information about travel patterns and network conditions in more detail and with more accuracy than ever before, and clearly offers the opportunity to provide more powerful and more predictive models

When it comes to getting from A to B, we live in an increasingly multi-modal world. Travel patterns become more complex and less predictable year on year as apps and mobile technologies provide information to travellers that enable and encourage changes in behaviour, changes which in turn incentivise the creation of new technologies in a dizzying feedback loop. Keeping up with the changes can seem daunting enough, let alone anticipating what comes next. As the legendary baseball coach Yogi Berra once said: ‘It’s tough to make predictions, especially about the future.’

Yet great challenges bring great opportunities. Every time you put your GPS-connected mobile phone in your pocket before stepping out of the door, you join millions of people who are inadvertently creating a precise record of their journey, a record that will become one more drop in an enormous and growing ocean of data waiting to be analysed (or ‘mined’) for patterns. Add to this the data from SatNavs, walking and cycling apps, bus and goods vehicle tracking, automatic number plate recognition data; public transport smart-card use, ‘intelligent’ traffic signal systems and automatic traffic counters, and you have ‘big data’. 

Big Data provides more information about travel patterns and network conditions in more detail and with more accuracy than ever before. Big Data clearly offers the opportunity to provide more powerful and more predictive models.  But if we simply respond to the new conditions with the old reflexes we could be missing a trick. If the data is good enough, fine- grained enough, perhaps it might replace the need for a model, at least in some cases. This is what David Connolly, Director of Innovation at Systra, suggests as a possible new trend that could provide quicker, more light-footed analyses of transport needs:

‘The point of models is appraisal, and the near real-time data provided by mobile technologies can make basic decision-making appraisal possible without going to the expense of building a model, so long as you have the people who can handle really big datasets and who can distinguish the most important patterns from the noise. Or to look at it another way, sometimes the data could be your model.’

Testing the future

Of course, as David points out, data only gives us a picture of what is happening now, whereas a model allows us to alter parameters and test what would happen in the future if we changed this or that condition. In this sense, multi-modal models will remain essential in many circumstances, such as for testing multiple scenarios over time in a large urban area. But there are situations where the data set can provide a ‘ready reckoner’; an accurate (enough) appraisal that gives decision makers a solid basis for deciding if developing plans further is worth the work. Mobile device tracking and vehicle tracking collected from motorway traffic over a year, for example, can contain enough information about speed-flow relationships, congestion and variability across the network to estimate the benefits or disbenefits from changing these traffic flows.  As David puts it: ‘If the answer is ‘yes’ or ‘maybe’ then build and calibrate your detailed model, but if it’s ‘no’, save your money’. 

So a deep understanding of what big data offers could transform not just power and accuracy, but the fundamental ways in which transport modelling is approached. And the key to this understanding will lie with people and experience, not with systems. The importance of the human factor in transport modelling is easily lost in discussions of the predictive and analytical powers of technology, but it isn’t forgotten by Richard Hancox, Deputy Divisional Director of Transport Planning at Systra: ‘In more than 25 years of developing transport models, contributing to the comprehensive model for Greater Manchester that has helped transform that city’s infrastructure, and also to the London Transportation Studies (LTS) model, it has been the ability to match the model to the details of the requirement that has mattered more than anything. One size does not fit all. Models have to be very much bespoke to the urban area, the problems to be tackled and the range of potential solutions, and that is where our experience and knowledge counts.’ 

This emphasis on refining approaches, instead of simply ramping up the power, is shared by many colleagues in the modelling world. Ingrid Petrie, Principal Consultant for Systra in Glasgow, puts it like this: ‘You have to be careful not to get lost in data. You can get the idea that if you just get more data and plug it into a more powerful computer, the results will be even better, but my feeling is it’s not as simple as that. We have to use what we have in better ways. You can model all you like, but if you don’t have someone to interpret it, it means nothing.’

Ingrid reminds us that there is still room for the old-fashioned virtues of collecting information on the ground, still value in communicating directly and asking what transport users think, not just counting up what they do. For Ingrid, an organisation with a diverse workforce from a wide range of professional backgrounds is every bit as important as the technical tools and datasets that can be brought to bear on a problem: ‘I’m incredibly optimistic about big data and I love a model as much as the next person, but there is still an essential human element. Data is exciting but we have to use it right.’

Behaviour change

Making the best use of human resources will be what separates the good from the best in modelling in the coming years, no matter how much new data can be collected. As David Connolly points out, there are bodies of data that the mobile technologies cannot provide, and still many serious questions that current models do not seem well-placed to address. ‘Current models do not account well for how the behaviour of the connected traveller is affected by the technologies they use. For example, an app can provide real-time information about bus arrivals but we don’t know whether that will increase the use of buses as they become more predictable and demand less waiting, or if it will encourage a change in behaviour as passengers compensate for delays by walking more, leading to less overall bus use. The complexities are immense because feedback is complex and unpredictable. How information changes behaviour is a PhD subject in its own right.’

David identifies several other areas of darkness in the current models that will need a lot of thought, work and ingenuity to address, such as ‘cohort change’; the fact that the life experiences and social attitudes of the young will affect their future behaviour as adults. In many ways, we remain extremely bad at predicting. How will patterns of cycling change among this cohort, for example, and how will cycling be used in combination with other modes of transport? A glance at the poor predictions around the recent increases in bicycle use in London should give everybody pause for thought.

Electric and driverless cars are hot topics, but focusing on them can distract from the broader problem of how little modellers know about what vehicles, and how many, we will choose to own or otherwise access. How will policy affect the fleet as a whole? Will your future new electric car be an addition, or a replacement? How do ‘environmental’ apps like Uber or car clubs impact on vehicle ownership choices, not just driver behaviour?

As a well known US politician might have put it: for all the upheavals ahead, the ‘known knowns’ are still the bedrock of the business and we have a good idea of what is needed to meet the many challenges of the ‘known unknowns’ . As for the ‘unknown unknowns’, well, they will have to take care of themselves but, like the revolution that came in the wake of Apple’s iPhone (launched only eight years ago), the chances seem good that they will bring as many opportunities as headaches to the Modelling World.