AI-Optimised Pathways for Schedule Execution

Machine-learning from past programmes to improve schedule planning.

Last updated: 24th February 2021

Date uploaded:

Approved for use

Innovation Lead: Sherrie Rad
Project number: 104795 and 105877
UKRI funding: £641,118

Website:
nplan.io


Edit this story

Summary

Even experienced planners working on large-scale construction projects can inadvertently add their own bias or subjective intuition when they design project schedules. nPlan has used AI and machine-learning to process over 200,000 historic projects and capture data around the difference between planned schedules and the reality of delivery. These learnings were fed into the AI-Optimised Pathways for Schedule Execution - a platform that can improve accuracy in scheduling and give planners and clients greater certainty and confidence around project delivery. The platform was then tested out on a live project with Network Rail to demonstrate how past evidence can improve future delivery, and save time and money in the process.

Innovation type: Digital
Organisation type: Innovative SME

Story building blocks:

Story building blocks:

Learn more about this innovation below and use these building blocks to craft your own story to share.

Open allClose all

Project pioneers

nPlan had already been applying machine-learning to improve project management in construction. They recognised there needed to be greater certainty in the process and a scalable solution that could be used across the industry.

The problem

Currently 77% of all large-scale, complex construction projects are at least 40% late, and 98% suffer cost overruns of over 30%. These projects are heavily reliant on human input from planners, schedulers and managers. The planners are highly-skilled and experienced, but their experience can unintentionally add human bias or intuition to the reality of a project's delivery. This subjectivity could cost a project as much as 12% of its entire budget.

Vision

Machine-learning from hundreds of thousands of historical schedules will ensure past evidence informs future planning. The AI-Optimised Pathways for Schedule Execution essentially brings real-world, empirical data into planning decision-making. Experienced planners no longer need worry that their own biases could unduly impact a programme schedule, especially on complex construction projects where the volumes of data are large. By applying past learnings in this way, timings can be met, productivity increases, and costs come down. The benefits will be felt across the entire planning process and the wider industry improving transparency and increased certainty in the supply chain.

Key Insight

nPlan is a Software-as-a-Service SME start up that uses AI to map historic planning schedules against the reality of each project's delivery. It knew AI and machine-learning could add valuable real-world evidence to future programmes it just needed the vast historic data set from which it could put this to the test.

First step

The funding from the Transforming Construction Challenge then gave nPlan the opportunity to develop the accuracy of the software further. 200,000 historic schedules from Kier's database were made available to nPlan allowing it to train its AI algorithm and draw on the machine learning expertise of the team at Cambridge University. It then set out to test it on two live projects with Network Rail to investigate how it could scale and benefit the industry as a whole.

Barrier

The bigger and more complex the construction project, the harder it is for teams to untangle the interconnectivity and apply objective data to forecast project outcomes accurately. And even experienced planners can inadvertently let their intuition or biases influence programme schedules, leading to projects running over time and running over budget. Machine-learning from historical schedules provides more accurate data and can renew trust and certainty in the scheduling process, and will lead to significant savings for the entire value chain.

Digital Innovation

The AI model is based on over 200,000 previous construction projects, which is the largest dataset of schedules in the world. The platform learns from the projects by looking at what was planned to happen and what actually transpired. The data is analysed, similar tasks and relationships are automatically grouped, patterns are drawn using AI, and the platform can then predict the most likely outcome for every task and for the overall project. The platform is also able to process 90% of the world's schedules in three formats, so it can cope with almost any format the original schedule was created in. Using machine-learning also means that the technology is inherently scalable, and can grow in its own intelligence as more data is added and can work autonomously. Project teams within construction companies will benefit from the rapid generation and validation of optimised schedules, as well as early warnings around issues. This greater certainty will reduce delays, overspend or rework - increasing productivity. In turn this will improve confidence in the industry's ability to plan accurately and deliver on time and on budget.

Collaborators

nPlan (lead-SME) provided the vision and the initial platform. This then opened the door for Cambridge School of Engineering to apply their in depth knowledge of machine-learning and data analytics to open up the scale of the project, which in turn would not have been possible without the raw data provided by Kier.
As a result nPlan was able to expand the remit of its platform, the University gained further insight into the construction industry and Kier has been able to refine and improve its scheduling process. nPlan is working with Network Rail to test the platform across its portfolio of projects, potentially transforming the way major rail projects are delivered across Britain.

  • Cambridge University
  • Kier
  • Network Rail
  • nPlan

Lead support

The Transforming Construction Challenge funding has supported the development of the software and testing of the ML algorithms. nPlan has been on the i3P SME Engagement programme leading to the partnership with Network Rail to test the software. nPlan has also benefitted from a TCC Covid Continuity Grant to see them through the financial burdens of the pandemic.

Long Term Vision

Machine-learning from historic projects can improve planning and scheduling on large, complex construction projects so timings and budgets are met, productivity increases, and costs come down. AI-Optimised Pathways for Schedule Execution will improve decision making through intelligent historic data. Alongside human experience, it will increase certainty and lead to better outcomes for on-time and on-budget delivery. Tested on two of its largest rail projects, Network Rail has already shown that, by leveraging past data, it could save up to £30m of a £3bn programme.

Human Stories

There has been great collaboration between private, public and academic organisations in this project. nPlan (lead-SME) provided the vision and the initial platform. This then opened the door for Cambridge School of Engineering to apply their in depth knowledge of machine learning and data analytics to open up the scale of the project, which in turn would not have been possible without the raw data provided by Kier. Now Network Rail is testing the software on two of its largest projects, the Great Western Main Line and Salisbury to Exeter Signalling project.

Powerful Processes

Combining the vision of nPlan with the breadth of knowledge from Kier and the machine learning depth of knowledge from Cambridge University, this unprecedented project has created an accurate planning tool for the industry that can galvanise teams around scheduled deadlines. It brings historic data and real-world evidence together with human experience to create more accurate forecasting on programme planning.

Fascinating Facts

AI-Optimised Pathways for Schedule Execution draws on over 200,000 historic construction projects, the largest data set of schedules in the world. nPlan estimates the platform will save 5% of the time normally taken from project inception to project sanction. More accurate project scheduling will provide improved certainty of budgets and timing, reducing contingency by 20%. Improved certainty with minimise delays and reduces the cost of rework or unexpected resource. This in turn will increase productivity and has the potential to improve profit margins by an estimated 15-25%. Tested on two of its largest rail projects, Network Rail has already shown that, by leveraging past data, it could save up to £30m of a £3bn programme.

Benefits

Cost
Improved certainty will minimise delays and reduces the chance of overrunning programme, rework or unexpected resource. This has the potential to improve profit margins by an estimated 15-25%. This is primarily achieved by flagging unknown risks to the project team, allowing them to mitigate those risks before they occur at significantly lower cost than if they are missed or ignored. Network Rail tested the platform on two of its largest rail projects – the Great Western Main Line and Salisbury to Exeter Signalling project. It found that by leveraging past data, it could save up to £30m of a £3bn expenditure on the Great Western Main Line project.

Investment
Kier have now employed nPlan’s services on their Guildford test project. The work and outputs were used to learn from and inform Kier's tender for the project.

Productivity
Productivity is at the heart of this platform, using learnings from over 200,000 project schedules to create accurate, objective and hyper-rigorous planning programmes. Teams across projects, from within and across contractors, will perform better with this great certainty. And benchmarking delivery models in this way will lead to standardisation and best practice across the industry, improve productivity.

Time
nPlan estimates the platform will save 5% of the time normally taken from project inception to project sanction. More accurate project scheduling will provide improved certainty of budgets and timing, reducing contingency by 20%.

Uptake
Network Rail is using the platform across its portfolio of projects, potentially transforming the way major rail projects are delivered across Britain. Network Rail tested the platform on two of its largest rail projects – the Great Western Main Line and Salisbury to Exeter Signalling project - which represent £3bn of expenditure. It found that by leveraging past data, it could save up to £30m on the Great Western Main Line project alone. Network Rail will now embark on the next phase of deployment, rolling out the software on 40 projects before scaling up on all projects by mid-2021. Data from over 200,000 programmes will increase prediction accuracy, reduce delays, allow for better budgeting and unlock early risk detection, leading to greater certainty in the outcome of these projects. Alastair Forbes, Network Rail’s programme director (affordability) said: “By championing innovation and using forward-thinking technologies, we can deliver efficiencies in the way we plan and carry out rail upgrade and maintenance projects. It also has the benefit of reducing the risk of project overruns, which means in turn we can improve reliability for passengers.”