UK Ministry of Housing, Communities and Local Government (MHCLG) 
Back to Digital Planning Programme Case Studies
Challenge
Consultations are a vital part of planning, yet participation rates for large consultations have historically been relatively low and involved significant resource requirements and lengthy processing times. Increased rates of participation are desirable, but they tend to make data processing slower and to extend analysis timeframes.
Approach
Plymouth City Council, South Hams District Council and West Devon Borough Council partnered with Commonplace, developers of a leading citizen engagement platform. Together they trialled a new approach using Natural Language Processing (NLP) to improve the efficiency of consultation analysis and reporting. Natural Language Processing focuses on computers interpreting human language using techniques from linguistics, machine learning, and artificial intelligence. PropTech Innovation funding of £200,000 was made available for this work, aiming to deliver a combination of improved local engagement, time savings and cost savings.
The project developed new workflows for data management, analysis and reporting. It involved the use of a standardised set of tags for analysis across Regulation 18 and Regulation 19. It tested whether NLP models could automate processing tasks to use time and resources more efficiently.
The goal was to achieve same day reporting on consultation data at the most granular level, whether it be about an individual site, subject area, or consultee, so planners could react and explore emerging trends.
The delivery team was responsible for scoping and developing a model with Commonplace with over 100 tags that could be applied in different local authority contexts anywhere across the country. The model included sentiment analysis – a way of classifying peoples’ responses as positive, negative or neutral based on the language used in their comments. The aim was to provide planners with instant consultation feedback and data trends without extensive manual processing.
Results
The project results were compared to previous Regulation 18 and 19 consultations and showed significant time and cost savings. Firstly, there was a 66% improvement in processing and validation time per consultation response, saving up to 10 minutes per response. Secondly, evidence suggested scalable savings of between £19,000 – £38,000 in officers’ time spent reviewing, processing and summarising consultation data. After accounting for the technology licence cost, savings ranged from £7,000 – £30,500 depending on engagement levels and the NLP product chosen.
The pilot demonstrates that the more successful a Local Authority is in engaging communities, the more value can be extracted from this approach to data management and analysis. Once a consultation reaches a level of 3,000 respondents, the NLP product can cover its cost, with the same value of officer time able to go back into planning services.
The goal of same day reporting was not found to be possible due to changing engagement levels, but within a week of consultation close was assessed to be achievable in test conditions with significantly reduced resource demand.
Next Steps
The project shows a promising picture, but first more refinements to interfaces are needed to improve data accuracy and further reduce officer validation times. Once addressed, there is potential to deliver significant benefits that can scale, including more flexible and collaborative consultations and improved efficiency and relevance of reporting for different audiences.