Case Studies

Our case studies

We met with stakeholders across the organisation to understand the business problems and to turn these problems into a series of focused questions. We identified useful data sources, built pipelines to extract, transform and load data to a platform ready for analysis. We then recruited a data science team and guided them to deliver the first successful machine learning models in the history of that company.

I/O were asked by a major scientific publisher to create a data science strategy covering their content acquisition domain.

I/O developed a model to categorise manuscripts and editors using NLP and then predict the probability of a match using machine-learning. This model was integrated into the editorial workflow and dramatically improved the success rate of editor invitations.

A major publisher of scientific literature wanted to improve the matching of manuscripts to editors in order to reduce editorial turnaround times.

I/O was asked by the HM Prison and Probation Service to predict which offenders and prisons might be involved in future violent incidents. We built a random-forest machinelearning model to calculate a probability of violence for each offender on a daily basis. Governing Governors commented that these scores were very helpful in managing offenders and reducing violence.

The rate of violence in prisons was increasing, putting the safety of both offenders and prison officers at risk.

I/O data science developed a system to use optical character recognition and a recurrent neural network to partly automate the redacting of documents. This system allowed HMPPS to reduce processing times and clear a backlog of requests.

HM Prison and Probation Service is required to answer subject access requests from offenders without revealing sensitive information about other offenders.

I/O worked with data owners across the organisation to build pipelines to load data into an analytical platform. A series of dashboards were then built to monitor the completeness of this data and present real-time performance indicators. These dashboards dramatically reduced the time taken to prepare for governance meetings and allowed policy-makers to make decisions based on accurate, up-to-date information.

The Environmental Quality Directorate in DEFRA wanted to make data more recent and accessible to support decision-making in governance meetings.

I/O data science built a Monte-Carlo simulation to predict the stocks and flows of manuscripts moving through each stage of the submission system. This model revealed that the apparent rising times were the result of a sampling bias and the true turn-around times were stable.

A major publisher was concerned by rising turn-around times in a newly built manuscriptsubmission system.

We worked with DEFRA to build a model to estimate vehicle churn and emission rates following the introduction of low emission zones in cities. This model allowed DEFRA to understand the benefits to public health and informed the roll-out of low emission zones across the UK.

The Air Quality team in DEFRA were exploring options for reducing Diesel emissions following the discovery that Nitrogen Dioxide was far more damaging to public health than previously thought.

Our client in the Resources team asked us to build a model to predict future UK recycling rates to inform these negotiations. We developed a cost-curve driven simulation to predict the flow of materials around the UK waste management system for the next 30 years. Armed with this information DEFRA were able to negotiate recycling rate and landfill targets that were achievable for the UK.

The Waste and Resources team in DEFRA represented the UK in the negotiation of EU wide recycling targets.

We worked with DHSC to develop machine learning models to analyse patient data and socio-economic factors and predict cancer incidence, prevalence, mortality and survival across regions. These models allowed DHSC to allocate resources and plan infrastructure according to need.

The Department of Health & Social Care wanted to understand the prevalence and survival rates for cancer across the UK.

I/O worked with the NHS to develop a microsimulation to track the flow of patients through diagnosis and treatment pathways. This simulation allowed NHS England to plan GP and secondary care services to cope with increased demand during the roll-out of healthcare interventions.
NHS England wanted to predict the demand on healthcare services following interventions such as screening programmes and awareness campaigns.

Contact info

Springer Nature

Springer Nature is a publisher of scientific text-books, magazines and more than 3000 scientific journals including Nature. I/O data science was contracted to establish an in-house data science team and build the necessary data infrastructure, including an analytical platform and associated data pipelines. We also worked with our client to deliver a number of bespoke data science products.