Swipe Icon
home
Organisational aspects of implementing ML based data editing in statistical production
Appendix 2: Use Case Template
With the explosion of data now available, modern methods such as machine learning have gained significant traction in everyday life by companies such as Google, Amazon, and Microsoft amongst many others. However, the same can not necessarily be said when it comes to National Statistical Organisations (NSOs) where the uptake has been less than in the private sector. This template is to gather some insight on why the uptake by NSOs has been slower than in the private sector and how it could be accelerated.
Guiding questions:
  1. Throughout the journey from experiment to production, what has been stopping your organisation from applying data science and modern methods in data editing?
  1. Focusing on the organisational aspects along the productionisation process, how did the project manager overcome those obstacles (e.g., how to explain the processes and methods of data editing to stakeholders within and beyond the organisation and proof that it is a good value for money)? Please identify the problems and stakeholders involved (e.g., the dynamics between senior management, research team, business area, IT team, etc.) and the actions taken to resolve the issues and improve multi-level engagement throughout the different stages listed in the use case template.
  1. To help contextualise the use cases, please discuss the pros and cons of applying these modern methods in data editing (in terms of accuracy, explainability, transparency, cost effectiveness, or other metrics) that help get buy-in from stakeholders.
  1. What are the lessons learned and best practices that would be useful for other NSOs?
Title of the use case and the name of the organisation
(Please provide a title of the use case with enough information so that readers can understand the context.)
Project overview
Please provide an overview of the project and describe its strategic importance to the work of the organisation, including the statistical programme in question, the business needs of the project, whether the proposed method is replacing an existing method or is a new application and why a modern method is being considered.
Organisational readiness
When completing this section, please keep in mind the readiness of the organisation related to aspects such as the IT infrastructure, the capacity of the organisation in terms of knowledge of the ‘tools’ required to set up the method and to also maintain it, as well as the openness of the organisation to adopt modern methods
Understand business needs (Who needs what)
Please provide information on the context around motivation of the project. Include information such as the business need, who asked/sponsored/paid for the project and enough information for readers to understand what the real business need is so that they can draw parallels with any projects they may have within their organisation.
Assess Preliminary Feasibility
Please indicate what assessments were made in deciding to investigate the method(s) chosen. These assessments could include considerations such as the appropriateness of the method given the data (continuous vs discrete) or the problem at hand, the availability of required resources (both IT and human resources), and the expected improvements over an existing method (if it exists).
Develop proof of concept
Please provide insight on how the proof of concept was developed and include information such as obstacles and how they were overcome (or not), any ‘adjustments’ that had to be made to the planned implementation of the method and lessons learned (both positive (keys to success) and negative (blockers)).
Approach/method used
Please provide a detailed description of the approach or method used to develop the proof of concept, model or solution being discussed in the use case. This could include information such as the algorithm used, the data sources and pre-processing techniques, the hyperparameters and training approach, and any other relevant details about the approach or methodology used.
Prepare a Comprehensive Business Case
Please provide insight on what factors influenced the success of the business case, what were the most important components of the business case to achieve acceptance/agreement, were there any obstacles to the preparation of the business case, and how were these overcome.
Deploy the model
Please include the challenges faced in integrating the model into a production system. These could include aspects such as
- redeveloping the model (to suit production systems and/or data),
- availability of IT human resources (specialised or not),
- reluctance to potentially put a production system at risk,
- availability of specialised IT infrastructure (e.g., ML platform),
- the need for documentation and training of end users,
- availability of funds required.
Results
Please provide information about the outcomes that have occurred if the model has been deployed in production.
Latest status and next steps
Please provide information about the status of the project (e.g., implemented, being programmed into the production system, etc.). If the method is not in production yet, please explain why and share the implementation plan if it exists.
Lessons learned & recommendation
Please consider the following sub-themes: IT infrastructure, IT capacity, organisational knowledge of the proposed method, maintenance of the method once in production and acceptance of the method by business areas.
Reference
Please provide any helpful links and supporting materials.
Contact
Please provide a contact person with an email address.