Machine learning is all the rage nowadays, and for good reason. In manufacturing, machine learning models can help companies build better forecasts for short and long-term planning. They can be used to improve workplace automation and worker safety, and avoid errors on the shop floor.
What you may not know is that machine learning can also be an equally powerful tool to design and improve your standard operating procedures (SOPs) and work instructions. At SwipeGuide, we’re working on two use cases that apply machine learning models.
SwipeGuide builds on instructional design principles to help organizations create effective work instructions, resulting in standardized ways of working. Instructional design can be defined as the process of methodically designing, creating, and sharing instructional materials including work instructions, one-point-lessons, SOPs, and educational courses. The goal of instructional design is to offer user-friendly and effective knowledge acquisition to the end-user.
We embarked on a mission to combine our knowledge of instructional design with machine learning techniques like natural language processing and pattern recognition. The end goal: to deliver recommendations that can improve the quality of SOPs, guides, and instructions - both short-term and long-term.
Use case 1: Improving instructions as they’re created.
Although our platform leverages instructional design principles, not everyone who writes work instructions has instructional design training. This means instructions can sometimes be too long to consume quickly, on the job. Mistakes in grammar and spelling or incorrect word usage can also occur. Combined, these mistakes can confuse the reader, resulting in workflow inefficiencies.
Using natural language processing, we are building an automatic system that recommends improvements to the writer, which can improve clarity for the reader. The system looks for the following best practices:
- Correct spelling and grammar, along with correct technical terms for specific actions.
- Short sentences, no longer than 20 words.
- Each sentence can only contain one instruction, unless more actions occur at the same time.
- Instructions must be written in imperative form.
- Instructions must be written in active form.
Real-time recommendations will be available for work instruction creators, much like the spell-checking functionality in word-processing tools.
Use case 2: Improving instruction effectiveness over time.
The second use case we’re working on uses pattern recognition to measure the effectiveness of work instructions over time. By looking at user behavior and sentiment at the Step, Instruction, and Guide level, we can analyze the effectiveness of content on the platform. The insights from our pattern recognition analysis will help us deliver content recommendations for customers, who can then apply them to:
- Improve Steps, Instructions, and Guides that are underperforming.
- Develop new instructional material based on recommendations.
Future-ready work instructions.
Machine learning has many interesting applications in both manufacturing and instructional design. SwipeGuide is combining the best of both worlds to help companies optimize their work processes and embrace continuous improvement. By digitizing instructions and using machine learning techniques to improve them, every employee can access not just the necessary information but the most effective information at the right place and the right time. This improves productivity and minimizes errors on the shop floor.
Join customers like Heineken and Pepsico and start building future-ready work instructions today.