Background Image

5 steps for successfully implementing

computer vision in manufacturing

BY KHIZER HAYAT

5 MINUTE READ

Pilot purgatory is a common problem that organizations face when implementing new technologies such as computer vision in their operations. Manufacturing is no different and in many cases, projects get stuck in the pilot phase for multiple years before dying off due to the high risk of implementation in production.

There are 5 crucial steps that manufacturing organizations should consider when implementing new technology such as computer vision into their production operations to ensure that the technology can scale beyond the pilot phase and into production.

Identify & quantify business problems

One of the biggest mistakes most organizations make when implementing technologies such as computer vision is that they don’t take into account the business problems and only implement technology for the sake of looking innovative. In order to ensure that a technology will make it to production, it is essential to gain a clear understanding of the business problems the organization is facing so that the solutions implemented ensure value from the start. For example, a manufacturing organization might be struggling to meet demand due to issues in their production operations causing them to lose $1 million in lost demand per month.
Once the business problems have been identified, the losses should be quantified in monetary terms. By ‘losses’ we mean the financial impact to the business if the problems remain unaddressed. After the financial loss for each business problem has been quantified, the business problems should be ranked in order of financial impact/loss i.e. the highest financial impact/loss problem should be number 1 on the list with the lowest financial impact/loss problem at the bottom. This ranking helps prioritize the problems.

Form project teams to tackle each of the top 5 business problems

To avoid getting stuck in pilot purgatory, form a project team that is in charge of solving each of the top 5 business problems. Depending on the size of the organization, multiple project teams may need to be formed to tackle the problems. Team members should be picked based on their knowledge of the business problems and be intimately familiar with the impact of the problems on the organization. Ideally, the team members should also be knowledgeable about the technological solutions in the market as that will help with the analysis of the solutions. However, the task of solving business problems should not be outsourced to an innovation team that is not involved in the day to day operations or lacks familiarity with the problems.
Each team also needs to have executive support and the ability to make decisions independently to avoid delays in solving the problems. For example, a team formed to solve production problems could directly have the COO of the company as an executive sponsor and given a budget that they can use without additional approvals.

Each team should determine the root cause and come up with a list of all solutions to solve a business problem

When coming up with potential solutions, it may be tempting to think of technological solutions as the best solutions however it is imperative to consider all options before opting for one solution. For example, the root cause of issues in production operations might be lack of visibility into the operations which makes it difficult to improve them. Possible solutions here could be running time studies multiple times per month, implementing a barcoding system for tracking production output, implementing a computer vision system for continuous monitoring of the production operation and implementing a simple white board system that is filled at the end of every shift with the production output. A process such as DMAIC should be used to streamline the process of determining the root cause and the potential solutions.

Determine the pros and cons for each solution and rank them based on expected Return on Investment

After coming up with all possible solutions, it is important for the teams to analyze them in terms of the pros and cons of each solution. It is also necessary to estimate the cost of implementing each solution to gauge what the Return on Investment (ROI) is expected to be. When assessing technological solutions such as computer vision, it is also necessary to have a realistic understanding of what the technology can achieve. For example, it is unrealistic to expect a computer vision system to generate an accuracy of 99%+ in a dynamic production environment but achieving accuracies between 90% and 95% is a fair assumption to calculate the ROI.
Oftentimes, organizations fall into the ‘Peak of Inflated Expectations’ when implementing new technologies only to be disappointed by the outcome when the technology fails to live up to the lofty expectations resulting in the technology being deemed a failure and never implemented again. Therefore, expectations should be tempered at the onset by conducting thorough research into the technologies and what can realistically be expected of them in a production environment.

Start small at a facility where innovation is welcome

Once it is determined that a given technology is the best solution to solve a business problem in terms of the expected ROI, implement it at a facility where the culture welcomes innovation for improving operations. These are usually facilities that are mini-startups in an organization and are constantly looking to stay ahead of the curve through constant innovation. Their management team is known for implementing innovative solutions and are happy to champion efforts to introduce new technology. It is best if a project team member is part of this facility so they can champion the project.
Once a facility is selected, run the pilot test on a full production line for a period of 6 - 12 months. Anything less than 6 months is too short to truly realize the potential of a technology since there is a ramp up period for implementing and using the technology. Once the technology is fully implemented, the team should constantly work on getting users engaged with the technology and determine the value gained. As soon as the value is realized, the team should begin developing an expansion plan for the technology. Ideally the technology would first be implemented across the entire production facility and then to other facilities.
At i-5O we’ve successfully helped multiple clients productionize computer vision in their organizations using these 5 steps. Contact us to find out how!

Try it for Yourself

Schedule your 1st Free Consultation and Start your AI Journey Today