Moderator: Dr. Barbara Kowalcyk, Director of CFI
Panelists: Melody Ge - VP of Governance, Intelligence, Analytics at Corvium; Founder of Women in Food Safety
Andrew Kennedy - New Era Technology Team Leader at FDA U.S. Food and Drug Administration
Suzy Sawyer - Food, Safety, Quality & Regulatory Data Lead at Cargill
Webinar recording date: October 27, 2021 12:00 PM - 1:00 PM EDT
Experts from academia, industry, and government have long recommended the development of a data-driven approach for improving food safety. Vast amounts of data are collected throughout the food system – on the farm, in processing, during distribution and at retail – but limited sharing makes it difficult to access this information for effective decision-making.
What is needed to effectively and efficiently use existing data to inform decision-making? The panelists will share their perspective and identify how to move food safety towards this goal of better information to make informed food safety decisions.
Webinar Questions and Responses From Panelists
Responses from Melody and Suzy are below.
- This question is for Melody: Great insights! Can you speak to data visualization at Corvium? Do you use different visualization approaches for senior leaders versus plant employees?
- (Melody) No, we use an industry known BI tool, just like all other tools to visualize the data. We also developed a platform for our own users based on the BI tool, and the users will have different views of the report they see based on their management level or job functions.
- This question is for Suzy. What is your favorite plant floor quality data software tool. What external public health data does Cargill use?
- (Suzy) Cargill is a large, multinational organization with a diverse product portfolio. Given this, we have multiple data software tools used in our plant locations globally and there is not one to call out more than others. Also, I believe there is great importance in how the software tools are deployed and governed for them to be successful. This includes having defined processes to manage the data, standards/rules and monitoring in place. This goes beyond the features and usability offered by the software tool itself.
- What sort of data related to food safety is publicly available? What data do you think should be publicly available?
- (Melody) There are quite a few data that are available publicly, such as recall and FDA inspection reports. I am not a data rights expert to answer the second part. But I think more importantly, as industry professionals, we should share more publicly the outcomes from the data.
- (Suzy) FDA provides publicly available data through their Data Dashboard at FDA Dashboards - Home. This includes visualizations and APIs for data retrieval on Inspections, Compliance Actions, Recalls and Imports. USDA provides publicly available data sets at Data Sets and Visualizations | Food Safety and Inspection Service (usda.gov) Data.gov provides a catalog of publicly available government datasets and includes a Topic Category of Food Safety and Nutrition.
- In your experience, how much change in operations or data management occur in a company as a result of building these end-to-end systems? How much does these needs for quality data or intelligence change operations? Are these hidden costs? What are the big costs if there are changes?
- (Suzy) Change management is built into our project plans where we are implementing common process, data and technology to gain adoption of the new solution. The change and project activities to address this can vary depending on what they are going from as their legacy solution. It can be a non-tech or low-tech enabled process where we tend to see less resistance with the new technology. Or it can be replacing an existing technology solution which has been in place for a long period of time and the value to change is not seen directly by a set of users who may resist the change. A hidden or cost increase can occur on these projects if the delta of change is not identified and addressed in the project. This tends to create scope creep and / or delays in the project adding to the overall cost to implement.
- There are some important issues that still need to be addressed. The ideal case is that we have zero outbreaks. However, this is probably cannot occur in practice. So how do we know that we are doing better using data analysis? Data collection and analysis have the potential to be very useful, but they could also be extremely resource expensive. What are the protocols for deciding when we would be better off dedicating resources in other directions?
- (Melody) I don't disagree. However, just like the mentality six sigma has, will we ever achieve the perfect six sigma in operation? Maybe not, however, we always find ways of working towards the perfect six sigma. Same thing here, all that we do as food safety and quality professionals is doing our best to achieve zero outbreaks, as close as possible (infinitely closer to zero). On the second part of the question, data is everywhere in our production, more importantly is the interpretation part and action on it. Otherwise, data is only just data. Lastly, this is like all other resources you will ask from your management team, yes, to make the changes might be costly, however, the ROI is tremendous, which is important to share with the management team when we make the decisions.
- (Suzy) Prior to starting a project, the business case needs to be captured which include the value of investing in the effort. There also needs to be success criteria established. With projects planning to scale, conduct a pilot or proof of concept to prove the hypothesis or business case has the value stated. This can be challenging in compliance or risk management. Often times, there is also efficiency gains or visibility to new insights which can be quantified as well.
- How do you ensure additional pertinent data that is not a part of your agency/company decision making is included? What mechanism do you suggest to ensuring that additional data is incorporated and considered?
- (Melody) This depends, and it's case by case. It depends on what data is missed or what data do you need? I will suggest asking yourself what do you want to learn from the current dataset, and start backwards on asking do we have enough data to proof that? Once you find those gaps, then let's see whether those missing data are publicly available, or you will need to partner with another data owner.
- (Suzy) Establishing a governance model where there is good cross team representation to make decisions on which changes are valid to introduce.
- In terms of data quality, do you check that the appropriate methods were used to generate the data? I am thinking of specific methods prescribed in regulations for detection of specific microorganisms, for example. Methods need to be appropriate for their specific purpose and the matrix being tested to produce quality data.
- (Suzy) Data quality practices are not validating the method to generate the data. The methods would be monitored for compliance through process minimum requirements, standard operating procedures and audit methodologies.