Learning & Innovation Research Manager| Project Management Institute (PMI)Spain
Data quality and quantity is particularly important as we think about leveraging AI on projects. Considerations include the diversity and comprehensiveness of the data that is available to us.
Have you ever encountered unexpected challenges or pitfalls while using data in your projects? How did you navigate the situation and find a resolution? Saving Changes...
Sherri-Gae ScottProgram Manager, Medical Innovation| American Gastroenterological AssociationBaltimore, MD, United States
The most common pitfall my organization encounters is the lack of or non-retention of data and even not collecting the right type of data. This makes it difficult to make informed decisions on certain projects. We are now implementing PM standards through the development of a PMO. Saving Changes...
Thank you Nikita Jha for sharing this illustrative experience with data. Since we are encountering "new" problems we have to be creative in the solutions and yours seems to be a great option. I guess that finding the right resources for that fast-tracking approach was a challenge. Did you include specific data-related profiles? Have you detected any new roles needed in the context of project + data? thank you
I would say that ensuring applicability of the data is the greatest pitfall. Meaning finding data that is most appropriate for the task at hand. Saving Changes...
I agree that human input poses and inherent risk of data quality issues. Therefore, the requirement to manually review the data for accuracy.
Agreed. I anticipate data quality and clean-up to be significate undertaking to ensure internal AI projects have the best results and output. Decades worth of legacy process and procedures that are out of date will skew results. Saving Changes...
Chanukya RajagopalaDirector - IT Strategy - R & D| iPOCA Private LtdUnited Kingdom
Handling unexpected data challenges is an integral part of any data project, especially in complex projects and programmes.
Project Details
The project involved developing a Electronic Health Record (EHR) system.The aim was to create a robust, secure, and user-friendly platform that adhered to the DISHA guidelines governing health data in India.
Challenges Encountered Despite meticulous planning, several data-related challenges emerged as we moved deeper into development
A - Data Standardisation Issues Medical data was being sourced from various formats: legacy hospital systems, diagnostic labs, and user-uploaded documents. Each source followed different standards, leading to inconsistencies in data types, terminologies, and formats.
B - Data Integrity and Completeness Large chunks of historical data lacked critical details, such as timestamps, units, or identifiers. Additionally, duplicate entries were prevalent, which could potentially affect analytics and decision-making.
c - Integration with External Systems Third-party systems, such as wearable devices and telemedicine platforms, offered APIs that were either poorly documented or inconsistent. Integrating these systems posed challenges in maintaining a unified data structure.
D- Compliance and Privacy Concerns Adhering to DISHA guidelines required an extra layer of encryption and data anonymisation to safeguard patient information, which introduced performance bottlenecks during data processing.
Approach to Resolving Challenges
Addressing these challenges required a combination of strategic problem-solving, collaboration, and technical execution:
Data Standardisation Strategy Conducted a comprehensive audit of incoming data sources to identify key variances. Designed a data transformation pipeline that converted non-standard data formats into FHIR-compliant structures. This included mapping terminologies to standard codes (e.g., ICD-10 for diagnoses).Collaborated with data providers to establish clear guidelines for data formatting and submission.
Ensuring Data Integrity and Completeness Developed a data validation framework that flagged missing or inconsistent entries for review. Implemented automated deduplication algorithms leveraging fuzzy matching techniques. Introduced a user-driven correction mechanism, allowing patients to update incomplete records with necessary details.
Streamlining External System Integrations Conducted exploratory testing to uncover nuances in third-party APIs and documented our findings.
Built an API gateway to standardise data flow between our platform and external systems, ensuring consistency despite differing protocols.
Established a sandbox environment to simulate integrations before deploying live connections.
Addressing Compliance and Privacy
Leveraged modern encryption algorithms to secure data both at rest and in transit. Adopted a pseudonymisation technique, ensuring that personally identifiable information (PII) was decoupled from clinical data wherever feasible. Conducted regular audits and penetration tests to assess system vulnerabilities and maintain compliance. Outcome By adopting a systematic approach, the project overcame these challenges and achieved its goals. The platform successfully aggregated patient data into a unified, accessible, and secure repository. It not only met but exceeded compliance standards, earning trust from users and stakeholders alike.
Key Takeaways
Anticipate Variability: Expect diverse formats and data irregularities, especially when integrating data from multiple sources. Iterative Problem-Solving: Break down complex challenges into manageable tasks, solving them incrementally. Stakeholder Collaboration: Engage data providers, end-users, and third-party vendors early and frequently.
Continuous Improvement: Treat challenges as opportunities to enhance system robustness and scalability.
Navigating data challenges is never straightforward, but with the right mindset and tools, they become stepping stones rather than roadblocks. This experience reinforced the importance of adaptability, diligence, and a solutions-focused approach in achieving project success. Saving Changes...
Joseph NosakowskiBusiness Technology consultant II - Program Management| BlueCross BlueShield of MichiganOwosso, Mi, United States
Through the use of very robust data sharing agreements and establishing who, what, when and where types of data being used in a project provide clear direction and oversight on how to manage data use. Saving Changes...
Anonymous
- Determine nature, scope, scale of the data challenges
- Determine root cause of data challenges - address root cause (i.e. is it a data quality issue, data reliability, other issues)
- Determine risk of the data challenge and execute risk management plan
- Communicate with project team, owner, and stakeholders as needed
- Document the challenge & it's root cause, the intervention to fix/mitigate the data challenge, and the results of the intervention. May need to iterate on potential solutions depending on the nature of the challenge. Saving Changes...
Anonymous
the greatest pitfall is having good foundational information for which to train an AI model. As time progresses and AI becomes more of a standard, I would expect that the collection and storage of data becomes more consistent, offering more reliable AI results Saving Changes...
Anonymous
the greatest pitfall is having good foundational information for which to train an AI model. As time progresses and AI becomes more of a standard, I would expect that the collection and storage of data becomes more consistent, offering more reliable AI results Saving Changes...
font _msthash="646" _mstmutation="1" _msttexthash="2319143879"في مشروعي السابق ، كان الهدف هو تقليل تقادم تقنيات قواعد البيانات في جميع أنحاء العالم ويعرف أيضا باسم العديد من المناطق. استغرق الإطار الزمني للاتصال بكمية هائلة من البيانات دورة أسبوعين للاستجابة وخطة العمل. ومع ذلك ، استمرت البيانات في النمو حيث لم يتم إدراج العديد من الأصول في منطقة معينة حيث فشلت واجهة برمجة التطبيقات في استرداد المعلومات في الوقت المناسب. ومن ثم ، فقد اتخذت هذه المنطقة كمشروع فرعي بموارد مخصصة وتتبع سريع لجعل المنطقة على قدم المساواة مع المناطق الأخرى. لقد كان الأمر صعبا للغاية ولكنه في النهاية كان مجزيا من حيث الإنجاز وتقدير أصحاب المصلحة./font Saving Changes...
Roberto Miguel HernandoSENIOR INTEGRATION ENGINEER| HITACHI DATA SYSTEMSBarcelona, Spain, Spain
Yes, data quality and quantity have been critical factors in our AI-driven projects. One unexpected challenge we encountered was dealing with incomplete or biased datasets, which affected the accuracy of our model outputs. To navigate this, we implemented a data validation process that included anomaly detection and cross-referencing multiple data sources to ensure diversity and representativeness. Additionally, we incorporated continuous monitoring and feedback loops to refine the data over time. Establishing clear data governance policies also helped us mitigate potential pitfalls. I'd be interested to hear how others have tackled similar challenges. Saving Changes...
PMI Team Member
