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...
Ling Yu TSOIProgramme officer| The Hong Kong Academy of Gifted EducationYuen Long, -, Hong Kong
Inadequate data quality and quantity hamper AI effectiveness. To overcome this, I prioritize data cleansing and augmentation, addressing inconsistencies and missing values. Collaboration with data scientists and subject matter experts ensures diverse and comprehensive data, mitigating biases. Regular monitoring and validation help identify and resolve issues promptly. Saving Changes...
Anonymous
How do you navigate unexpected data challenges in your projects?
By continuously monitoring the quality and authenticity of the data inputs and the behavior of the outputs. Saving Changes...
As data grows and time passes, quality and relevance of data are going to deteriorate. There are continual changes in design standards, quality and HSE standards, statutory acts / laws, material prices, productivity of equipment, reporting formats etc. considering the dynamic nature of these data, the model should be designed to update itself on all these reference points; Also, time limits can be set for the relevance of the data and after that the usage of these data for any purpose need to be validated by the experts in the relevant areas. Saving Changes...
You always need to act pro-actively about unexpected changes. Question is if it is a risk or it is / you can turn it into an opprtunity. Saving Changes...
Donna OttDirector of PM/PMO| ETSMaple Shade, Nj, United States
Feb 01, 2024 4:13 PM
Replying to Verónica Elizabeth Pozo Ruiz
...
The main pitfalls that we can encounter when working with large amounts of data are inaccuracies, incoherency, and duplicated or outdated data. It's appropriate to use Data Quality Management Software to identify and correct these issues, and also monitor and synchronize data across the company, keeping database quality perfect.
In accuracy and duplicated data has always been a problem. We have had client requests to retain personal data for 50 years and with that long a period owner of the data don't remember data on fil3e, This may be a dumb question, but when AI is pulling data from various sources how do you make know that proper security and authorization to use data is in place? Saving Changes...
Lucy BellissimoProgram Manager| York UniversityToronto, Ontario, Canada
We had to implement a data governance process and structure in the organization in order to resolve issues of data in disparate databases that were not reconciled. That is, the data point for example may be labelled the same way but one system calculates real time and one has a once a day update and so depending what was used, they could look different. A data governance process with stewards, accountability, shared approved definitions and documentation helped us clarify what we needed for our project. Saving Changes...
Anonymous
When it comes to HR and hiring data, there was a case where we had a sales leader who had no diversity within the team. We wanted to understand why this was and go through the hiring data. At first glance, the hiring manager appeared to not be open to have diversity within the team, but his predecessor also had the same challenge. The current hiring manager was promoted and was a new leader and didn't have the training in diversity and his predecessor didn't have an appetite for it. Saving Changes...
Jonathan BurleyMr| Cornwall Fire and Rescue ServiceFalmouth, Cornwall, United Kingdom
We used vast amounts of data as part of an options project in which to identify the most efficient means of managing the emergency response to calls. The data was extracted from a command and control or computer aided dispatch system. Data was exported into excel via CSV entries within the proprietary database. Initially NULL entries were ignored as generally these data sets were not populated. However what this masked when we reviewed the raw data was a failure of the system to export due to unknown, random and temporary faults which had been unnoticed. We were able to model averages for the period we had then introduce them back into the data set. So the lessons are do not take source data at face value and consider ways to rebalance or mitigate inaccuracy. Saving Changes...
Mike FrenetteManager, IT PMO| Halifax Water (retired)Halifax, Nova Scotia, Canada
As I see all the responses about data, it occurs to me that one of the issues we have with data and will have more of is poor data design caused by apps developed using no code/low code. While these apps will generate fairly solid code, they are often predicated by data provided by users not schooled in data design. This can result in redundant data that is not kept up to date with the "source of truth" as well as poorly designed data structures that are not normalized. Many people trip up on one of the most basic question regarding relationship cardinality, that is, the relationship between two data items, whether it is 1:1, 1:M, or M:M. The last one causes the most problems, since it requires the introduction of a third record to manage the M:M relationship. These are called Associative Entities. Another area people trip up is in the identification of a primary key, such that a record is not uniquely identified, causing more relationship-type issues. For example, using a person's name as a key when it is obvious that name are not unique. and attempts to relate two records failing due to the imroperly identified primary keys being used as secondary keys in related records.
Of course, there are many other data design issues. The examples here are just the tip of the iceberg.
The point here is that if you are dealing with data conversion involving user-created data used to feed low code/no code applications, be very wary of the source, accuracy design issues within the data. Saving Changes...
My journey in researching on various project database across the globe has underscored the critical importance of data quality and quantity. Leveraging AI effectively hinges not just on the volume of data at our disposal but, more crucially, on its diversity, accuracy, and comprehensiveness.
One such challenge I faced was the outputs I received on project execution models and successful execution techniques. The dataset we initially relied on was extensive yet heavily skewed towards data from a few high-income countries. This lack of geographic diversity and representation poses a significant risk of bias, potentially leading to inaccurate predictions for East or Southern nations. Collaborative effort to broaden the data sources is essential to mitigate this bias. Ensuring data quality, diversity, comprehensiveness and identifying potential biases and taking proactive steps to mitigate them is important in utilizing AI. Saving Changes...
PMI Team Member
