Mondays: So Wrong And So Right

Thank you for subscribing and for all the feedback this week. Last week, I updated my certifications and what I am writing here based on your recommendations. I am grateful to everyone who takes the time to send me comments and emails. Even likes (or a lack of likes) are helpful feedback. We are all busy now, so I know the value of your time.

Many new followers are getting this Monday post for the first time. Some come from Eric Webber's Substack, From Data To Product. He has some excellent content and a rich library going back to last year. Welcome to this growing community. Your feedback benefits everyone. Drop me comments, DMs, or emails to let me know your thoughts.

I was wrong earlier this year. I told everyone to pull in hardware purchases, especially GPUs. YIKES. That GPU take was way off. Why? I did not look at the network effects of the crypto crash.

I talked about it but did not connect the demand for GPUs to declines in crypto price. I should have because the link is obvious. Macro factors play a significant role in accurate models. What's the fix?

I talk to more people before I publish a prediction. My circle of peers and advisors is growing. I am building network diversity to get a broader range of perspectives. I am discovering new sources of high-quality content that also bring new knowledge into my thought process.

My certification courses use a new design methodology that teaches intelligent processes like strategy and leadership. Traditional course design theories are built to teach logical processes with specific knowledge and defined steps. I created a curriculum that teaches systems, models, and frameworks.

There are steps, concepts, and knowledge. However, intelligent processes require adaptability and synthesis to novel scenarios. Intelligent processes are mastered through years of experience. My challenge was to build a course design to deliver 8 years of AI strategy or 15 years of leadership and the core concepts.

I neglected to anticipate how disruptive the course design would be to students. I spent last week and will be working through this week to include constructs from traditional course design. I must use novel design methods while integrating them back into traditional ones. I failed to think about helping students transition from what they are used to into the new methodology.

I didn't test the new methodology rigorously enough. My validation group already understood what I had built, so it was not as disruptive. They are people I worked with, so they also had a grasp of the core curriculum in practice. Rookie mistake. My test audience was not representative of my students.

Being wrong in public is critical. I am accountable for improving, but in the influencer community, it's easy to start believing your own hype. Stagnation comes from ignoring mistakes or covering them up.

Quantitatively measuring outcomes is critical for improvement, but so is accountability. Self-accountability does not hold the same power as being accountable to others. That is how I stay close to my customers. I ask for feedback and revealing my mistakes creates an open environment for criticism.

In a LinkedIn post last week, I talked about upskilling senior software engineers into data engineer and machine learning engineering roles. Several comments disagreed, but in this case, I am not wrong.

Large tech companies have reskilling programs for software engineers to migrate into the data team's engineering roles. I have built training programs for years, and they are successful. I know several people who made the same transition and are effective data and machine learning engineers today. My core thesis has strong support.

The majority held belief can be wrong. A new way of thinking or operating starts in small communities, and the first to evangelize their discovery is a target. Many of the comments were personal attacks or over the top.

I responded to well-thought-out points with counterpoints. Sometimes, I left them alone to let others see a different perspective without my views. I fired back at the people who were rude and unprofessional. Tolerating them is how toxicity spreads. Stupid should hurt.

Over the last decade, I have been the target on other ideas. I talked about making data science more accessible, and the gatekeepers came out in force. I described a more effective method to hire and assess data science talent. Recruiters and gatekeepers came out in force.

Part of being in data science is telling brutal truths that contradict firmly held beliefs. Data professionals are ineffective without developing the capability to deliver reality that most would rather ignore. Admitting when I am wrong is part of my capabilities.

Trust is at the heart of what we do. My clients expect me to give them my best assessment and notify them as soon as I realize there could be an error. We all operate on the best available information, and best doesn't mean perfect. Risk is part of the job.

The opposite end of trust is not backing down when firmly held beliefs are wrong. My clients trust me to deliver hard truths, support them, and stand by them through tough questioning. That is every data scientist's job.

There will always be pressure to back down, and it will always be easier to bend the message. No one invites conflict, but our job cannot always avoid it. A younger me was afraid of getting fired. The older me knows that I don't have much value if I cannot stand behind my assessments and will eventually be let go anyway.

Being wrong is our source of authority to hold firm when we are right.

Vin