Building a data-driven world

Joined May 2013
730 Photos and videos
Someone said your data pipeline is "broken" and now everyone is panicking. Let's start from the beginning. 🧵
1
207
Misconception: broken pipeline means the data is gone. Reality: it means the data stopped moving. The source still has it. The pipeline just needs to be fixed or restarted.
1
50
A good pipeline is invisible when it works. You only notice it when it breaks.
42
Every org knows about technical data debt: messy pipelines, inconsistent fields, outdated exports. There is a second type that rarely gets tracked: trust debt. 🧵
1
258
The repair for technical debt: fix the pipeline. The repair for trust debt: demonstrate reliability consistently, over time. No shortcut. You have to earn it back.
1
35
Right now, orgs are asking if their data is clean enough for AI. That is the technical question. The harder one: does your team actually believe the data? If not, AI output gets the same skeptical treatment as everything else.
34
Someone just said “we need to fix the data model,” and everyone nodded like that was a normal sentence. Let’s translate. 🧵
1
105
Misconception: a data model is just technical plumbing. Reality: it shapes what questions your organization can answer.
1
30
A dashboard built on a messy model is just a very confident-looking mess. A good data model makes data easier to find, easier to explain, and much harder to accidentally misuse.
35
AI governance sounds like a policy conversation. It is. But in practice, it is also a workflow conversation. 🧵
1
98
AI governance gets more practical when you treat it like data operations: Define the source of truth. Document the metric. Set permission boundaries. Make review visible. Keep a record of what changed.
1
46
The organizations that make AI useful will not be the ones with the longest policy document. They will be the ones that turn governance into a working system people can actually follow. Responsible AI happens in the workflow.
43