pogorelov.dev | Writing

Standards that don't remove bottlenecks are a tax

Sun, 12 Jul 2026 19:46:00 GMT

A standard that doesn’t remove a bottleneck isn’t governance — it’s a tax.

We wanted to know where our data teams actually lose time, so instead of running a maturity assessment we mined the backlog itself: 40,000 tasks, read for what the work was about and where it stalled. The answer was lopsided. Stalls concentrated in data preparation, the unglamorous cleaning and reshaping that happens before anyone can use the data, and not in the areas maturity checklists spend most of their pages on.

Then we held every standard against that map and asked one question: which bottleneck does this remove? Standards that pointed at real pain stayed and got investment. Standards that existed because a checklist says a serious organization has them, whose only visible output was a document and a review step, we killed.

A maturity model describes an ideal organization; the task tracker describes the one you run. When the two disagree, believe the tracker — it bills you daily.

The deliverable of a governance review is a kill-list, not a longer checklist.

Notes are live

Sat, 11 Jul 2026 09:00:00 GMT

Short-form notes live here now — quick thoughts, links worth keeping, and things I’m reading on data and AI, in between the longer posts. Expect these more often than essays and less polished by design.

Designing decision systems for executive teams

Sun, 05 Jan 2025 00:00:00 GMT

Executive teams want clarity, not dashboards. The goal of a decision system is to turn signals into action with as little friction as possible.

A decision system checklist

North-star metrics that align teams in every function.
Cadence rituals that lock in decisions and owners.
Narrative context so data points land with urgency.

What makes the difference

When the operating model is clear, analytics platforms stop being a cost center and become a compounding advantage.

Blueprint for high-performing analytics platforms

Fri, 15 Nov 2024 00:00:00 GMT

Building an analytics platform is less about tooling and more about clear operating principles. The best teams treat performance, cost, and trust as design constraints from day one.

What matters most

Clear product ownership. Every dataset needs a steward and a KPI.
Service-level expectations. Align latency, freshness, and completeness targets to business needs.
Automation as hygiene. Testing and observability are non-negotiable.

Playbook highlights

Start with the top 10 business questions.
Map each to a data product with an owner, SLA, and dashboard.
Measure adoption weekly and prune what isn’t used.

The result is a platform that executives trust and teams actually use.