More than five years ago, I sat in yet another meeting listening to someone say, "You know what we really need? A gold layer for our data lakehouse."
Heads nodded around the table. Everyone agreed it was a smart idea. Obviously necessary.
Then everyone went back to their desks and did nothing.
The idea wasn't revolutionary. It was sitting right there on the surface: visible to anyone who looked at our Bronze-Silver architecture and asked, "How does the business actually use this data?"
But here's what I learned: Ideas are everywhere. People who actually build them are rare.
That day, I stopped waiting for someone else to make it happen. I proposed the initiative, assembled my team, and we started building.
This is the story of how we completed a data lakehouse architecture that nobody mandated, while entire market stayed comfortable with the status quo.
The Reality: An Incomplete Architecture
When I joined as Senior Manager of Business Analytics, we moved to what many organizations would call a "modern data platform":
Bronze Layer: Raw data ingestion from all sources into our Hadoop ecosystem ✓
Silver Layer: Cleaned, validated, standardized data ready for processing ✓
Corporate IT in collaboration with us, data analysts, had built a solid two-layer data lake. Standard infrastructure. Nothing special.
But I saw what was missing - the layer that actually serves the business.
Our analysts were spending 60% of their time doing the same transformations over and over:
Aggregating data for weekly executive reports
Joining multiple tables to create comprehensive views for dashboards
Pre-calculating KPIs so BI tools wouldn't time out
Restructuring data so stakeholders could run simple queries
Every analyst. Every week. Reinventing the same wheel.
Meanwhile, across our industry:
Our industry stayed comfortably on Bronze/Silver layers. They weren't even considering migration. Tableau adoption? A couple of dashboards built reluctantly, checking a box for "digital transformation."
Other markets weren't moving either. Everyone was satisfied with Bronze and Silver or hadn't even gotten that far.
The question sitting on the surface, waiting for someone to ask:
"If everyone keeps rebuilding the same datasets manually, why don't we just build them once, correctly, and make them available to everyone?"
The answer: a gold layer.
The idea was obvious. What wasn't obvious was whether anyone would actually do the work to build it.
The Decision: I Proposed It, Then Built It
I'll be honest: I didn't invent the concept of a gold layer. The idea was right there, visible to anyone familiar with medallion architecture or data lakehouse best practices.
But in 2022-2023, almost nobody in our market was doing it. And in our region? Nobody.
Why not?
They didn't see the need. They had data. It somehow worked. Why change?
And the same was true for our headquarters, they didn't mandate it. Bronze + Silver felt "complete enough."
My peers were skeptical. "More layers mean more complexity. Why bother?"
Senior leadership was cautious. "Show us the ROI before we commit resources."
Nobody was saying it was a bad idea. They just weren't willing to own it.
Only my boss told me: ok, you can do it, make the best BI adoption in our country. Or even in our region. And maybe in the world.
So I made a decision and my boss given me a mandate for it: I would propose it, take ownership of it, and build it without waiting for organizational air cover.
I submitted the business case. I pitched the vision. I outlined the roadmap. And when I got cautious approval to "explore the concept," I assembled my team and we started building in production.
This wasn't about having a brilliant idea. It was about being willing to execute when everyone else stayed comfortable.
What We Built: The Gold Layer
While others kept saying "someone should do this," my team and I built the missing piece of our data lakehouse.
Gold Layer: Business-Ready Analytics Foundation
We designed and implemented the third layer, the one that transformed our data lake into a true lakehouse:
Pre-aggregated business metrics:
The data that answered 80% of common business questions instantly - from performance tracking to customer behavior insights. All pre-computed, optimized, and ready to query in seconds.
Dimensional models:
Structured data that business users could work with directly - no complex joins, no technical knowledge required.
Optimized datasets:
Purpose-built for dashboard performance and self-service analytics.
Self-service data products:
Business-ready datasets that enabled analysts to become insight generators instead of data preparers.
The Core Insight
This wasn't about adding more data. It was about creating the right structure so data became instantly actionable.
Instead of every analyst rebuilding the same transformations manually, we created a centralized, reusable gold layer that enabled instant self-service analytics across our consumer business.
The idea was simple. The execution was hard.
The Execution: How We Actually Built It
Execution beat strategy every single time in this journey.
While others debated the "perfect" design in meetings, my team and I shipped working datasets regularly.
My Execution Framework
1. Started small, moved fast
I didn't ask for a massive budget or a dedicated team. We started with one high-impact use case and proved value before scaling.
Within 12 weeks, we had moved from concept to production, with executive dashboards running on gold layer data and self-service analytics adopted by our first business unit.
While others were still in "planning mode," we had users in production.
2. Built with my team, not alone
This wasn't a solo story. I led the vision. Data architecture. My team executed the technical work:
built transformation pipelines
designed data models
optimized dashboard performance
ensured data quality and governance
I was the architect and the shield clearing obstacles, securing resources, managing stakeholders, building optimal data model, so my team could focus on building
3. Moved faster than perfect
Our first gold layer datasets weren't flawless. We iterated and improved them based on user feedback, not theoretical design sessions. United some, split some.
They were real. They were useful. And we refined them in production.
4. Turned skeptics into advocates through results
Every month, I demonstrated measurable impact:
Significant reduction in analyst time spent on manual data prep
Dashboard performance dramatically improved
Self-service adoption growing across business units
Data quality incidents approaching zero at the consumption layer
Numbers don't lie. Skeptics became sponsors.
What Made Execution Possible
Clarity over complexity: I broke the massive "gold layer" vision into achievable increments my team could deliver.
Ownership over consensus: I didn't wait for everyone to agree. I took accountability for the decision and the outcome. But my boss approved it 😀
Speed over perfection: We shipped functional datasets regularly, iterating based on user feedback.
Team empowerment: My team had autonomy to solve problems their way, as long as we hit our milestones.
The Moment That Defined It
Six months in, we faced with dramatic (up to 50%) decrease in number of incoming ad-hoc requests for analytics. Our self service system decreased the simplest part of ad-hocs: all the needed metrics were readily available in BI system so our product, sales, digital and other teams started using self-service analytics.
Our requesters saw the gold layer in action: hundreds of users, instant dashboards, self-service analytics working seamlessly.
Our colleagues from fintech market asked: "We've been talking about modernizing our data architecture for two years. How did you actually build this?"
My answer: "We stopped talking and started building."
The Results: From Infrastructure to Impact
Three years later, the gold layer we built speaks for itself.
Infrastructure Achievement
Completed medallion lakehouse architecture:
Extended corporate Bronze/Silver layers with business-ready Gold layer: first in our region to do so.
Gold layer innovation:
My initiative that bridged the gap between data storage and business intelligence.
Industry leadership:
While industry stayed on 2 layer or even 1 layer architecture, with incomplete BI adoption, we built a full three-layer lakehouse on Hadoop ecosystem.
Operational Impact
- 50% reduction in analyst time on manual data preparation work
- Dashboard performance: Optimized data delivers insights in seconds instead of minutes
- Self-service analytics for 500+ users: business users work with data directly without technical support
- <1% data incident rate down from 10%, as gold layer enforced quality at consumption
- 70% reduction in stakeholder complaints about data inconsistencies
Strategic Impact
**ML models accelerated: streamlined data access reduced development time
400% increase in analytical output: analysts became insight generators instead of data preparers
Zero audit findings on data quality: comprehensive governance embedded in gold layer
Foundation for advanced analytics: Customer intelligence, real-time personalization, predictive models, all powered by gold layer data.
The Business Transformation
From "we have data" to "we make decisions"
From analysts as data preparers to analysts as insight generators
From waiting days for custom reports to instant self-service insights
That transformation happened because someone saw an obvious gap - and actually filled it.
What I Learned: Execution Beats Ideas Every Time
Looking back on this journey, here are the lessons that mattered most:
1. Execution beats ideas every single time
Ideas are abundant. Execution is scarce.
The gold layer wasn't a revolutionary concept. It was sitting right there on the surface: visible to anyone who understood medallion architecture.
But visibility doesn't create value. Execution does.
Every organization has people who say:
"Someone should build this"
"We really need that"
"It would be great if...”
Translation: "I'm not going to do it, but I hope someone else will."
The leaders who change organizations are the ones who say "I will", and then actually do it.
Here's what execution required:
Taking ownership without permission: I didn't wait for a mandate. I proposed the initiative, took accountability, ok from my direct boss and built it.
Building with my team, not alone: My team turned vision into working systems. I led. They executed. Together, we delivered.
Moving faster than perfect: Our first datasets weren't flawless. But they were real, and we improved them based on actual usage, not endless planning.
The uncomfortable truth:
Most organizations don't lack good ideas. They lack leaders willing to take the risk of execution.
2. Identify the missing piece that completes the system
Everyone was comfortable with Bronze and Silver. "We have a data lake. Mission accomplished."
But I asked: "Can the business actually use this data?"
The answer was no. Analysts were manual transformation engines. Business users couldn't self-serve. Dashboards were slow. Every insight required custom work.
The gold layer wasn't about adding more data. It was about making all existing data useful.
Sometimes the best innovation isn't building something new. It's completing what everyone else considered "finished."
3. Build without organizational air cover, if you believe it's right
I had no regional executive sponsor. No dedicated budget. And no regional or corporate framework. Only “do the best analytics system in the world” from my boss.
What I had:
- A clear vision of what needed to exist
- A team willing to build it
- The conviction to take accountability for the outcome
- My brave boss
Waiting for perfect organizational alignment is how good ideas die in meetings.
Sometimes you have to build proof points before you can secure support.
4. Speed creates momentum; momentum creates support
I didn't spend six months designing the perfect architecture. We built working solutions quickly and put them in front of users.
That first dashboard, performing dramatically better than before, became our business case.
Stakeholders don't get excited about PowerPoints. They get excited about results.
5. Industry context matters: being first has compounding value
In a different market, building a gold layer might have been "catching up." But in our region and industry, where others stayed on Teradata or Oracle or something else, and others weren't moving, we became the reference architecture.
Being first, even with an "obvious" idea, creates disproportionate impact.
6. Honesty builds credibility
The gold layer wasn't my invention. The idea was on the surface. What differentiated us was execution when others stayed comfortable.
Claiming credit for execution is earned. Claiming credit for obvious ideas is not.
Own what you built. Be honest about what you didn't invent.
To Every Leader Sitting on an "Obvious" Idea
Today, when I know how to build the system, where business analysts pulling real-time insights without technical help, or executives making confident decisions from instant dashboards, I remember that meeting three years ago.
*"Someone should build a gold layer."
Everyone nodded. Everyone agreed it was smart. Everyone went back to their desks.
And then I proposed it, and my team and I actually built it.
While industry kept talking about modernization, we shipped.
While others debated the "right" approach, we iterated.
While our industry colleagues stayed comfortable with Bronze and Silver, we completed the architecture.
The idea was obvious. What wasn't obvious was whether anyone would do the work.
Here's what I learned:
The world doesn't need more people identifying obvious gaps. It needs more people willing to fill them.
Your "obvious" idea that everyone nods at in meetings? Someone needs to build it. That someone could be you.
Don't wait for the perfect mandate. Don't wait for consensus. Don't wait for someone else to take the risk.
You need to propose it, to own it and finally build it.
The difference between organizations that talk about transformation and organizations that deliver it?
Leaders who stop saying "someone should" and start saying "I will."
And teams who execute.
What's the "obvious" idea everyone in your organization agrees on but nobody's actually building?
I'd love to hear your story in the comments.
All views and opinions expressed are my own and do not represent those of my current or former employers.
About the Author: Aygul Aksyanova is a data analytics leader with 22+ years of experience, including 14+ years managing project/data teams at Fortune top 45. She has led customer data platform implementations, reduced team turnover by 80%, and mentored dozens of professionals who've advanced to leadership roles. She now helps data leaders accelerate their careers and build world-class analytics capabilities.
Top comments (0)