"Data without analysis is just organized noise."
The phrases "data is the new oil" and "data is the new gold" have become mandatory clichés in any corporate event. Everyone repeats them on slides, few understand what they require in engineering, and fewer still do anything practical about it on Monday morning.
In the corporate trenches, I see a constant paradox: organizations that pay exorbitant invoices to maintain gigantic data lakes, modern visualization tools, and respectable data engineering teams, but where the most critical decisions of the month continue to be made based on the executive gut feeling because no one trusts the numbers on the screen.
The problem is not collecting: Ackoff's Pyramid and the cost of bad data
Collecting and storing data has become cheap and technically simple. Integrating APIs, triggering telemetry events, and dumping terabytes of files into cloud storage takes a few days. Today, most companies have data to spare.
The real bottleneck lies in four chronic problems:
- Data of dubious quality at the source;
- Data scattered in silos that do not talk to each other;
- Data without an owner and without clear governance of concepts;
- Data that no one in operations knows how to translate into a decision.
In systems theory, this distinction was immortalized in 1989 by Professor Russell Ackoff in the article From Data to Wisdom, which gave rise to the classic DIKW Pyramid (Data, Information, Knowledge, Wisdom):
- Data: raw symbols and records without inherent meaning (e.g., a table with millions of log lines);
- Information: processed and contextualized data that answers "who, what, when, and where";
- Knowledge: application of information to understand the "how" and see cause-and-effect patterns;
- Wisdom / Decision: practical judgment that uses knowledge to act and change the future of the business.
When a company merely accumulates records on the first step of the pyramid without climbing to reliable information and decision-making, it does not have a gold mine: it has only an expensive digital liability depot.
Even more serious is the effect of the classic computing principle known as GIGO (Garbage In, Garbage Out): if garbage enters at the source, garbage comes out in the report and in the artificial intelligence response. In a study published in Harvard Business Review (2016), expert Thomas C. Redman demonstrated that low-quality data costs about US$ 3 trillion per year to the United States economy alone, because they force analysts and managers to spend up to half their time checking spreadsheets, fixing duplicate records, and reconciling conflicting reports between departments.
A practical way to look at data maturity (the 6 levels)
To diagnose where an operation is stuck, I use these six steps as a practical conversation script (not as a bureaucratic market ranking, but as an engineering compass):
Level 1: Data exists Someone collected the record in some transactional system or spreadsheet, and it is stored somewhere.
Level 2: Data is accessible Whoever needs to investigate a problem can locate and query the database without depending on a three-week technical queue.
Level 3: Data is reliable Source quality is monitored, business concepts are standardized (sales and finance agree on what an "active customer" means), and data lineage is traceable.
Level 4: Data generates insight Frequent analyses leave the paper and dashboards illuminate real bottlenecks in margin, retention, or operational efficiency.
Level 5: Data helps in decisions Following the statistical tradition of W. Edwards Deming (Out of the Crisis, 1986), evidence truly enters the decision table every week, combined with the context and judgment of those who know the operation.
Level 6: Data supports real-time products and services Internal APIs, automations, predictive models, and features of the product itself begin to operate on top of clean and auditable data.
In my practical experience, the most difficult and most valuable leap is between Level 2 and Level 3: moving from simply stored data to data in which leadership truly trusts when allocating money.
What is worth monitoring in the coming years
For those structuring a company's data strategy today, three fronts deserve pragmatic attention:
1. Artificial Intelligence as the analyst's copilot (and not a substitute for critical thinking)
Language models greatly accelerate the writing of SQL queries, catalog documentation, and initial database exploration. However, if the database is dirty at Level 1, AI will only hallucinate pretty answers on top of broken numbers. The human analyst remains indispensable for validating the result, knowing the business rule, and noticing when the question itself is wrong.
2. Clear responsibilities for data (beyond fads)
Decentralized architectures like Data Mesh are an excellent option for large organizations with multiple independent domains, but they are not a mandatory structure for every company. What is never optional is basic governance: knowing nominally who is responsible for the quality of each data at the source and what the rules of use are.
3. Automated quality checks (Data Observability)
Automated integrity tests in pipelines, tracking of data lineage, and anomaly alerts help discover that a table broke at 7 AM, before the director opens the wrong dashboard at 9 AM. But remember: no tool replaces defining what to measure and who takes action when the alert goes off.
My practical advice
Before signing the check for the next trendy data platform, invest in data culture and source hygiene. Standardize the five vital metrics of the company. Teach leadership to ask better questions. Create the weekly ritual of looking at the same single source of truth and demanding action plans based on evidence.
The most expensive tool in the world is useless when no one trusts the number or knows what to ask. Data may even be the new gold, but mining requires method, cleansing, and engineering, not just buying a shinier pickaxe.
Continue deepening: read also Complete Guide to Data and BI for Companies and 2041: the year the Brazilian population begins to shrink. Also explore the practice of Data Engineering and BI and the Case Studies.
References and verifiable readings
- Redman, T. C. (2016). Bad Data Costs the U.S. $3 Trillion Per Year. Harvard Business Review.
- Ackoff, R. L. (1989). From Data to Wisdom. Journal of Applied Systems Analysis, 16, 3-9.
- Deming, W. E. (1986). Out of the Crisis. MIT Press.
Good ideas keep the conversation going.
Share this article or publish the visual deck to your Stories and Feed.

Cesar Zanis
15+ years leading software architecture, database performance, FinOps, and applied AI across enterprise operations. Built and scaled an engineering organization from zero to 20+ developers.
Enjoyed this article?
I publish practical analyses on critical systems architecture, Cloud FinOps, data, AI, and engineering leadership directly on LinkedIn.
Continue Reading

Database Performance and Data Engineering Under Fire: Wait Stats, Indexing, and Zero-Downtime Migrations
1 min read

CDP, Customer 360, and Martech at Scale: Real-Time Identity Resolution with Privacy Governance
2 min read

Modern Data Architecture: Why Data Lakes Turn Into Swamps Without Governance
6 min read
Take this analysis further with data and sources
All visual cards of this article are ready above to share on your Stories, Feed, WhatsApp, or LinkedIn with full sources.
