I’ve been working with diverse data for 20 years. Over the past 6 years, my focus has been on building data integrations, data warehouses, and data products.
Every project follows a similar pattern: the client invests significant resources to bring all their key business data into one place—cleaned, organized, high-quality. Then we build the reports, hand over the access, and say: “Here it is. Enjoy.”
It’s like that old joke:
Step 1: Build a data warehouse.
Step 2: ???
Step 3: Profit.
For me, the real challenge has always been Step 2 – turning data into profit. Ideally automatically, to maximize the return on investment from a data warehouse.
From Reports to Proactive Intelligence
While exploring this challenge, I came across the concept of augmented analytics, an area now rapidly evolving thanks to AI and LLMs.
I see two main approaches:
- Reactive – You ask an intelligent system a question in natural language, and it gives you an answer. This can look impressive, but it comes with weaknesses:
- Conceptual confusion (who is a customer? what is a product?).
- You need to know that you should ask in the first place.
- Proactive – The system tells you when something important is happening.
That’s my vision.
The Key is in Anomalies
When reports don’t change, I stop looking at them. Time should be spent only when something changes—that is, when anomalies appear. But anomalies are not one-dimensional. Their meaning depends on context:
- Type: gap, spike, trend change, cumulative shift, change in data character.
- Scope: company-wide vs. subsegment (a single product in one region).
- Significance: a small change in a key category can matter more than a big change in a marginal one.
- Frequency: a first-time occurrence vs. a recurring, well-known event.
- Evolution: today’s spike might be tomorrow’s trend; today’s outage might be resolved data tomorrow.
- User perspective: a CEO’s priorities differ from a local manager’s.
- Aggregation: a drop in sales for 2 products in one region is more valuable insight than 2 unrelated alerts.
- Prioritization: handle the biggest business impacts first.
- Speed of reaction: learning about an issue 3 months later is often too late.
- Quantification: immediate access to numbers, duration, affected dimensions, comparisons to history.
From Principles to Product
From these ideas, I developed AnomalyGuard—a tool that:
- Automatically pulls data from your systems (databases, CRM, data warehouse).
- Analyzes them based on your requirements (configurations).
- Detects and aggregates anomalies.
- Sorts them by significance and delivers them to the right people.
- Displays them in the app or sends them into your reports and internal tools.
If this perspective on business data resonates with you, check out our offering or get in touch. We’d be happy to talk about your data—and how to turn them into real business profit.
Subscribe to Our Newsletter