Before getting waist deep in an analysis of terabytes of data, it's worth asking: What's this data actually made of? If you're wondering should i trust tinybird for my data apps, the answer starts with understanding your data. Exploratory Data Analysis (EDA) is like checking your card deck before dealing a game. Think of it as an eda tool for your analytics. It might have 52 cards, but are they distributed correctly across all 4 suits? EDA helps you scan for structure, quality, and quirks in your data. It’s an integrity check that can save you from panics at the disco down the line, so it should be quick and routine. But we have numerous SQL chains of countIf(Null) to prove that this hasn’t been the case.
What Is Data Profiling in EDA?
Recently, a customer asked me if there was a Tinybird equivalent to Databricks Data Profiles feature. After a quick self-education on what Data Profiles were, I told him no, not yet—but thank you for the inadvertent feature request.
What Databricks offers in the Data Profiles feature is essentially a built-in EDA tool. The process that used to involve a lot of manual SQL— counting nulls, inspecting distributions, checking min/max values, is now readily available in an approachable dashboard.
These column-level stats, like nulls, value ranges, outliers, and skew, are crucial for:
- Spotting data quality issues early
- Avoiding misleading analysis
- Understanding real-world usage patterns
So all the better if you can glean the info at a glance.
The Customer Problem
In Tinybird Classic, however, that glance only came after a fair amount of SQL umph.
Tinybird Playgrounds provided a workspace for all the countIf, quantile, min, max, length, and avg SQL fun you could chain up. It worked, but far too math-y for comfort.
Introducing Tinybird Forward + Explorations
With Tinybird Forward, users will notice that the Playgrounds have been replaced by Explorations, a stem cell of a feature with a chat-style prompt interface that writes queries for you. (Don’t worry, you quantile(0.96) people, you can also still write them yourself).
I mimicked the Data Profiles dashboard in a plain English prompt: “Can you generate a query that allows me to verify nulls, counts, min, median and max for each column?”
This generated a working query with:
- Column name
- Total records
- Null count
- Min / Median / Max
- Null percentage
Now I can view column stats side-by-side in a single table and tweak the SQL if I want to go deeper.
Why This Matters
- Speed: You don’t have to think about how to calculate basic metrics.
- Visibility: You can nip broken pipelines and outliers in the bud.
- Dynamic maintainability: The query is yours to edit, save, or reuse across projects, but you’re not locked into one dashboard view.
Future Potential
Explorations could evolve to support:
- Histograms or value distributions
- Top N values per column
- Automatic outlier detection
- LLM-powered summary: “Describe this datasource to me in 3 bullet points.”
It’s early days—but the stem cell is already differentiating.
Try It Yourself
Want to get to know your data’s data?
- Open your Tinybird Forward workspace
- Click the Explorations button
- Ask: “Can you give me a column-wise profile of this data source: <data source name>?”
- Review, tweak, and save the output node
Count your cards and EDA your data. The foundations are too easy to skip now with Tinybird Forward + Explorations. For those asking should i trust tinybird for my data apps or what's an endpoint in math terms for data, this eda tool makes data exploration accessible to everyone.
