From your data to intelligence.
A condensed view of the platform: connect, organize, transform and put data to work.
Six topics, one journey.
Bring the right data, at the right frequency.
Lakehouse, Catalog and layers.
Teach what the business means.
Model what has become routine.
Take data to decisions and action.
Operate with confidence and control.
From source to action, one integrated architecture.
The question comes before the connector.
Connecting everything increases cost and complexity. Start with the sources and tables that answer the first use case.
What can come in
- Applications
CRM · Ads · ERP - Databases
Postgres · MySQL - Files
CSV · Sheets · S3 - Unstructured
PDF · text · images
Four decisions
- Streams
which tables? - Fields
what's needed? - Sync
full or incremental? - Cadence
when does it need to update?
Media shows the click. CRM shows the revenue.
The value appears when different sources start to answer the same question together.
Bronze preserves. Silver organizes. Gold serves.
The lakehouse keeps history and separates data as it arrived, as it was cleaned and as it will be consumed.
Bronze
Faithful copy of the source and the basis for traceability.
facebook_ads_campaignsSilver
Types, names, keys, deduplication and reusable rules.
campaigns_cleansedGold
Table designed for a decision, report or activation.
marketing_daily_performanceThe Catalog explains the data. Context explains the business.
Description in the Catalog
- Table
assinaturas - Column
status - Granularity
1 row / subscription - Relationship
customer_id
Context document
- Active customer
active contract - Revenue
confirmed payment - Churn
canceled in the period - Exceptions
tests and internal excluded
One definition prevents four answers.
When “revenue”, “active customer” and “conversion” have an explicit rule, people, dashboards and agents use the same meaning.
- Without context: each consumer interprets columns and exceptions.
- With context: the rule travels with the data.
- Result: less rework and more reliable answers.
## Recurring revenue MRR counts only active subscriptions with a confirmed payment in the month. Exclude internal accounts and test environments. Official source: Gold.mrr_monthly
Query when the data is enough. Notebook when code comes in.
| Tool | Best for | Examples | Result |
|---|---|---|---|
| Query SQL | Filters, joins, aggregations and relational rules. | Funnel, income statement, cohort, daily metrics. | A new table in the Catalog. |
| Notebook | Python, PySpark, external APIs, files, AI and ML. | Forecast, enrichment, scoring, classification. | One or more consumption-ready tables. |
Rule of thumb: start with SQL. Move to a notebook when logic, volume or integration truly calls for it.
If the question repeats, the answer should become a model.
- Input: cleaned campaigns in Silver.
- Rule: standardize channel and sum by day.
- Output: a stable Gold for BI, MCP and activation.
- Trigger: run after the sources update.
SELECT date, channel, SUM(investment) AS investment, SUM(leads) AS leads, SUM(revenue) AS revenue FROM Silver.marketing_unified GROUP BY date, channel
Notebooks turn data into new products.
Deal history becomes probability by stage and monthly scenarios.
Conversations become topics, recurring questions and documentation gaps.
Companies and addresses receive attributes from specialized APIs.
The pattern repeats: read from the Catalog, apply specialized logic and save a table the rest of the platform can use.
Data needs to reach where the decision happens.
Data MCP
Agents query the Catalog and context in natural language.
Data API
Applications access tables programmatically.
Destinations
Clean data flows back to CRM, ads and operations.
BI
Dashboards read the same Gold used by other channels.
Configure once. Roll out agents with control.
Example: Claude analyzes all the history available in Nekt and, via the Gateway, queries or takes actions in the connected tools, always within the permissions set by the organization.
Idle Gold is just a potential report.
Activation turns the pipeline into a cycle: data drives action, action generates new data, and the model improves.
The CRM works better when it knows the real revenue.
Confidence means knowing what ran, what failed and who has access.
Status and history of each execution.
Technical detail and failure summaries.
The team is notified before the consumer.
Runs on a schedule, by event or on demand.
Access by user, group and resource.
From the Gold table to the original source.
A playbook connects acquisition, revenue and retention.
The architecture repeats. Context changes per client.
- Catalog per client for external sources and deliverables.
- Internal catalogs for operations, benchmarking and management.
- Isolated Bronze preserves each source without mixing accounts.
- Standardized Gold compares channels and clients with the same rule.
The first use case doesn't need to model the entire company. It needs to produce a reliable answer and reach whoever will act on it.
academy.nekt.com