Ran both Sigma and Lightdash on a standard dbt + Snowflake stack. Tested: model selection, metric definition, dashboard build speed, and query performance.
**Setup:**
* dbt Cloud with 15 core models (dimensions, facts).
* Snowflake X-Small warehouse.
* Test dataset: ~5 million rows in primary fact table.
* Connected both BI tools directly to the dbt-managed schema.
**Findings:**
* **dbt Integration:**
* Lightdash: Native. Reads `schema.yml` directly. Metrics defined in dbt.
* Sigma: Uses "Workbook" paradigm. Must map to dbt models manually. Less tight sync.
* **Metric Definition:**
* Lightdash:
```yaml
metrics:
- name: total_revenue
label: Total Revenue
model: ref('fct_orders')
calculation_method: sum
expression: revenue
timestamp: order_date
time_grains: [day, week, month]
```
* Sigma: UI-based aggregation on a joined dataset. No version-controlled metric layer.
* **Performance (Avg Query Time - 5 user concurrency):**
* Simple aggregation (day grain): Lightdash ~1.2s, Sigma ~1.8s.
* Complex join with 3 dimensions: Lightdash ~3.5s, Sigma ~4.9s.
* **User Experience:**
* Lightdash: Feels like a BI wrapper for dbt. Limited visualization polish.
* Sigma: Spreadsheet-like interface is powerful for ad-hoc. More chart types.
**Verdict:** For a pure "lightweight BI on dbt" use case, Lightdash is more integrated and faster to deploy. Sigma offers more analytical flexibility but adds a layer of abstraction from the dbt project. Choose based on whether you want a metric layer extension (Lightdash) or a separate analysis canvas (Sigma).
Benchmarks don't lie.