Query layer optimizer for BI tools on Snowflake and BigQuery

The moment this becomes urgent: a data analyst publishes a dashboard their team actually uses, and within two weeks it's so slow that people stop opening it. They know the bottleneck is somewhere between their cloud warehouse and their BI layer — but they don't know exactly where, and fixing it requires either a data engineer's help or weeks of trial and error with query rewrites.

This gap persists because BI vendors like QuickSight and Superset are incentivized to sell you more compute or storage (SPICE capacity, for instance) rather than help you need less of it. The optimization problem sits awkwardly between two teams — the analyst who owns the dashboard and the data engineer who owns the warehouse — and neither has a clean tool that belongs to them for this specific job.

What existing tools get wrong: BI tools show you that a query is slow, but not why in warehouse-specific terms. Warehouse query profilers show you execution plans, but have no concept of a dashboard or how a visualization is hitting the data. The analyst who hears 'optimize your queries to maintain dashboard performance' (a real QuickSight complaint) is being handed a task that requires skills they don't have and tooling that doesn't speak their language.

This is a business and not a feature because the need recurs every time a dashboard gets new users, new date ranges, or new filters added. A team that ships five dashboards a month hits this ceiling constantly. The cost without it is either an engineer pulled off other work to manually profile and rewrite queries, or dashboards that quietly get abandoned because they're too slow to use.

What to build

Build a browser extension and CLI tool that intercepts BI tool queries hitting Snowflake or BigQuery, profiles them against live warehouse metadata, and surfaces specific rewrite recommendations — partition pruning, clustering key mismatches, missing aggregation layers — in plain language tied to the specific dashboard element causing the slowdown.

Where to start

Start with Superset-on-Snowflake deployments at mid-market companies, where the open-source nature of Superset means there's no vendor support to call and the pain is acute and unaddressed — these teams are already comfortable installing tooling themselves.

The hard part

Getting query interception right across BI tools without requiring IT approval or warehouse credential changes is technically brittle, and the first version will likely only work cleanly for one BI-warehouse pairing (e.g. Superset on Snowflake), making early customer acquisition narrow.

How it makes money

Monthly subscription per workspace, priced by number of dashboards monitored, with a free tier for up to five dashboards to drive bottom-up adoption by individual analysts before a team license conversation.