Salesforce Patents a Draw-Your-Query Tool for Data Exploration
What if you could find a dataset by sketching the shape of the trend you're looking for? That's the core idea behind Salesforce's latest patent — turning a freehand drawing into a live database query.
What Salesforce's sketch-to-dataset search actually does
Imagine you're a business analyst who has a hunch that sales dipped in Q2 and then spiked sharply in Q4. Instead of writing a SQL query or clicking through filter menus to find data that matches that pattern, you just draw it — a rough line that dips then climbs — on a blank canvas.
Salesforce's patented system takes that sketch, breaks it down into line segments, and uses those segments to search a database for real datasets whose trends match the shape you drew. The matching datasets are then turned into proper charts and shown back to you on the same screen.
The idea is to make data exploration feel more like sketching on a whiteboard and less like writing code. You express what you're looking for visually, and the system does the translation work to find it in your data.
How a freehand sketch becomes a database query
The system works in a few distinct steps. First, it presents you with a blank drawing canvas inside a user interface — think of it as a whiteboard inside your BI tool.
When you draw a line or shape on that canvas, the system converts your sketch into a set of line segments — breaking a freehand curve into a series of straight-line approximations. For each segment, it extracts parameters like slope, direction, and relative magnitude.
Those parameters are then used to run a query against a database of "linearized data" — a pre-processed representation of your datasets where trends have already been broken down into comparable line-segment form. This is the key architectural trick: by pre-linearizing stored data, sketch queries can be matched efficiently without scanning raw time-series values.
The system retrieves the dimensional datasets (think: sales by region, revenue by product line) whose shapes best match your sketch, and renders them as standard data visualizations — charts, graphs — displayed directly in the UI. The patent also references ad hoc annotation, suggesting users can layer notes or markers onto the canvas alongside their sketches.
What this means for non-technical data analysts
For data teams, the bottleneck in analytics is often formulating the question, not running the query. Sketch-based search flips the workflow: you start with a visual intuition and let the system figure out which data matches it. That's a meaningful UX shift, especially for business users who know what a trend looks like but can't easily write the query to find it.
This fits neatly into Salesforce's broader push with Einstein and Tableau to make data tools more accessible to non-technical users. A sketch-to-query interface could lower the floor for self-service analytics without requiring natural language processing — it's purely gestural, which sidesteps a lot of ambiguity that plagues NLP-based query tools.
This is a genuinely interesting idea that addresses a real friction point in analytics workflows. The pre-linearization architecture is the clever part — without it, sketch matching would be too slow to be useful at scale. Whether it survives contact with messy, high-cardinality enterprise data is the real question, but the approach is sound enough to take seriously.
Get one Big Tech patent every Sunday
Plain English, intelligent commentary, no hype. Free.
Editorial commentary on a publicly published patent application. Not legal advice.