The pitch on ScienceSheet's homepage is aimed at a specific person. The business analyst whose week is consumed by searching, cleaning, and organizing data before a data scientist can touch it.
The Palo Alto company's argument is that Excel formula syntax is, in practical terms, one of the most widely used programming languages on earth. The analyst already fluent in it should be the one staging data for machine learning rather than waiting three to six months for a traditional data science process to come back with an answer [ScienceSheet]. That is the wedge. ScienceSheet wants the spreadsheet to be where ML prototyping starts.
The bet
Founded in 2019 and headquartered at 228 Hamilton in Palo Alto [LinkedIn], ScienceSheet sells what it describes as a Spreadsheet AI that generates ML apps [Crunchbase].
The more concrete product claim, drawn from the company's own site, is that an analyst can use ScienceSheet to generate data features and prototype an open-source model on behalf of an in-house data scientist [ScienceSheet]. A second product line, Sparksheet, is positioned as a bridge between the two roles the company believes actually create machine learning artifacts in the enterprise. Data scientists and business analysts [ScienceSheet].
The ideal customer profile is narrow and worth naming plainly. Mid-market and enterprise analytics teams that already run on Excel. Have at least one data scientist on staff. Have a backlog of feature engineering and model prototyping work that the data science team cannot service fast enough.
The buyer is most plausibly a director of analytics or a head of data science. The budget owner is the same. The procurement cycle is the standard SaaS annual contract with a security review attached.
Why it could be big
The shape of the market favors tools that meet analysts where they already work. Microsoft has spent the last two years pushing Copilot into Excel.
The broader category of AI-assisted data preparation has attracted serious capital. This because the underlying pain (analysts spending the majority of their time on data prep before any modeling happens) is durable and well-documented.
ScienceSheet's framing, that the heavy lifting of data prep for ML can be done inside a familiar grid, is consistent with how enterprise buyers have historically adopted analytics tooling. Through the spreadsheet first, the notebook second.
If the company can demonstrate that an analyst-generated feature set and prototype model is good enough to hand to a data scientist for production hardening, the time-to-first-model compression is the sales pitch that travels.
The company is actively fundraising, with a public investor request page on its own site [ScienceSheet]. Third-party databases place headcount at 10 to 19 employees per ZoomInfo and 11 to 50 per LinkedIn's self-reported band. ZoomInfo estimating revenue between $1M and $5M [ZoomInfo, LinkedIn].
Those are estimator-grade figures rather than audited disclosures. They should be read as a range, not a fact.
| Signal | Value | Source |
|---|---|---|
| Founded | 2019 | PitchBook |
| Headquarters | Palo Alto, CA | |
| Headcount band | 10 to 19 | ZoomInfo |
| Headcount band (self-reported) | 11 to 50 | |
| Revenue band (estimated) | $1M to $5M | ZoomInfo |
| Disclosed funding | Undisclosed | Crunchbase |
The team and traction
Publicly captured sources do not name the founding team in the structured record this article is built from. The most useful thing to report is what the company itself has put on the record.
A Palo Alto address. A partner program with a published contact at sales@sciencesheet.com. An EULA that contemplates evaluation access via AWS as a precursor to a paid subscription [ScienceSheet].
The AWS evaluation path is a meaningful detail. It signals that the company is structured to land via cloud marketplace motion. This is increasingly the default procurement route for data and ML tooling inside large enterprises. Buyers draw down committed cloud spend rather than open a new vendor line.
The honest counterfactual
What bears will say is straightforward. The competitive set around analyst-facing AI for data prep and ML prototyping is real and well-funded.
Microsoft Copilot in Excel is the default that every buyer will compare against. Simply because it ships inside the tool the analyst already opens every morning.
Beyond Microsoft, the realistic competitive set includes Google's Gemini in Sheets. Einblick (acquired by Databricks). Hex's notebook-plus-spreadsheet hybrid. Sigma Computing for cloud-warehouse-native analytics.
The long tail of AutoML platforms (DataRobot, H2O.ai, Dataiku) target the same backlog from the data scientist's side rather than the analyst's.
To win, ScienceSheet has to be meaningfully better at the specific handoff between analyst-built features and data-scientist-owned production models than a Copilot prompt is. It has to prove that case to a procurement organization that already pays Microsoft.
What bulls answer is that none of the incumbents are organized around the handoff itself. Copilot accelerates the analyst. AutoML accelerates the data scientist.
The friction in the middle (reproducible features, an open-source model the data scientist will actually accept) is a real seam. ScienceSheet's Sparksheet positioning is aimed exactly at that seam [ScienceSheet]. A small focused company can ship to a seam faster than a platform vendor can.
What to watch
The next twelve months should answer three questions a buyer would actually ask. First, does ScienceSheet close a priced seed round with a named lead? The investor page is live and the company is openly fundraising [ScienceSheet].
Second, does it land on the AWS Marketplace with a published listing? This would convert the EULA's evaluation language into an actual procurement path [ScienceSheet].
Third, and most important for anyone underwriting the business, does it begin to disclose customer logos, seat counts, or net revenue retention? Growth charts without retention numbers are not a story. Renewal cohorts are.
ICP, stated plainly: analytics teams of 20 to 200 inside mid-market and lower-enterprise companies. Excel-native. With at least one data scientist on staff and a documented backlog of feature engineering work.
Realistic competitive set: Microsoft Copilot in Excel, Google Gemini in Sheets, Hex, Sigma, Einblick (Databricks). The AutoML incumbents (DataRobot, H2O.ai, Dataiku) approaching the same problem from the modeler's side.
Pipe Haddad covers enterprise and SaaS for Startuply.