Explainable company scoring
Enter the company's parameters. The model computes a ranking score for investment attractiveness and shows which features drove the result (SHAP analysis).
Input
Company details
We read the site and extract the model's features. Works for non-Russian companies too.
Adds verified registry data. A non-Russian company has no EGRUL record — leave this blank.
—
| Phrase | Volume* | Demand | Competition |
|---|
average number of searches per month.
Values tagged "AI guess" and "domain heuristic" are guesses, not verified facts. The model extracted them from the site's text. Check them before relying on the score, and correct them by hand where needed.
Structured features
Result
Company rank (percentile, 0–100)
The rank is the company's place among funded Crunchbase companies of the same age that the model never saw in training: 80 means the profile scores above 80% of them. The model is trained to tell companies that reached an acquisition or IPO from the rest and compares profiles with each other — it is not a probability of success. Rounds, funding, milestones and relationships count in the company's favour: the more of them, the higher the rank. Each feature's contribution is shown below.
Explanation
Feature contribution (SHAP, log-odds)
Green bars raise the final score, red bars lower it. The values are dimensionless (log-odds): the base value plus the sum of every bar reproduces the model's final logit exactly.
* — this feature was left blank. The bar shows how the model handles a missing value, not data you entered.
Details
Factor breakdown
| Feature | Value | SHAP |
|---|
Rows marked "not specified" are features you left blank. Their contribution reflects how the model handles a missing value, not a value you entered.