the-data-nerd/vc-deal-flow-signal

Startup GitHub Engineering Velocity Panel

A longitudinal dataset of public GitHub engineering-activity signals for venture-backed startups. It is published under CC BY 4.0 for reproducible research, data journalism, and analysis of alternative data in venture capital.

• 219 startup-period observations
• 55 unique startups
• 18 sectors
• 4 quarterly periods: Q3 2025, Q4 2025, Q1 2026, and Q2 2026
• No missing values in the primary table
• Version: 1.0.0

The 219 rows are startup-period observations, not 219 funding rounds.

Concept DOI: 10.5281/zenodo.19650919
Version DOI: 10.5281/zenodo.19650920
Research paper: SSRN 6606558
Website: https://gitdealflow.com
Live data: https://signals.gitdealflow.com/api/signals.csv
Methodology: https://signals.gitdealflow.com/methodology
MCP server: https://github.com/kindrat86/mcp-deal-flow-signal
MCP install: npx -y @gitdealflow/mcp-signal

Dataset scope

Each observation represents one startup in one quarterly period. The data is derived from public GitHub organization and repository activity. It measures:

• 14-day commit velocity
• Change in commit velocity versus the preceding window
• Contributor count and contributor growth
• Newly created public repositories
• A rule-based engineering signal type
• Broad stage and geography fields

This Hugging Face release is a stable research snapshot. The live GitDealFlow product tracks a larger, changing panel of 350+ startup organizations across 15 current product sectors. The two counts describe different artifacts and should not be merged.

Files and configurations

Config	File	Rows	Description
startup_signals	startup_signals.csv	219	One row per startup and quarterly period.
sector_aggregates	sector_aggregates.csv	72	Sector-level summaries by period.
signal_type_timeseries	signal_type_timeseries.csv	15	Signal-type counts and shares by period.

Primary schema

Column	Type	Description
period	string	Quarterly period slug, such as q2-2026.
sector_slug	string	Sector identifier.
sector_name	string	Human-readable sector name.
startup_name	string	Public GitHub organization slug.
stage	string	Broad funding-stage field.
geography	string	Broad region grouping.
commit_velocity_14d	integer	Commits in the measured 14-day window.
commit_velocity_change_pct	number	Percentage change versus the preceding window.
contributors	integer	Unique contributors in the observation window.
contributor_growth_pct	number	Percentage change in contributor count.
new_repos	integer	New public repositories in the window.
signal_type	string	Rule-based engineering activity classification.
github_url	string	Public GitHub organization URL.

The intended primary key is (period, startup_name).

Load with Datasets

from datasets import load_dataset

signals = load_dataset(
    "the-data-nerd/vc-deal-flow-signal",
    "startup_signals",
)

print(signals["train"].num_rows)
print(signals["train"].features)

Load an aggregate table by replacing the config name with sector_aggregates or signal_type_timeseries.

Methodology

1. Curate venture-backed startups with observable public GitHub organizations.
2. Pull repository activity, contributor statistics, and repository metadata from the public GitHub REST API.
3. Compute recent commit velocity and compare it with the preceding window.
4. Classify the dominant engineering-activity pattern with deterministic rules.
5. Emit one record per startup and period, then produce sector and signal-type aggregates.

See the published methodology for the full definitions and caveats.

Research context

The accompanying study reports an observed 21 to 47 day interquartile range between the engineering signal and subsequent public fundraise announcements in its labeled sample. This is an observational lead-time result, not a causal claim or an investment guarantee. This v1.0.0 dataset does not include ground-truth funding-event labels.

Limitations

• Selection bias: startups without meaningful public GitHub activity are under-represented.
• Private work is invisible: private repositories and private contribution activity are not measured.
• Stage fields are approximate: treat them as broad context, not audited financing records.
• Short windows are noisy: releases, holidays, migrations, and repository moves can create temporary spikes.
• No funding labels in this release: users must join against a separate verified funding-event source for outcome studies.
• Not investment advice: verify every signal independently before making a financial decision.

Citation

@dataset{the_data_nerd_2026_github_velocity,
  author    = {{The Data Nerd}},
  title     = {Startup GitHub Engineering Velocity Panel},
  year      = {2026},
  publisher = {Zenodo},
  version   = {1.0.0},
  doi       = {10.5281/zenodo.19650920},
  url       = {https://zenodo.org/records/19650920}
}

For the research method and lead-time analysis, also cite:

The Data Nerd. (2026). A Longitudinal Panel of GitHub Engineering Velocity for Venture-Backed Startups: Dataset and Early Observations. SSRN. https://ssrn.com/abstract=6606558

License

Creative Commons Attribution 4.0 International (CC BY 4.0). You may share and adapt the data for any purpose, including commercial use, provided you give appropriate credit, link to the license, and indicate whether changes were made.

Contact

Corrections and replication questions: signals@gitdealflow.com