Turn an investment thesis into a ranked startup shortlist

The goal

Start with a thesis, not an endpoint

Most scouting workflows begin as a sentence. The useful work is turning each part of that sentence into a filter the API can apply consistently.

This guide resolves the three human-readable taxonomy values at runtime, sends one company list request, and sorts the result by signal score. The live query currently finds 16 companies and returns the top eight.

Step 1

Authenticate once, then let the client refresh

Create a Programmatic M2M key in Dealroom and keep both credentials on the server. The downloadable script uses OAuth2 client credentials and retries once with a fresh token after a 401 response.

DEALROOM_CLIENT_ID=your_client_id
DEALROOM_CLIENT_SECRET=your_client_secret
DEALROOM_USER_AGENT=your-company-thesis-shortlist/1.0

Install the three small dependencies used by the example:

pip install authlib requests python-dotenv

Do not put a Programmatic M2M secret in browser code. Run this script on a server, in a scheduled job, or locally.

Step 2

Resolve names to taxonomy IDs

Filters use stable IDs, but a thesis uses names such as Europe, Industrial Automation, and Artificial Intelligence. Resolve those names before constructing the filter instead of copying IDs into your application.

def exact_value(items, label, source_type=None):
    for item in items:
        item_label = item.get("label") or item.get("name", "")
        item_type = item.get("source_type") or item.get("type")
        if item_label.casefold() == label.casefold() and (
            source_type is None or item_type == source_type
        ):
            return str(item.get("value") or item["id"])
    raise LookupError(f"Could not resolve {label!r}")

europe_id = resolve_location("Europe", "continent")
industrial_id = resolve_tag("Industrial Automation", "sector")
ai_id = resolve_tag("Artificial Intelligence", "technology")

Location discovery returns name/id. Cross-filter search returns label/value. The helper accepts both response shapes, then requires an exact case-insensitive label and the expected taxonomy type.

Step 3

Translate each thesis clause into a filter

The filter DSL keeps the query readable. Each clause corresponds to one decision in the thesis.

filters = [
    f"tag_id[in_all]:{industrial_id}|{ai_id}",
    f"hq_location[eq]:{europe_id}",
    "is_startup[eq]:true",
    "company_status[eq]:operational",
    "launch_date[gte]:2020",
    "total_funding[lte]:50000000",
    "signal_rating[gte]:65",
]

company_filter = f"and({','.join(filters)})"

The in_all operator matters here. It requires both tags. Using in_any would also return general AI companies and industrial automation companies without AI.

Step 4

Rank on the server

Ask the list endpoint for the top eight records ordered by descending signal score. Set include_total=true so the response also tells you how many companies matched before the limit was applied.

result = client.get(
    "/data/companies",
    params={
        "filter": company_filter,
        "sort": "-signal_rating",
        "limit": 8,
        "include_total": "true",
    },
)

print(result["page"]["total"])
print(result["data"][0]["name"])

Company rows already include the fields needed for a first-pass shortlist, including the tagline, headquarters, founding year, funding summary, team size, tags, hiring status, founders, and signal score.

The output

Snapshot generated 2 Sep 2026. Funding values are shown in USD. Dealroom signal scores are discovery aids, not investment recommendations.

Next steps

Turn the shortlist into a workflow

This example stops at a ranked list. A production scouting workflow can add a small amount of state and human judgment:

• Run the query on a schedule and store the company UUIDs you have already reviewed.
• Flag newly matched companies and meaningful changes in funding, headcount, or hiring.
• Add your own score for thesis fit instead of treating the Dealroom signal score as a final decision.
• Write analyst notes and decisions to your CRM, database, or internal scouting tool.