How to Delegate Investor Research Without Losing Quality

Delegate investor research by defining the decision each record must support, requiring dated evidence and checking a small sample before expanding the assignment. A spreadsheet full of names is not a finished return. The useful output is a reviewable set of possible investors, with knowns, unknowns and reasons for inclusion.

This lesson teaches how to commission and inspect research. For the underlying research task, use the guide to a VC firm's investment thesis. Research does not authorize outreach or establish that anyone will invest.

1. State the decision and exclusions

Describe the company using founder-approved facts: what it does, stage, relevant geography and the type of conversation sought. Identify constraints that would change fit, such as a specific sector, customer model or known portfolio conflict. Separate mandatory criteria from preferences.

Tell the researcher what not to decide. They can identify evidence of fit; they should not mark an investor “will fund us,” infer undisclosed capital availability or guess a private email. Public research and permissioned company context are sufficient starting inputs. Do not share confidential materials simply to make the task feel complete.

Specify source priority: the investor's own thesis, team and portfolio information first, followed by dated original announcements. A third-party directory can provide a lead, but should not silently replace current source verification.

2. Agree what an acceptable record contains

Research acceptance specification

Field	Acceptable return	Reject or hold
Firm and person	Correct identities and current role evidence	Company name confused with a similarly named firm
Mandate match	Source supports required stage, sector and geography, or unknown is explicit	“Technology investor” treated as proof of every stage
Relevant activity	Dated evidence, with limits on what it proves	Old deal called current capacity
Conflict question	Potential overlap flagged for founder review	A definitive conflict ruling without context
Recommendation	Include for review, investigate, or exclude with reason	False certainty or an unsupported numeric fit score

Ask for a concise rationale beside links. A link alone does not explain which sentence supports the decision. Save the evidence date and distinguish a firm's stated mandate from an inference about one investment.

3. Review the method before multiplying it

1. Commission a small calibration set. Include likely fits, a borderline case and an obvious exclusion. The sample size is an operating choice, not a statistical claim.
2. Reproduce the research. Open the sources. Check identity, relevance, dates and whether the cited passage supports the conclusion.
3. Classify the error. Separate inaccurate facts, missing sources, poor fit rules and unclear instructions. These need different corrections.
4. Correct the brief. Update the rule and version. Do not leave the researcher guessing which founder comments apply to the rest of the list.
5. Accept by status. A record can be complete research while the investment fit remains unknown. Research-complete and outreach-approved must stay separate.

The Techstars Mentor Manifesto distinguishes facts from opinions. Apply that discipline to research handoffs: the investigator should expose uncertainty, not hide it to satisfy a quota.

4. Worked example: why 20 names can contain only 12 usable records

Illustrative example: Harbor Sensor delegates 20 records. Its acceptance rule requires correct identity, an accessible supporting source and an explicit decision on the mandatory fit criteria.

The founder reviews all 20 in this first batch. Twelve meet the record standard. Four cite only an aggregator, two confuse a firm's stage focus and two lack current team evidence. The completion rate is 12/20 =60%. This measures research acceptance, not the chance of raising money.

Do not combine that percentage with investor outcomes. The four aggregator-only rows might ultimately be good prospects, but their evidence is incomplete. The stage errors require a method correction across all records, not just the two rows where they were noticed.

Completed correction log

Batch: Harbor research v 1,20 records.
Accepted:12.
Needs primary evidence:4.
Stage-rule error:2; recheck the whole batch against corrected rule.
Role evidence missing:2.
Correction: Stated investment stage must have a source or explicit unknown.
Next return: Corrected 8 rows plus a short method note.
Outreach approval: None implied.
Acceptance reviewer: Founder.

5. Copy the delegation brief

Company brief / approved version:
Decision the research supports:
Mandatory criteria:
Preferences:
Exclusions:
Source priority and evidence date:
Required fields and rationale:
What must remain unknown when unsupported:
First calibration sample:
Reviewer / feedback deadline:
Acceptance states:
Error categories:
Brief version / method correction:
Who approves any external action:

When quality slips

• Rows get longer but not clearer: Require a short source-to-criterion rationale instead of copied biographies.
• AI produces plausible citations: Open every cited source during calibration; unverified citations do not count.
• The sample passes but later work drifts: Review new batches and edge cases; do not treat one sample as permanent assurance.
• The founder changes the target midstream: Version the brief and identify which earlier rows need re-review.

Completion check: A reviewer can reproduce the inclusion rationale, identify unknowns and trace corrections. That is a research system you can delegate without pretending that research replaces investment judgment.