Every screening process involves judgement, and judgement is shaped by experience, habit and first impressions. In venture, where decisions are fast, information is thin and most deals arrive through personal networks, those impressions have plenty of room to work. The result is that founders with the same quality of business can get very different treatment depending on how they look, sound or present.
This is a process problem as much as a personal one. Bias in startup screening is easier to reduce by changing how decisions are made than by asking people to try harder to be fair. The sections below summarise what peer-reviewed research has found, then set out the practical changes a fund or programme can make, with honest limits on what any process or tool can do.
What research has found about bias in startup screening
Two studies are widely cited because they isolate the effect of who is pitching from what is being pitched.
Brooks, Huang, Kearney and Murray, writing in the Proceedings of the National Academy of Sciences in 2014, studied three entrepreneurial pitch competitions in the United States and ran two experiments. They found that investors preferred pitches presented by male entrepreneurs over pitches made by female entrepreneurs, even when the content of the pitch was the same, and that physical attractiveness increased persuasiveness for men but made no difference for women. In one experiment, reported by Harvard Business School Working Knowledge, the same narration was dubbed with a male or a female voice. 68.33% of participants favoured the ventures pitched with a male voice.
Kanze, Huang, Conley and Higgins, in the Academy of Management Journal in 2018, analysed question-and-answer sessions at TechCrunch Disrupt New York between 2010 and 2016. Investors tended to ask male entrepreneurs promotion-focused questions, about gains and growth, and female entrepreneurs prevention-focused questions, about losses and risks. Founders tended to answer in the same frame, and those asked promotion-focused questions raised significantly more funding. The questions asked accounted for the effect of gender on funding.
The second finding is especially useful for process design, because it points to something a fund controls directly: the questions it asks.
Where bias enters the screening process
Bias is not one moment. It can enter at every stage, and each stage has its own practical safeguard.
| Stage | How bias can enter | Practical safeguard |
|---|---|---|
| Sourcing | Most deals come through networks that resemble the team | Run an open application route and track where deals come from |
| Application form | Questions reward founders who know venture jargon | Plain-language questions with word limits and examples |
| First read | Names, photos, schools or accents shape the first impression | Read the business answers before founder details where you can |
| Criteria | Vague standards like "strong team" let instinct fill the gap | Written, checkable criteria agreed before reading |
| Missing information | Short or unconventional applications are treated as weak | Mark missing facts as unknown and ask, rather than fail |
| Interviews and calls | Different questions for different founders; polish mistaken for substance | Same core questions for every founder, assessed on content |
| Discussion | The first or most senior opinion anchors the rest | Independent scores before discussion, then debate |
| Decision | Reasons are not recorded, so patterns are invisible | A short written reason for every advance and pass |
Structure the process before you read
Evidence from hiring supports structured assessment. Google's re:Work guide to structured interviewing describes using the same questions and standardised rubrics for every candidate, and cites research finding that structured interviews increase predictive validity and decrease differences between demographic groups. Startup screening is not hiring, but the same mechanisms apply.
- 1Write criteria first. Turn your thesis into hard rules, weighted preferences and questions before the cycle starts. Our guide on how to write an investment thesis shows how.
- 2Describe good answers. For each criterion, write what a strong and a weak answer looks like, so reviewers share a standard.
- 3Separate fit from quality. Check eligibility rules mechanically, then judge quality only for companies that fit.
- 4Score each criterion separately before forming an overall view, so one impressive or off-putting detail does not decide everything.
- 5Collect independent scores before any discussion, and discuss the largest disagreements rather than averaging them away.
- 6Record a reason for every decision, in a sentence or two, where the team can see it.
Ask every founder the same core questions
The Kanze research suggests a direct change: decide your core questions in advance and ask them of every founder, in the same frame. If you ask one founder how big the company could become, ask every founder. If you ask one founder how they will avoid running out of money, ask every founder.
- Write a short set of core questions for first calls and interviews, balanced between growth and risk.
- Add follow-ups based on the founder's own deck and claims, not on assumptions about who they are.
- Assess the substance of each answer and the evidence behind it.
- Exclude accent, fluency, presentation polish, appearance and background from any assessment. None of them is evidence about the business.
- Give founders the same time and the same format, and offer a text option for anyone who prefers to write.
Our guide on questions to ask founders has a question set built this way, and our guide on asynchronous founder interviews explains how a fixed-format interview, live with an AI interviewer or written, can help.
Check your own decisions, and know the limits
A process only helps if you look at what it produces. Once a quarter, review your pipeline by source and by stage. Where do applications from open routes drop out compared with referrals? Which recorded reasons for passing appear most, and are they really about the business? Where did reviewers disagree most? You do not need demographic data to start. Reasons, sources and stage conversion already tell you a lot.
Be honest about the limits. No process, scorecard or software makes screening free of bias, and claims that a tool does should be treated with suspicion. AI systems can reflect patterns in the data they learned from, and a model scoring founders on presentation could amplify the very effects the research describes. The safest use of AI in screening is narrow: extracting facts with sources, applying criteria you wrote, and flagging what is missing, with people making every decision and able to see why each suggestion was made.
Common mistakes: training sessions with no change to the process, blind review in the first round followed by unstructured calls, treating "culture fit" as a reason without explaining it, and measuring nothing.
Where Delsivo fits
Delsivo is designed around these safeguards. It does not make screening fair on its own, and no tool can. In the pilot release, your thesis is written as versioned rules and preferences before screening. Each criterion shows the rule, the value found, the source and the reasoning. Missing information is marked Unknown and becomes a question, never a failure, and a mismatch is surfaced for a person to review, never an automatic rejection. Overrides need a written reason.
Live AI interviews, planned for the first release, give every founder on a template the same core questions, persona and time plan, and a text interview under the same rubric is always available. The AI assesses only the transcript: the substance of each answer and the evidence behind it. Accent, fluency, presentation polish, appearance, background and the video and screen images are excluded, and speech metrics such as pace or filler words are not computed for investors. Analysts can override any assessment with a reason, and people decide every advance and pass. See founder interviews for the planned design.
Sources
- Brooks, Huang, Kearney and Murray: Investors prefer entrepreneurial ventures pitched by attractive men (PNAS, March 2014), MIT DSpace record (checked 6 November 2026)
- Harvard Business School Working Knowledge: Venture investors prefer funding handsome men (30 April 2014) (checked 6 November 2026)
- Kanze, Huang, Conley and Higgins: We Ask Men to Win and Women Not to Lose (Academy of Management Journal, 2018), London Business School research record (checked 6 November 2026)
- Google re:Work: A guide to structured interviewing for better hiring practices (checked 6 November 2026)
Delsivo publishes this guide and makes software for venture funds, so we have a stake in the topic. Facts about other companies, research and regulations come from the sources listed, on the dates shown, and can change. Corrections are welcome through our contact page.