Key takeaways
- Applications arrive through your branded portal, private referral links or records analysts create, so every deck enters one queue (Pilot).
- Every extracted fact links to its slide, and slides that couldn't be read are flagged rather than guessed.
- A missing value is Unknown and becomes a question. A mismatch goes to a person, never to an automatic rejection.
- Thesis fit, evidence quality and diligence readiness are shown as three separate readings, not blended into one score.
Planned feature: not in Delsivo yet, and details may change before it ships.
Why inbound screening falls behind at a lean fund
Inbound reaches a small fund through a website form, forwarded emails, warm introductions and decks attached as PDFs or links. Each lands somewhere different. Analysts end up reading in arrival order rather than fit order, and the reason for a pass lives in someone's memory or in a spreadsheet cell nobody reads again.
Volume is only half the problem. The other half is that a first review produces no reusable record. The facts read off the deck, the criteria checked and the reason for the decision are lost, so when the same company applies a year later, or a partner asks why it was passed, the work starts again.
How to screen inbound startup applications in Delsivo
- 1Publish your thesis as a rubric: hard rules such as stage, sector, geography and round size, weighted preferences, exceptions and the questions to ask when a deck is silent (Pilot).
- 2Share your portal link on your website and send private referral links to co-investors and portfolio founders. Founders save and resume, then get a receipt and a secure link. Analysts create records for decks that arrive by email (Pilot).
- 3Each PDF deck is read: slide text, legible charts and structured company facts, every fact citing its slide. Weak extraction is flagged (Pilot).
- 4Your rubric is applied. Each criterion shows the rule, the value found, the reasoning and the source, with a result of Met, Mismatch or Unknown (Pilot).
- 5The deal joins the analyst queue in fit order. Possible duplicates are suggested by company domain and founder identity, and an analyst confirms any merge (Pilot).
- 6An analyst advances the deal, passes with a reason or sends a short form about the unknowns. A partner approves every message before it reaches the founder (Pilot).
What each criterion result means
| Result | What it means | What happens next |
|---|---|---|
| Met | The deck satisfies the rule, and the slide is cited | Counts toward thesis fit |
| Mismatch | The deck explicitly falls outside the rule | Surfaced for an analyst to review, never rejected automatically |
| Unknown | The deck doesn't say | Becomes a question for the founder, never a failure |
| Override | An analyst disagrees with the suggested result | Allowed, with a recorded reason, and the original stays visible |
Re-running a screen after a rubric change creates a new run and keeps the old one, so you can see what the change did.
Three readings instead of one score
A single score hides the reason behind it. A deal can fit your thesis well and still have almost no evidence behind its numbers, or have a full data room and sit at the edge of your mandate. Delsivo shows thesis fit, evidence quality and diligence readiness separately, so an analyst can see which of the three is holding a deal back.
Weighted preferences still order the queue. The order is a suggestion with its working shown, and analysts are expected to disagree with it when the reasons look wrong.
Limits in the pilot
- PDF decks only at first. PowerPoint follows once extraction quality is measured on real decks.
- English first. Decks in other languages can be uploaded, but extraction is designed and checked for English.
- Pitches forwarded to a shared mailbox don't become deals on their own yet. An intake mailbox is planned for later.
- No CRM import or export in the pilot. That is planned for a later release.