# The Demo Day Batch Triage Playbook: Cohort List to a Sized Meeting Slate

*Take a full demo day cohort through six timed filters and leave with a Tier 1 slate sized to your meeting capacity, a Tier 2 list, and a 4, 8, 12 week watchlist.*

- Canonical URL: https://www.refolk.ai/guides/demo-day-batch-triage-playbook
- Pillar: Investing and deal sourcing
- Format: Playbook
- Published: 2026-10-31
- Last reviewed: 2026-10-31
- Reading time: 21 min
- Keywords: how to triage a demo day batch, accelerator demo day shortlist for investors, how to prepare for demo day as an investor, screening accelerator cohort companies, demo day companies that did not raise follow up

## Key takeaways

- In the Gompers et al. survey of 885 VCs at 681 firms, each closed deal took about 28 management meetings, so a Tier 1 demo day slate should be sized at roughly 28 first meetings per check you intend to write.
- A 15-company demo day slate carries about 0.54 expected closed deals at the survey's conversion ratios, so booking 15 meetings and expecting a deal is a capacity error.
- A 5% weekly growth claim compounds to 1.80x over a 12-week batch but reads as 12.6x when annualized, so ask for the absolute base and recompute the 12-week multiple before trusting a one-chart pitch.
- One VC described companies whose contracted revenue figure ran 70% higher than their actual recurring revenue, so every demo day ARR number needs to be split into collected, contracted, and run-rate before the meeting.
- Start the triage clock at the cohort announcement, not the demo day date, because the most promising companies finalize term sheets in the days and weeks before the event.
- Michael Seibel said there is little correlation between the most in-demand demo day companies and the ones that become extremely successful, which makes the long tail a value pool worth re-ranking at 4, 8, and 12 weeks.

This playbook is for early-stage investors, platform and talent partners at funds, and angels who have just received an accelerator cohort of 150 to 250 companies, all raising on the same calendar. It runs that list through six filters, cheapest first, on a clock that starts four weeks before demo day. At the end you have three things: a Tier 1 slate sized to the meetings you can actually take, a Tier 2 list, and a long-tail watchlist with re-rank dates at 4, 8, and 12 weeks.

## What does demo day batch triage have to produce, and by when?

Batch triage produces three lists: a Tier 1 slate you book on demo night, a Tier 2 list you meet if slots open, and a dated watchlist for everything else that fits your thesis. The cheap filters have to be finished before demo day, because the strongest rounds close within days of it.

### This is a YC-scale job

Only Y Combinator runs cohorts this large. Past YC batches have presented 149 and 189 companies, and one ran to nearly 400. Techstars classes are a different shape. They run about 5 to 11 companies (London had 11, Boulder 9), and Techstars invests $220,000 in each. With a class of 11 you can diligence every company and don't need a funnel. YC has moved to four cohorts a year, so if you invest at this stage the job below comes around every quarter.

### When the rounds close

No one has published a time-to-raise breakdown by program or by cohort size. What exists is a set of YC-specific claims, and they mostly agree.

| Window after demo day | Claim | Source |
|---|---|---|
| 48 hours to 2 weeks | Top 10 to 20% close | Single-accelerator guide |
| About 1 month | "Vast majority" finish | Axios |
| 1 to 2 months | "Vast majority" complete | Michael Seibel, via TechCrunch |
| 4 to 8 weeks | "Most close" | Founder-side guide |
| Any time after | About 80% of one year's companies | YC spokesperson, via Newcomer |

Read the table as directional. One unaffiliated site splits a cohort 40/30/20/10 (raised by 4 weeks, by 8 weeks, by 6 months, never). It states no method, so I don't use it.

### When the list becomes visible

This is disputed. YC lets approved investors preview companies through its startup directory before demo day. One unaffiliated site puts that preview at 24 to 48 hours before the pitches. Another source says company lists and materials are visible weeks early and the meeting calendar fills before the event. Axios reports that unofficial lists of the companies and their fundraising have long circulated among investors before demo day. No one has published how many days before demo day the calendar fills.

> **Rule:** Start the clock at the announcement, not at demo day
>
> The most promising companies finalize term sheets in the days and weeks leading up to the event. Date your triage from the day the cohort becomes visible to anyone, including through unofficial lists, and never from the demo day date.

### Why the queue is crowded

At one YC demo day, more than 1,500 investors tuned in and generated almost 50,000 investment leads across nearly 400 companies. That works out to roughly 125 meeting requests per company (my derivation). One founder-side guide tells founders to block 25 to 40 investor meetings across the two weeks after demo day. So most requests are never accepted. That is why triage has to be finished before demo night and not done during it.

You also aren't competing only with funds.

**6,109 - US professional profiles with "Angel Investor" in the title, in Refolk's index**

In a 25-profile sample, 9 were in the New York area and 8 in the SF Bay Area, the same two metros where demo day demand concentrates.

## How many meetings does one closed check take?

The best benchmark is about 28 first meetings per closed deal, from a survey of 885 institutional VCs at 681 firms (Gompers, Gornall, Kaplan, and Strebulaev, "How Do Venture Capitalists Make Decisions?"). Size your Tier 1 slate from the number of checks you intend to write at that ratio, then cap it at the meetings you can really hold.

#### What one closed deal costs upstream

| Stage | Figure | Note |
| --- | --- | --- |
| Considered | 101 | early-stage firms report 119 |
| Met management | 28 | your Tier 1 meetings |
| Partner meeting | 10 |  |
| Diligence | about 5 | 4.8 in the survey |
| Term sheets | 1.7 | per 1 closed deal |

*For each deal closed, surveyed firms considered about 101 companies and met about 28 management teams.*

The same ratios scaled to a 15-company slate:

| Stage | Per closed deal | Per 15-company slate |
|---|---|---|
| Met management | 28 | 15 |
| Partner review | 10 | about 5.4 |
| Diligence | 4.8 | about 2.6 |
| Term sheets | 1.7 | about 0.9 |
| Closed | 1 | about 0.54 |

The per-deal column comes from the survey. I derived the slate column by multiplying each figure by 15/28, or 0.536.

**28 - First meetings per closed deal in the Gompers et al. survey**

A 15-company demo day slate therefore carries about half an expected deal, not one.

### The slate formula

- **Slate needed** = target checks from this batch × 28
- **Meeting capacity** = first meetings the partners can genuinely hold in the two weeks after demo day
- **Tier 1 size** = the smaller of the two
- **Expected checks** = Tier 1 size ÷ 28

At this ratio, a 10-company slate gives about 0.36 expected checks, 15 gives about 0.54, and 20 gives about 0.71. If capacity is below the slate you need, you have three honest choices. You can lower the target, add a partner to the meeting load, or accept a lower probability and write that down.

### Limits of the ratio

The survey ratio covers a firm's whole deal flow, not a pre-filtered accelerator batch. Early-stage firms in the same data considered 119 companies per closed deal, against 101 across all firms. A slate that has already passed six filters might convert better than 1 in 28. That is not established, so log your own conversions each batch and replace the 28 once you have a number of your own.

There is a speed problem too. In the survey the average deal took about 83 days from first meeting to close, with about 118 hours of diligence. Demo day rounds mostly close within 4 to 8 weeks. My inference is that a fund running its normal pace will lose most of its Tier 1 slate. The way out is to do the evidence and claim work before demo day so the meetings start further down the funnel.

## The triage procedure, stage by stage

The procedure runs six filters in rising cost order: capacity, thesis, conflict, founder evidence and momentum, claim plausibility, then booking. Each filter is cheaper per company than the one after it, so the expensive judgement is only spent on survivors.

#### The triage calendar

1. **Weeks 4 to 3 before** - Pull the cohort, write the capacity cap
2. **Weeks 3 to 2 before** - Thesis filter and conflict check
3. **Weeks 2 to 1 before** - Founder evidence, momentum, claim reruns
4. **Demo night** - Request every Tier 1 meeting
5. **Weeks 1 to 2 after** - Hold Tier 1 meetings, promote from Tier 2
6. **Weeks 4, 8, 12 after** - Re-rank the watchlist

*Every cut that needs no conversation happens before demo day, and the watchlist runs for 12 weeks after it.*

#### Cohort list to sized slate

1. **Pull the cohort into one sheet** - Three to four weeks before demo day, export the directory, launch posts, and announcements into one row per company with URL, one-liner, and founders. Takes 1 to 2 hours.
2. **Write the capacity cap** - The partner writes down the maximum check, maximum valuation band, and maximum number of companies, and sets the Tier 1 cap at target checks times 28, limited by real meeting hours. Takes 30 minutes.
3. **Run the thesis filter** - Tag sector and stage from each one-liner at about 30 seconds per company. Expect roughly a quarter of the cohort to survive.
4. **Clear portfolio conflicts** - The partner checks every survivor against the CRM at about 1 minute each, then marks and drops conflicts.
5. **Score founder evidence and public momentum** - Compare repo activity with the company's own baseline and verify prior roles, about 2 minutes per company. Write one line saying what would make it interesting and what would kill it.
6. **Rerun every headline claim** - Classify each metric as collected, contracted, or run-rate, find the month the run-rate rests on, recompute growth over 12 weeks, and set a confidence flag.
7. **Request Tier 1 meetings that night** - On demo night, request every Tier 1 meeting through the portal's request-to-meet button and by direct ask, until Tier 1 is booked.
8. **File Tier 2 and date the watchlist** - Move the next band to Tier 2, and give every other thesis-fit company re-rank dates at 4, 8, and 12 weeks with an empty trigger log.

### What each stage costs

| Stage | Owner | Time | Done means |
|---|---|---|---|
| Pull the list | Analyst or platform lead | 1 to 2 hours | Every company has a row |
| Capacity cap | Partner | 30 minutes | Written slate cap |
| Thesis filter | Analyst | About 30 sec per company | Sector and stage tagged |
| Conflict check | Partner plus CRM | About 1 min per survivor | Conflicts dropped |
| Founder and momentum | Analyst | About 2 min per company | One-line case for and against |

The 2-minute figure for founder evidence is my working estimate, not a published benchmark. No one has published timings for checking domain age, hiring posts, or prior employers. Time your own first batch.

### Worked arithmetic for a 200-company cohort

- Thesis filter: 200 × 30 seconds is about 100 minutes.
- About a quarter survive, so about 50 companies go to the conflict check: about 50 minutes.
- Founder evidence on about 50: about 100 minutes.

That is under a working day of analyst time, before the partner's claim reruns. The 1-in-4 survival rate is inferred from the survey's ratio of meetings to opportunities considered (28 of 101). It is not a measured cohort figure. The cuts after the thesis filter have no published survival rate at all. Your capacity cap sets them, not the cohort.

Fill one row per company and leave nothing blank. A blank cell is where a weak company slips through.

**Triage sheet columns (one row per company)**

```
Company | URL | One-liner | Founders | Sector tag | Stage tag | Thesis fit (Y/N) | Conflict (none / adjacent / direct) | Prior roles verified (Y/N + source) | Repo activity vs own baseline | Headline claim as stated | Claim type (collected / contracted / run-rate) | Run-rate month | Base for growth figure | Confidence (high / medium / low) | Interesting if | Killed by | Tier (1 / 2 / W) | Re-rank dates (4w / 8w / 12w) | Trigger log
```

*Paste as a header row; add your own fund-specific columns to the right, never in the middle.*

## Which signals can you check in two minutes, and how do they lie?

The two-minute signals are the company's public code activity, the directory one-liner, and the founders' prior roles. Each one proves something narrow and fails in a known way, so record what it proves alongside the reading.

| Signal | What it proves | How it lies |
|---|---|---|
| Repo commits and contributors | People are shipping code in public | Forks, bots, mirrored repos |
| Directory one-liner | The category the founders chose | Positioning, not product |
| Prior employer and school | Claimed background | Fabrication happens |
| Headline ARR | Some revenue exists | Contracted or spike-month figures |
| Weekly growth chart | Direction of travel | Tiny base, annualized |
| Investor demand | Price and access | Little link to outcomes |

### Repo activity

For technical batches, the companies' public GitHub orgs are usually live before demo day, so you can measure commit activity before you see the pitch. Compare each company against its own baseline: contributors and commit cadence over the last 30 days against the 30 before that. Don't compare it against other companies. A three-person team shipping steadily looks small next to an org padded with forks and bot commits, and a ranking across companies will reward the padding.

Pulling that baseline for 50 companies by hand is the slow part of this stage. [Refolk](/) takes the question in plain English and returns the founders and their companies' public code activity together, so the momentum column fills before you open a repo.

I ran this search: `Founders in the current YC batch building developer tools whose company GitHub org has had commits from 3+ contributors in the last 30 days` - [see the full result list](https://www.refolk.ai/s/2epv0c03yt).

*Returns founder profiles tied to company GitHub orgs with recent multi-contributor activity, ready to paste into the momentum column of the triage sheet.*

### Prior roles

Verify the prior employer and school on your own before the meeting. Don't rely on the deck or the directory for this. Credential fabrication is documented: Medobed was expelled from a YC batch because its founder had fabricated his educational history and inflated growth metrics. The same Refolk search that finds a founder also shows their public employment record, which turns the check into a comparison instead of a hunt.

### Writing the one-liner

Every survivor of this stage gets one line in two halves: "interesting if..." and "killed by...". Fill in both halves. If you can't name what would kill a company, you haven't looked closely enough to put it in Tier 1.

## How do you rerun a demo-day metric before the meeting?

Rerun every headline number in three ways. Split ARR into collected, contracted, and run-rate. Find the month the run-rate is built on. Recompute weekly growth over the batch window instead of a year. This filter comes after the cheap ones, and you can't skip it, because the bar founders are measured against keeps rising.

Tom Blomfield has described the shift. About two years earlier, a good YC demo day target was $150k of annualized revenue. The lowest target he now saw was $800k, and most companies were aiming for $1 to 2 million. A rising bar raises the incentive to report the most flattering number available.

### Three questions for every revenue claim

1. **Collected or contracted?** Contracted ARR (CARR, revenue signed but not yet live or billed) is often reported as plain ARR. One VC described companies whose contracted figure ran 70% higher than their actual recurring revenue.
2. **Which month?** Run-rate means one month's revenue multiplied by 12. If that month was a one-off spike, the annual figure is fiction. A claim like run-rate revenue of more than $1 million within eight weeks of launch isn't wrong on its face, but it needs the month-by-month series behind it.
3. **Recurring?** One-time setup fees and pilots aren't recurring revenue.

> **Watch out:** Contracted is not recurring
>
> If a founder can't split the headline number into collected, live, recurring revenue and everything else, flag the claim low confidence and say so in the meeting. Gaps of up to 70% are documented.

### Discounting a growth chart

Paul Graham set the reference point: "A good growth rate during YC is 5-7% a week. If you can hit 10% a week you're doing exceptionally well." The trouble starts when that weekly rate is annualized. One TechCrunch column took the growth figures claimed at a demo day, projected them twelve months out, and got revenues ranging from $1 million to $159 billion.

| Weekly growth | 12-week multiple | Annual, as often reported |
|---|---|---|
| 1% | 1.13x | 1.7x |
| 5% | 1.80x | 12.6x |
| 7% | 2.25x | not computed |
| 10% | 3.14x | 142x |

I derived the 12-week column as (1 + r) raised to the 12th power. The annual figures come from published notes on Graham's growth essay. At 5% a week the batch actually delivered 1.8x, and the slide says 12.6x. No one has published a formal discount factor. My rule is to evaluate the 12-week multiple on the absolute base and ignore the annualized number.

Send these questions before the meeting, so the answers arrive before you sit down.

**Pre-meeting claim check (send after the meeting is booked)**

```
Hi [first name], looking forward to [day]. So I can come prepared rather than spend our time on definitions, three quick questions on the number in your deck:

1. Of the [figure] you presented, how much is collected, live, recurring revenue, and how much is contracted but not yet live?
2. Which month is the run-rate built from, and what were the two months before it?
3. What was the absolute starting base for the weekly growth rate, and over how many weeks?

A rough answer is fine. I would rather hear the real number now than discover it in diligence.
```

*Swap in the founder's own figure and metric name; keep all three questions even if one seems already answered in the deck.*

## How demo day batch triage goes wrong

Most triage failures come from one of eight mistakes. Each has a cheap check. Use this section as a pre-flight list before demo night.

1. **Starting late.** The triager assumes the list arrives 24 to 48 hours out, while the top companies have already priced. *Check:* date the work from the cohort announcement, not from demo day.
2. **Taking ARR at face value.** Contracted ARR or a one-month spike gets annualized. *Check:* ask "collected, live, recurring?" and allow for gaps of up to 70%.
3. **Believing the growth chart.** Short windows compounded off a tiny base produce huge annual numbers. *Check:* ask for the absolute base and redo the 12-week multiple.
4. **Treating demand as a signal.** Oversubscription gets read as quality. Seibel says there is little correlation between the companies most in demand on demo day and the ones that become extremely successful. *Check:* score on evidence before looking at who else is in.
5. **Repo false positives.** Forks, bots, and mirrored repos inflate activity. *Check:* compare commits and contributors with the company's own baseline, never with other companies.
6. **Fabricated credentials.** Medobed's founder fabricated his education and inflated his metrics. *Check:* verify prior employer and school on your own before the meeting.
7. **Overbooking against a false expectation.** Booking 15 meetings and expecting a deal, when the funnel implies about 0.5. *Check:* size the slate from target checks at 28 meetings each.
8. **Unearned re-rank dates.** Watchlist dates get set, but no one tracks whether the triggers predict anything. *Check:* log every trigger and its outcome to build your own base rate.

> **Rule:** Evidence first, demand second
>
> Fill the evidence and claim columns before you open any signal of who else wants the company. Demand belongs in the price column, not the quality column.

The matrix below is how I sort the survivors once both columns are filled.

#### Evidence against demand

Horizontal axis runs from Low investor demand to High investor demand. Vertical axis runs from Thin evidence to Strong evidence.

| Quadrant | What it means |
| --- | --- |
| Quiet and proven | Tier 1, request early; your price discipline holds here |
| Proven and crowded | Tier 1 only if inside your valuation band; request that night |
| Quiet and thin | Watchlist with a dated re-rank |
| Hyped and thin | Do not chase; re-check at 4 weeks |

*Quiet companies with strong evidence are where triage pays off, and loud companies with thin evidence are where it saves you time.*

Third-party batch scores have a place, but not in the evidence column. One fund reported that 35% of a single YC batch scored in the top 20% of all YC companies it had ever evaluated. If anything, that means quality is spread widely through a batch, which is one more reason not to let a single ranking stand in for your own filters.

## How to follow up demo day companies that did not raise

Put every company that fits your thesis but missed your slate on a watchlist, and re-rank it at 4, 8, and 12 weeks after demo day. The 4 and 8 week checks sit inside the reported close window, where most rounds close in the four to eight weeks after the event. The 12 week check catches the companies that raise outside it.

> The long tail of a batch is a value pool, not the leftovers from the people who booked faster.

The case for the long tail is historical, not statistical. Airbnb struggled to raise even after demo day. DoorDash closed its first round in the weeks after its demo day. Insiders get early looks, and demand doesn't track outcomes, so how fast a round fills doesn't rank the companies in it.

### What to check at each re-rank

No published study ranks re-rank triggers or times them, so the list below is a starting set, not a proven one. Each trigger comes with what it suggests and how it misleads.

- **Engineering job posts with no priced round announced.** Suggests the company is spending, and may still be raising. Misleads when the posts are evergreen or the round closed quietly.
- **Code activity rising against the company's own baseline.** Suggests the team is shipping. Misleads with forks, bots, and mirrors, as above.
- **A new metric the founder will split into collected and contracted.** Suggests the earlier gap has closed. Misleads if the run-rate month is a spike again.
- **A round announcement.** Tells you the round is gone or nearly gone. Use it to close the row, not to chase.

Do not use press coverage or visible investor interest as a trigger. Demand is what the triage was built to ignore.

Re-running the first trigger by hand across a batch every four weeks is exactly the repetitive search that slips. In Refolk you can save a plain-English ask such as "Demo day companies from the last two YC batches that posted engineering job openings in the past 60 days but have announced no priced round" and rerun it on each re-rank date.

### Building your own base rate

Each time a trigger fires, record three things: the date, what you did (ignored, reached out, met), and what happened by the next re-rank. After two or three batches you'll know which triggers were worth acting on for your thesis. That is better evidence than anything currently published.

## What to verify before the slate goes out, and how to keep the method current

The slate is ready when every Tier 1 company has passed all six filters with no blank cells, the slate size matches the written cap, and every watchlist row carries three dates. Run the checklist below on the afternoon of demo day.

#### Before demo night

- [ ] The triage clock was dated from the cohort announcement or first visible list, not from demo day.
- [ ] The partner's maximum check, valuation band, and company count are written down.
- [ ] Tier 1 size equals the smaller of target checks times 28 and real meeting capacity, and the expected-checks figure is recorded.
- [ ] Every Tier 1 company has a conflict status of "none" checked against the CRM.
- [ ] Every Tier 1 founder's prior employer and school were verified from a source other than the deck.
- [ ] Repo activity for every technical company was compared against its own prior 30 days, with forks and bot commits excluded.
- [ ] Every headline revenue claim is split into collected, contracted, and run-rate, with the run-rate month named.
- [ ] Every weekly growth claim has an absolute base and a recomputed 12-week multiple.
- [ ] Every Tier 1 row has a filled "interesting if" and "killed by" line.
- [ ] The pre-meeting claim check message is queued for every Tier 1 company.
- [ ] Every watchlist row has re-rank dates at 4, 8, and 12 weeks and an empty trigger log.

### Keeping the numbers honest across batches

YC runs four cohorts a year, which gives you four chances a year to replace borrowed ratios with your own. After each batch, record these:

- **Thesis survival rate.** Replace the inferred 1 in 4 with your actual share.
- **Meetings per check.** Replace the survey's 28 with your real number once you have closed a few.
- **Minutes per company** at the founder evidence stage, to retire my 2-minute estimate.
- **Trigger outcomes** from the watchlist log.
- **When you first saw the list**, compared with demo day, since that timing is disputed and may change.

Re-check the mechanics each cycle rather than trusting this page: when the directory preview opens, how the request-to-meet flow works in the investor portal, and how many companies are presenting. These are the parts of the job most likely to change between batches.

## Frequently asked questions

### How far ahead of demo day should I start triaging a YC batch?

Start three to four weeks before demo day, counted from when the cohort becomes visible and not from the event date. Sources disagree on when approved investors can see the list. One says 24 to 48 hours ahead, another says weeks, and unofficial lists circulate earlier still. The strongest companies finalize term sheets in the days and weeks before the event, so starting late means triaging a list whose top rows are already priced.

### How many companies should be on my demo day shortlist?

Size your Tier 1 shortlist at about 28 first meetings for every check you intend to write, then cap it at the meetings you can really hold in the two weeks after demo day. The 28 comes from a survey of 885 VCs, where each closed deal took about 28 management meetings. At that rate a 15-company slate carries about 0.54 expected deals. Treat that as a starting ratio and replace it with your own.

### What should I do about demo day companies that did not raise?

Put every thesis-fit company that did not make your slate on a dated watchlist, and re-rank it at 4, 8, and 12 weeks after demo day. Most rounds are reported to close within one to two months, so the 4 and 8 week checks fall inside the close window and the 12 week check catches stragglers. Log each trigger you act on and how it turned out, because no published study ranks re-rank triggers.

### Does this triage method work for Techstars or smaller accelerators?

The full funnel only pays off at YC scale, where past batches have presented 149, 189, and nearly 400 companies. Techstars classes run about 5 to 11 companies, such as London with 11 and Boulder with 9, so you can afford to diligence the whole class. With a small class, skip the thesis and conflict cuts and go straight to founder evidence and claim plausibility.

### How do I check whether a demo day ARR number is real?

Ask whether it is collected, live, recurring revenue, and which month the run-rate is built from. The documented failures are contracted ARR reported as plain ARR, and run-rate built on a one-off spike month. One VC described contracted figures running 70% higher than actual recurring revenue. Send the questions before the meeting so the answer is ready when you sit down, and set a confidence flag on every claim.

### Should I prioritize the demo day companies with the most investor demand?

No. Score on evidence before you look at who else is interested. Michael Seibel said there is little correlation between the companies most in demand on demo day and the ones that become extremely successful. Airbnb struggled to raise after its demo day, and DoorDash closed its first round in the weeks after. Demand tells you about price and access, not quality.

---

*From the Refolk guide library. I revise these guides rather than replacing them, so the current version is always at https://www.refolk.ai/guides/demo-day-batch-triage-playbook*