Seed fundraise: the data room and the numbers investors actually check
What goes in the data room, which metrics get scrutinised line by line, and the questions that kill rounds late.
A seed round is not won in the pitch. It is won in the second meeting, when a partner opens your model and starts checking whether the story and the spreadsheet agree.
I have raised across equity and venture debt, in good markets and bad ones. The pattern is consistent: founders lose rounds not because their business is weak but because their evidence is disorganised.
This playbook is what to prepare, in what order, and what will be checked.
Before you open the data room
Do not build the data room while you are pitching. Build it before the first meeting, so that momentum never stalls waiting on you.
- Cap table is clean, with every SAFE, note and advisor grant documented
- All founders have signed IP assignment agreements
- Bank statements reconcile to your revenue reporting for the last 12 months
- You can produce a cohort retention view without building it from scratch
- Customer contracts are signed and stored in one place
How to structure the data room
Six folders. Numbered. No duplicates, no 'final_v3'. An investor should find anything in under thirty seconds.
| Folder | Contents |
|---|---|
| 01 Company | Deck, one-pager, incorporation documents, cap table, prior round documents |
| 02 Financials | Historical P&L, monthly management accounts, bank statements, the model |
| 03 Metrics | KPI dashboard export, cohort tables, funnel data, unit economics workbook |
| 04 Customers | Signed contracts, pipeline export, churn log with reasons, references |
| 05 Product and tech | Architecture overview, roadmap, security posture, key dependencies |
| 06 Team and legal | Org chart, hiring plan, employment agreements, IP assignments, any litigation |
The numbers that get checked line by line
Investors do not read every number. They read these, and they cross-check them against each other. If two of them disagree, everything else becomes suspect.
- ◆Monthly revenue for the last 18 months, with new, expansion, contraction and churn split out
- ◆Logo and revenue retention by cohort, not blended averages
- ◆Gross margin, calculated honestly, with infrastructure and support costs included
- ◆Customer acquisition cost by channel, and payback period in months
- ◆Burn, runway, and the exact date the money runs out at current burn
- ◆Pipeline conversion at each stage, with the sample size shown
- ◆Headcount plan against the milestones the round is meant to buy
The fastest way to lose credibility is a blended metric that hides a bad cohort. Show the bad cohort and explain what you changed.
The model: bottom-up, always
Top-down models that start from market size and apply a percentage signal that you do not understand your own engine. Build up from the inputs you control.
- 01Start with activityOutreach volume, traffic, or partner referrals. Things a person or a channel actually produces.
- 02Apply your real conversion ratesFrom your own data, with the sample size noted next to each rate.
- 03Layer in retentionUse your worst plausible cohort, not your best one.
- 04Derive revenueRevenue is the output of the model, never an input you set and work backwards from.
- 05Cost the planEvery hire dated, every line traceable to a milestone.
Diligence questions that kill rounds late
These come in weeks six to ten, once a partner is interested enough to look properly. Prepare the honest answer in advance rather than improvising it.
- ◆Why did these three customers leave, and what specifically changed afterwards
- ◆What percentage of revenue comes from your largest customer
- ◆Which revenue is contracted, which is repeat but uncontracted, and which is one-off
- ◆Show me the month the numbers dipped and explain it
- ◆What happens to this business if your main acquisition channel stops working
- ◆What did you personally get wrong in the last twelve months
- ◆Who leaves if the round does not close
The right answer to almost all of these is specific, unflattering and followed by what you did about it. Investors are pricing your judgement, not your polish.
Running the process
- 01Build the list first60 to 100 funds, ranked by stage fit, sector fit and check size. Warm path noted for each.
- 02Batch the outreachRun first meetings in a compressed window so that interest arrives at the same time.
- 03Keep one trackerFund, partner, stage, last touch, next step, date. Reviewed daily during the raise.
- 04Never negotiate on price aloneBoard composition, pro rata, option pool and liquidation terms move outcomes more than valuation.
- 05Set your own deadlineMomentum is manufactured. Tell people when you intend to close and hold it.
- Deck finished and the narrative practised out loud ten times
- Data room live, permissioned and tested with a friendly outsider
- Target list built, with intro paths mapped
- Two mock pitches with people who will be harsh
- Runway calculated: you should start with at least six months left