The cheapest fix is the one made before an investor sees the deck. Run these five rounds in order; each catches a different class of failure, and each can be completed in an afternoon.
Round 1 - The 90-second read
- Give the deck to someone unfamiliar with the company for 90 seconds, then take it away.
- Ask them: what does the company do, who pays, how is it going, and what are they asking for?
- Any question they cannot answer identifies a slide that is not doing its job.
- Repeat with three people. Consistent gaps are structural, not personal.
Round 2 - The hostile read
- Ask a sceptical operator or investor to argue against the company for twenty minutes using only the deck.
- Write down every objection verbatim; do not defend during the session.
- Categorise: objections answerable with existing data, objections requiring new data, and objections that are genuine risks.
- Address the first two in the deck. Address the third openly on a risk slide or in the meeting.
Round 3 - The arithmetic audit
- Trace every number in the deck to a cell in the model or metrics workbook. Numbers with no source get removed.
- Check internal consistency: does the customer count on the traction slide reconcile with the revenue and the stated ACV?
- Check the ask: does the amount actually fund the milestone at the stated burn, including hiring lead times?
- Check the market model: multiply it out yourself and see whether the implied share is plausible.
Round 4 - The forward test
- Assume the deck is forwarded to someone who will never speak to you. Does it stand alone?
- Check the file: sensible filename with company and date, PDF, under 10 MB, fonts embedded, no broken links.
- Confirm nothing confidential is in it that you would not want circulated - because it will be.
- Confirm the contact details and the ask are on the last slide as well as the first.
Round 5 - The reference test
- For each claim about a customer, partner or advisor, confirm that person would repeat it if called.
- For each metric, confirm you could produce the raw data within 24 hours of a request.
- For each risk you chose not to disclose, decide now what you will say when it is discovered - and consider disclosing it instead.
After the test
Rebuild the deck once after all five rounds rather than patching between them; patched decks accumulate inconsistencies. Then freeze the version, log the date, and track which version each investor received. When the deck changes mid-process, tell anyone still reading the old one - it is a small signal of operating discipline and it costs nothing.