Price dependency risk and lead time
External work is the estimate's most common hidden cost: it does not appear in your task list but it controls your start. Make every wait explicit, give it a lead time and an owner, and decide the fallback before the lead time runs out.
Procedure
For each external dependency record four things: what you need, who controls it, the lead time, and the trigger date (needed-by minus lead time).
notes/deps.md - need: API key for vendor Z | owner: procurement | lead: 10 business days | needed: 2026-03-01 | trigger: 2026-02-13Sanity-check the lead time against history, not the vendor's promise. If last time took 15 days, use 15.
Insert the dependency as a real node on the critical path in
notes/deps.md, with its lead time as its duration. It is not free time.Set an alarm at the trigger date. If the dependency has not moved by then, execute the fallback:
# in a scheduler or calendar remindctl add --title "chase API key Z" --due 2026-02-13Name the fallback up front: mock it, use a competitor's service, descope the feature. "We'll figure it out" is not a fallback.
Add the dependency's uncertainty to the range — external waits have wide tails and rarely come in early.
Pitfalls
- Assuming the promised lead time is the real one; vendors quote best-case.
- Starting the clock when you remember to ask, not when you first knew you needed it.
- Treating the wait as free because no one is working, so it never lands on the critical path.
- No fallback decided, so the slip triggers an improvised scramble under deadline pressure.
- Forgetting approval-type dependencies (security review, legal, a signing ceremony) which have long, lumpy lead times.
Verification
grep -n 'trigger:' notes/deps.md
# passes when every dependency has a trigger date in the past-of-its-needed date and an owner
Report each dependency with its owner, lead time, trigger date, and the fallback you will execute if it slips.