Your stakeholder asks for a launch date before anyone has sized the work. You paste tasks into a spreadsheet, drag bars until the chart looks plausible, and two weeks later a vendor delay forces you to rebuild the whole grid. The failure is not the tool—it is treating the finish date as the input instead of the output.
Anchor the start, then stack durations
Pick one immovable start: contract signature, factory release, or the day permits clear. Every downstream date is start plus elapsed time. If design needs eighteen working days after kickoff, write that as a duration attached to the start, not as a free-floating cell you nudge until it "looks right." When a task slips, only its successors move; the anchor stays put.
On a small website refresh, the team locked content freeze to the Tuesday after legal review (nine calendar days from kickoff). Engineering duration stayed forty hours of effort, but the calendar span became twelve days because of a holiday and a shared QA window. Recording both numbers—effort hours and elapsed days—prevented the PM from promising a Friday ship when the elapsed chain pointed to Monday.
Separate working time from waiting time
Waiting is not work. Approval queues, print shops, and courier legs belong in a "lag" row with calendar days, while build tasks use the hours people actually touch the work. Mixing them is how schedules absorb invisible buffer until the first delay exposes none left.
Practical rule: if nobody on your team can pull the task forward on a given day, it is lag. If they can, it is work. Label rows that way in your sheet so reviewers see where risk sits.
Publish ranges, not false precision
Executives want one date; teams need a window. Share a committed internal date (median case) and a stakeholder date (median plus one lag you already named). When procurement adds five days, you adjust the lag row— not every task in the plan.
Before you email the launch date, sanity-check the chain with a Date Calculator: enter the anchor start and add each elapsed segment. If the total disagrees with your Gantt by more than a day, find the row where calendar days and working days were swapped.
When the plan must move
Slips happen. Update durations first (realistic remaining work), then lags (vendor truth), then scope last. Cutting quality to protect a date you never derived from durations is how teams repeat the same fire drill next quarter.
For cross-team handoffs, also note time-zone cutoffs: "end of day Friday" in Denver is already Saturday morning in London. State the zone on the handoff line.
Stakeholder questions worth rehearsing
Before the steering meeting, list the three dates people will actually ask about: earliest demo, contractual milestone, and revenue-recognition cutoff. Map each to a chain segment so you can answer without reopening the whole plan. If demo moves, show which lag row shrank—feature freeze or review slack—not a vague "we are slipping."
Students running capstone projects benefit from the same discipline: pick the day the prototype must sit on the table, subtract build and test durations from that anchor, and see whether the implied start is before your other courses' crunch week. The math is identical to a product launch; only the stakes feel smaller until they are not.
Recording assumptions next to the chart
Assumptions rot quietly. Next to the timeline, keep a living bullet list: "vendor quotes 10 calendar days," "legal review assumes one round," "QA shares a pool with mobile." When an assumption breaks, edit the list before you edit the bars. Reviewers trust plans that show what broke last time.
If you export a PDF Gantt, paste the assumption list on page two. Executives rarely zoom into row 47; they do read a short honest paragraph about risk.
Frequently Asked Questions
Should I plan in business days or calendar days?
Use business days for human tasks on your own team; use calendar days for anything involving shipping, government, or external calendars you do not control. Many slips come from using business math on a calendar-only process.
What if the start date is not fixed yet?
Run two scenarios with the same durations: earliest feasible start and latest acceptable start. The overlap window is what you can promise externally until the anchor firms up.
How much buffer belongs in the plan?
Put buffer in explicit lag rows with names ("client review slack") instead of hiding it inside every task. Named buffer can be traded away deliberately; hidden buffer just rots.
Do I need a Gantt chart?
No. A ordered list with start anchor, durations, and lags is enough for most small projects. Charts help communication; they do not replace the arithmetic.
How do I check a date chain quickly?
Add durations in order from a fixed start using the same day type (calendar or business) throughout. A one-minute check catches transposed months and weekend mistakes before they become commitments.
Related planning: counting days until a milestone with Days From Now Calculator helps when someone asks "how many days until beta" while you are still stacking durations.