The Cold-Start Problem: Forecasting a Project That Hasn't Shipped Anything Yet

Throughput-based forecasting has one honest weakness: it needs throughput. A brand-new project has completed zero tickets, which means there's no history to sample from — and ordinarily, no history means no forecast. If every weekly sample in your window is zero, the simulation has nothing to draw from and simply can't run.

That's a frustrating gap for exactly the moment a forecast is most wanted: kickoff, when a stakeholder is already asking "when will this be done?"

Borrowing evidence instead of guessing

PtahCast's answer isn't to make something up — it's to let you borrow real evidence from elsewhere. When you create a project, there's an optional starting weekly throughput field. The number you put there isn't a wild guess; it's meant to be informed by how the specific people joining this project have performed on other work. PtahCast's Per-person throughput report exists precisely to supply that number — each team member's actual historical completion rate, sitting right there for you to reference before you type anything into the field.

Enter a number, and PtahCast generates a 13-week provisional array using it as an average — with realistic week-to-week variation built in, not a flat, robotic repeat of the same figure thirteen times over. That gives the Monte Carlo engine something honest to sample from on day one, instead of nothing.

The part that matters: it disappears on its own

Here's the detail that makes this more than a placeholder trick. From the moment the estimate is set, each of the 13 weeks resolves to real data automatically as it elapses. Week 1, all 13 weeks are still estimate. By week 6, the six weeks that have actually happened are using your team's genuine completed-ticket counts, and only the remaining seven are still provisional. Once all 13 weeks have passed, the estimate has fully aged out — the project behaves exactly like any other established project, and the number you entered at kickoff has no further effect.

Nobody has to remember to clean this up. There's no toggle to flip, no manual reconciliation step. Real data simply replaces the guess, one week at a time, as it becomes available.

A detail worth knowing about the windows

Because weeks are fixed calendar weeks (Monday to Sunday) rather than a rolling window, a 30-day forecast and a 90-day forecast run on the same cold-start project, at the same moment, can show a different mix of real versus provisional weeks — one might still be showing a provisional banner while the other has already gone fully real. That's not a bug; the two windows are genuinely looking at a different slice of calendar time.

When any week inside your selected window is still provisional, PtahCast shows a gold banner above the forecast, so nobody mistakes a directional early estimate for a committed date. Treat it accordingly — useful for planning conversations, not for the number you put in a contract.

The lesson

A cold start doesn't have to mean flying blind. It means borrowing the best evidence you have — your own people's track record — and letting the system quietly swap that evidence for reality as reality arrives.