Forecast (Monte Carlo)

Answers "when will this be done?" without asking anyone to estimate.

This is a Monte Carlo simulation. It doesn't guess about individual items — it replays your team's own delivery history thousands of times to see how the future might unfold, and answers two questions: "when will N items finish?" and "how many will be done by a date?"

Monte Carlo forecast answering when 30 items will finish, at four confidence levels — **When will N items finish?** — 30 items, simulated 10,000 times from historical throughput. Each card is a confidence level: 50% by Oct 26, but 85% by Nov 8. The histogram shows the full distribution, with the low-confidence tail in red.

How it works

1 · Historical sample

For every calendar day in the selected window, FlowPaths counts how many items completed (reached a Done status) — including zero-completion days. That list is your real daily-throughput distribution.

2 · Resample

Each simulated future day is a random pick from one of your team's actual past days. A "future day" in the simulation looks like a randomly chosen day from your history — not a guess or a trend line.

3 · 10,000 trials

When will N finish? Keep drawing random days, summing completed items, until you reach N — record how many days it took. Repeat 10,000×, giving a distribution of finish times.
How many by date X? Sum random draws over the number of days until X. Repeat 10,000×, giving a distribution of item counts.

4 · Confidence = percentile

Results are sorted, then read off by percentile.

For "when", the 85% row is the day by which 85% of the 10,000 simulations had finished — so a higher confidence number means a later date.
For "how many", FlowPaths reports a lower bound ("at least this many"), so 85% confidence is actually the 15th-percentile count — a higher confidence number here means fewer items, since it's being more cautious about the promise.

Use the 85% or 95% rows for commitments — the same percentiles used on the Scatterplot and Aging WIP.

Monte Carlo forecast answering how many items will be done by a target date — **How many done by a date?** — the same simulation asked the other way round. By 31 October you can be 50% confident of 33 items, but 95% confident only of 22. The gap between those two numbers is the honest measure of your predictability.

Assumptions & limits

The future resembles the sampled past — no seasonality, trends, or one-off events are modelled.
It counts items by throughput, so every item is treated as equal size (no story points).
Scope changes and cross-team dependencies aren't modelled.
More history gives a more reliable forecast — very short windows give noisy results.

Lower-bound flag: if a meaningful share of simulations can't reach the target within the forecast's cap (very low throughput, or a target that's too high), FlowPaths flags the result as a lower bound rather than presenting a false level of confidence.

Forecasting a single stage

With every Doing status ticked, the forecast uses your whole throughput and answers a delivery question. Untick some, and the sample counts only completed items that passed through the ticked statuses, with the default target being the open items currently in them.

Read that difference carefully. A narrowed forecast answers “when does work that goes through these stages get done?” — and if many items skip those stages, the dates come out later than your real delivery rate would give. Tick every Doing status for a delivery forecast. A note above the chart says which mode you are in. See the Statuses dropdown.

Next: Configuration & Filters →