Simulate, this page
You type the throughput you already know. Good for a quick conversation, for testing scenarios, and for when the history lives in your head or in a closed report.
Simulate · Monte Carlo
When will the project be done? Forecast with probabilities
Stop giving a single optimistic number. Simulate thousands of futures from your real throughput and commit at a confidence level.
When will a project be done? Instead of a single guess, Monte Carlo draws thousands of futures from your historical throughput and returns honest probabilities.
How to use it
Two numbers are enough to start. Neither of them needs a per-item estimate.
-
Enter your historical throughput
How many items your team completed in each of the past weeks. Counting on the board, in the tool report or in a spreadsheet all work. It does not have to be tidy, it has to be real.
-
Say how many items are left
The count of what still has to be delivered. Items, not points or hours: the simulation works with quantity completed per period.
-
Read the three percentiles
The simulation returns p50, p85 and p95. The p85 is the number to take into the conversation: it holds in 85 of every 100 scenarios drawn from your own history.
Why three dates instead of one
Estimating by analogy fails for a structural reason: it assumes the future behaves like the average case. Knowledge work rarely does. Delivery depends on how much the team completes per period, and that amount varies because of holidays, incidents, external dependencies and items that turned out bigger than they looked.
A Monte Carlo simulation does not try to remove that variation, it uses it as raw material. Each run draws weeks from your own history and assembles one possible future. After a few thousand runs there is a distribution of dates, and the percentiles come from it.
That is why the answer comes as three numbers. One of them is not “the right one” with the other two being errors: each carries a different chance of happening. When someone asks for a single date, the missing question is how much risk they are willing to take.
Which tool to use
Both produce a probabilistic forecast. What changes is where the data comes from.
Analyze your real flow
You upload the board CSV and the tool computes throughput, cycle time, percentiles and the forecast. Good when exportable history exists and you want the full picture.
Use this calculator on your site
Free for blogs, internal wikis and course material. Paste the code into your page: the calculator runs in the visitor’s browser, no sign-up.
Preview
Frequently asked questions
What is a Monte Carlo forecast?
It simulates thousands of possible scenarios using your historical throughput. Instead of a single number, you get a distribution of dates with probabilities.
What do I need to enter?
How many items to deliver and a few past throughput values, meaning items completed per week. The more history, the better.
Why use the 85th percentile?
The median is right only half the time. The 85th percentile is the common balance of optimism and safety to commit to a customer.
How many weeks of history do I need?
Six to eight weeks are enough to simulate. Below that the result swings a lot, because the sample has not captured the team normal variation. Above twelve weeks the gain is small, and history that is too old may describe a team that no longer exists.
Do I need to size the items?
No, and that is the advantage of the method. The simulation uses quantity completed per period, so items of different sizes are already accounted for by the variation in throughput itself.
What if the scope changes midway?
The forecast holds for the number of items you entered. New scope needs a new simulation. Re-running it every couple of weeks is more useful than defending the old date.
Does it work for teams not using Kanban?
It does. The arithmetic depends on how many items a team completes per period, not on the method used to organize them.
Can I use this to promise a date to a customer?
You can, on one condition: state the probability alongside it. Promising the p85 and explaining that it holds in 85% of scenarios is honest. Promising the p50 and saying nothing is not.
Have the board CSV at hand?
The analyzer reads the file inside your own browser and computes throughput, percentiles and the forecast without you typing anything.
