Ask a team when something will be done and you usually get a single date. "The 14th." It sounds confident. It's almost always wrong — engineering estimates miss their targets by 30–50%, and a single date hides the one thing a stakeholder actually needs: how likely is that?
Monte Carlo forecasting answers that question honestly. And if your team runs on monday.com, you already have everything you need to do it — the data is sitting in your board right now.
What Monte Carlo forecasting actually is
A Monte Carlo forecast doesn't guess a date. It looks at what your team has actually delivered — how many items you finished each week over your recent history — and then plays the future forward thousands of times, each run randomly sampling from that real throughput.
The result isn't a number. It's a distribution: a range of outcomes, each with a probability attached. Instead of "the 14th," you get "80% chance by the 18th, 95% chance by the 25th." That's a forecast you can actually commit to, because it tells you the odds.
It's the same technique used in financial modelling and engineering risk analysis. The only input it needs is your own history — no story points, no planning poker, no optimistic guessing.
Why monday.com is the perfect place to do it
Every Monte Carlo forecast is built from one thing: throughput — the count of items your team completes per unit of time. And your monday.com board has been quietly recording exactly that.
Every time an item moves to Done, monday.com logs the status change with a timestamp in the board's activity log. String those transitions together and you have a complete, honest record of your team's real delivery rate — week by week, with no extra tracking and no spreadsheets to maintain. That history is the fuel for the forecast.
This is why forecasting from a monday.com board is so much better than estimating in a meeting: the board can't talk itself into an optimistic number. It only knows what actually happened.
How to run a Monte Carlo forecast from your board
The mechanics are straightforward, whether you do it by hand or let software do it for you:
- Pull your throughput history. From your board's status-change history, count how many items reached Done in each of the last 8–12 weeks. That list of weekly counts is your sample.
- Decide what you're forecasting. There are four useful questions: How long will this set of items take? When will they be done (a calendar date)? How much can we finish in a fixed window? How likely are we to hit a specific deadline?
- Simulate the future, thousands of times. For each simulated run, randomly draw weekly throughput from your history and accumulate it until the work is done. Do this 1,000 times and you've explored 1,000 plausible futures.
- Read the spread. Sort those 1,000 outcomes and look at the percentiles. That's your forecast — a range with probabilities, not a single brittle date.
You can do this in a spreadsheet, and it's a great way to learn. It's also fiddly to maintain, and it goes stale the moment your board moves on. (If you'd like to feel how it works first, there's a free interactive Monte Carlo simulator you can play with — drag the sliders and watch the distribution form.)
Reading the forecast: P50, P85, P95
The output of a Monte Carlo forecast is usually expressed as percentiles. Three matter most:
| Percentile | What it means | When to use it |
|---|---|---|
| P50 | Half the simulated futures finished by here. The median — a coin flip. | Internal stretch target. Risky to promise. |
| P85 | 85% of futures finished by here. | The commitment most teams should actually make. |
| P95 | Nearly all futures finished by here. | When failure isn't an option. |
The honesty is the point. If a stakeholder wants the P50 date but the deadline only carries a 30% chance, you now know that before you commit — not at the retro afterwards.
The catch: a forecast is only as honest as the behaviour behind it
Here's the part most forecasting tools miss. A Monte Carlo forecast tells you the range. It does not tell you whether you'll land inside it — because the future isn't only a function of throughput. It's a function of behaviour.
Outcome = Capability × Behaviour. Your throughput history captures capability. But the moment focus fragments, work-in-progress balloons, or quiet dependencies start stacking up, your real delivery rate drifts away from the history the forecast was built on — and the range quietly stops being true.
So the forecast gives you the odds. The behavioural signals tell you whether the odds still hold. That's why probabilistic forecasting and behavioural intelligence belong in the same place: one sets the line, the other warns you when you're about to fall off it.
Doing it automatically in monday.com
Running this by hand once is educational. Running it every week, across every board, by hand, is not realistic. That's exactly what we built Delivery Intelligence by IMIRT to do.
It's the first Monte Carlo forecasting app on the monday.com marketplace. Add it to a board and it reads your real status-transition history, runs 1,000 simulations, and gives you back the P50/P85/P95 answer to all four questions — How long? When? How much? How likely? — with colour-coded confidence and coaching that tells you what to commit to. On top of the forecast, it watches the behavioural signals — drift, bottlenecks, stacking dependencies — that decide whether you'll actually hit the range. It's free for two boards.
Stop committing to single dates your board can already prove are unlikely. Let the history do the forecasting, and spend your meetings deciding what to do about the odds instead of arguing about them.