Agile Project Forecasting With Monte Carlo QSRA: A Smarter Way To Predict IT Delivery Timelines

While your Agile team keeps pushing through sprints to catch up on progress, senior management keeps asking: “Can we deliver on time?”

Looking at the waterfall chart, you are unsure whether the project will be successfully delivered. It would be desirable to have a scientific way to forecast the delivery date with higher confidence while maintaining Agile principles.

To find the solution, we can refer to a proven method from the infrastructure project control discipline: Quantitative Schedule Risk Analysis (QSRA) based on the Monte Carlo technique.

Instead of using task duration as the basis of the estimate, which was the traditional building block of the analysis, we could use a more commonly accepted concept in Agile projects: the average range of story points the team can complete in each sprint.

What is Monte Carlo Simulation and QSRA?

To understand what Monte Carlo simulation is, we can use a dice roll example. To predict the result of the roll, you might guess seven, as it has the highest probability of occurrence. However, it would be even better to roll the dice a thousand times and present the results as a distribution.

This is what Monte Carlo simulation does: it uses repetitive sampling to randomly predict outcomes. After a thousand iterations, it gives you a range of possible outcomes, as well as the corresponding statistics.

Applying the same concept to project scheduling, Quantitative Schedule Risk Analysis (QSRA) simulates thousands of different scenarios on the project milestone, instead of presenting only a single date.

This way, the result becomes a probabilistic forecast that shows the likely range in which a project could finish, and how confidently it could meet the management target.

In infrastructure and capital projects such as airports, rail systems, healthcare facilities, and university campuses, QSRA has already established a strong foundation within project controls. These projects typically take more than five to ten years and involve hundreds, if not thousands, of tasks with various critical milestones.

They carry high reputational and financial risks. In such projects, QSRA models activities based on their durations, manpower availability, and scheduling logic to estimate three-point durations and construct the Monte Carlo model.

The simulation results help senior management understand which project milestones are affected and where risks may impact the critical path. As a result, management can prioritize mitigation actions to most effectively and efficiently reduce project delays and cost escalation.

This modelling technique is well-suited to traditional projects with waterfall logic and linear workflows, but may not be appropriate for the dynamic nature of Agile projects.

Why Traditional QSRA Doesn’t Fit Agile

Most IT Agile work is known to have repetitive and incremental delivery approach. Instead of having a waterfall straight line milestone process, teams in Agile work in shorter sprint circles that would usually have two to four weeks duration. Within each of the sprint cycle, the project team would deliver the product by small and functional increments.

Projects under such environment would prioritize teamwork, agility, and user feedback. The scope and priority would readjust according to progress and review insight. The project planning thus become continuous and project success is measured by consistent roll out of value with completed user story and working software, rather than achieving certain milestone status.

Agile projects do not inherit a fixed duration. Instead, they are about:

• Story points and backlogs as units of effort.
• Velocity to measure speed of delivery.
• Iterative process with incremental scope and priorities.

Because of such, using the QSRA technique to Agile projects by the first glance may not be a good fit and it often result to misalignment. Providing task durations for story points does not match the Agile principle. Alternatively, the Monte Carlo analysis should build on how many story points each team can accomplish for every sprint, but not the task duration.

A Better Fit: Three-Point Estimates for Delivery Velocity

Instead of trying to provide the time required of each task, an approach often unfit for the Agile working situation, our proposed Monte Carlo method redirect the process to what usually the team member can more easily control: their delivery speed, often presented as story points accomplished in each sprint, a measurement would provide insight on both past performance and team dynamics.

It could be achieved with viewing the problem using another lens by asking “what is the historical range of story points our team could accomplish in each sprint?” instead of “what is the duration of each of the task?”

With this in mine, we can then structure our problem with a three-point estimation technique. This technique consists of defining the minimum velocity (i.e. the slowest expected delivery velocity with a challenging conditions), the most likely velocity (i.e. based on the rolling average of sprint data), and the maximum velocity (i.e. the best-case performance scenarios).

With the three-point estimates on hand, this found the basic of Monte Carlo simulations, which would, in its simulation, create thousands of scenarios under different possible circumstances. The output of this exercise would then become a probabilistic forecast that is under the Agile philosophy and achieve the robust scientific requirements for project planning and provide senior management confidence.

To summarize, this process would require a three-point estimation for the story points velocity:

• Minimum: Velocity under unfavourable conditions (e.g., 15 story points/sprint)
• Most Likely: Velocity based on rolling average data (e.g., 30 story points/sprint)
• Maximum: Velocity under best case conditions (e.g., 50 story points/sprint)

By running the QSRA analysis with these three-point estimate across the involved teams and their total backlogs, it would result a probabilistic forecast on how many sprints would be required to finish the project, which we would understand how much time is required.

This method of simulation aligned with the Agile philosophy and, at the same time, provide a proven statistical scientific method to provide the confidence of delivery timeframe.

Case Study: Delivering Certainty in a High-Volume Transaction System

To demonstrate this approach in real life, we can consider a case study of a high-stakes IT legacy architecture upgrade project.

In a recent engagement, the author has helped conducted risk analysis for a large-scale upgrade of an IT platform that would require processing huge amount of transaction within a short period of time, particularly within certain peak usage intervals.

The specification of the system is comparable to one of the largest trading platform in financial exchanges, there is no tolerance for down time and error as there would be huge financial and reputational consequence.

The delivery team works in Agile sprint cycles, yet traditional QSRA forecasting cannot fit well into this project framework and unable to provide valuable insight to executive management. To resolve this problem, the author introduced a Monte Carlo QSRA based on three-point estimates of story points per sprint. By doing so, we were able to:

• Model delivery uncertainty using historical velocity data and expert input.
• Simulate thousands of delivery scenarios across the evolving backlog.
• Provide alert to management when actual performance differed from planned.
• Identify a data-driven completion date, derived from simulation outputs, as a foundation of go-live planning.

This robust way of modelling not only improved forecast accuracy but also allow management to conduct proactive decision-making and thus building trust with executive managements. The result is to assure that this system would be delivered within the planned launch window.

Why This Matters for Agile IT Leaders

Now a day with everybody talking about AI, good IT systems are critical of business to strive. It could be a trading system, a banking app, or a global-scale transaction platform, the allowance for delay or failure is massive.

With Monte Carlo QSRA adapted to Agile through story point-based modelling, this method provides a way to:

• Quantify uncertainty in delivery time frame.
• Support scenario planning for scope, resources, and deadlines.
• Bridge the gap between Agile teams and risk-averse stakeholders.

In the increasingly AI and software driven world, having confidence in delivery timeframe is a competitive advantage. The power of Monte Carlo QSRA would bring high value into the Agile world, not about constraining agility, but to provide confidence with clarity.

Mozart Chan is a risk management professional specializing in project and risk management across Asia Pacific, the Middle East, and the UK. With certifications in PMP, PMI-RMP, and machine learning, he brings a unique blend of technical rigor and strategic insight to complex infrastructure and digital transformation projects.