Estimating Epics in Agile: Predicting the Unpredictable
In Agile software development, Epics are large, complex work items that are broken down into smaller, more manageable user stories. Estimating the size of these Epics is critical to extending the planning horizon and managing expectations. In this blog post, we explore the key questions related to estimating Epics and present a systematic way to do so using Random Walk Monte Carlo (RWMC) simulation.
Why Do We Need to Estimate Epics?
Enable Long-Term Planning
Estimating the size of Epics is essential for long-term planning. With a rough estimate of how large an Epic is, project teams can:
Allocate Resources Efficiently: Knowing the size of an Epic helps teams allocate the right amount of resources, including personnel, budget, and time. It helps to avoid overcommitment or underutilization of resources.
Set Realistic Timelines: Estimating Epics helps in creating a timeline that is achievable. It guides when each Epic can be started and when it is likely to be completed, helping the teams to avoid last-minute rushes and the compromising of quality.
Understand Dependencies: A clear estimate of an Epic’s size allows teams to better understand how that Epic interacts with other tasks or Epics, enabling a clearer picture of task dependencies and helping in prioritizing work.
Quantify Risk
Estimating the size of an Epic helps in quantifying the risk associated with its delivery within a particular timeframe. It offers:
Insight into Completion Likelihood: An estimate provides a statistical basis for the likelihood of completing the Epic within the given timeline, allowing teams to assess and communicate risks more effectively.
Contingency Planning: By quantifying the risk, teams can develop contingency plans for Epics that have a high risk of delay or failure. This proactive approach helps to minimize the negative impact when things don’t go as planned.
Stakeholder Communication: Estimation allows for clearer communication with stakeholders about the risks associated with different stages of the project. It provides data-driven insights that can be used to set expectations and make informed decisions.
Cost vs Benefit Analysis
Understanding the size of an Epic is vital to assess if the effort invested justifies the value delivered. Estimating allows stakeholders to:
Weigh Costs Against Benefits: Before committing substantial resources to an Epic, it is essential to understand the expected returns. Estimation helps in analyzing if the potential value an Epic delivers aligns with the effort and cost required to complete it.
Prioritize Work Effectively: With a clear cost vs benefit analysis, teams can prioritize which Epics should be tackled first based on the value they are expected to deliver relative to their cost.
Make Informed Investment Decisions: For business stakeholders, understanding the size and associated cost of an Epic is critical for making investment decisions. It allows them to compare different initiatives and decide where to invest time and money for the best possible return.
How Do We Estimate Epics?
To illustrate the process, we’ll assume that we have to create 15 new Epics for which we have no clue how big they are or will be and use our existing project data to estimate them. We will use the RWMC simulation for this purpose.
Step 1: Gather Historical Data
Collect data from existing Epics that have already been broken down into stories or child work items. This data will be used as sample data for the RWMC simulation.
Step 2: Calculate Epic Sizes
For each existing Epic, calculate the number of child work items it comprises. These counts represent the sizes of the Epics.
Step 3: Run the RWMC Simulation
Use the Epic sizes obtained from step 2 as your sample data and run an RWMC simulation. This involves using historical data to generate a range of possible outcomes and their probabilities.
Step 4: Determine Confidence Intervals
After running the RWMC simulation, you need to establish percentile numbers that represent your team’s or organization’s appetite for risk. These percentiles help you to understand the potential range in which the actual size of an Epic might fall, given your comfort with risk.
For example, in this guide, we chose the 50th and 85th percentiles due to our specific appetite for risk. These numbers represent our comfort level with uncertainty:
The 50th percentile serves as a middle-ground estimate, where there is a 50% chance the actual size of an Epic will be higher and a 50% chance it will be lower.
The 85th percentile represents a more conservative estimate, where we are 85% confident that the actual size of the Epic will be at or below this level.
These are just examples, and your choice of percentiles should align with how much risk your team or organization is willing to accept. If your team is more risk-averse, you might choose a higher percentile (e.g., 90th or 95th). If your team is more comfortable with risk, a lower percentile (e.g., 70th or 80th) might be more appropriate.
It is important to have a discussion with your team and stakeholders about what these percentiles mean and how they will be used in planning and decision-making.
Step 5: Use in Overall Planning
These numbers are now ready to be used in overall planning, supporting decisions in terms of cost vs benefit, and aiding in resource allocation.
How Much Historical Data is Needed for the RWMC Simulation?
The confidence level in the predictions made through RWMC simulations hinges significantly on the amount of historical data, or sample size, used. As a rule of thumb, aim for a minimum sample size of 7 Epics to start. This gives a 75% chance that the actual size of the Epic will be at or below this level. The more comprehensive your sample data, the more reliable and robust your predictions will be. This is where Prediction Intervals come into play.
What are Prediction Intervals?
Prediction Intervals (PIs) are a range of values used in statistical modeling to predict future observations. Unlike confidence intervals, which estimate where the mean of the population lies, prediction intervals estimate the range within which a future observation will fall with a certain probability, given what we know from the existing data.
In the context of estimating the size of future Epics in an Agile project, Prediction Intervals give us a range of possible sizes (in terms of child work items) that a new Epic may have, based on the historical data we’ve collected.
For example, suppose a 95% Prediction Interval for the size of a new Epic is calculated to be between 10 and 20 work items. This means that we are 95% confident that the size of a new Epic will fall within this range.
The Importance of Sample Size
Here is a breakdown of the impact of sample size on Prediction Intervals:
- With child work item counts for 3 Epics, there is a 50% chance that a new Epic could fall outside of the size range of your existing Epics.
- With counts for 4 Epics, the probability of a new Epic being larger or smaller than the size ranges of existing Epics reduces to 40%.
The larger the sample size of historical Epics we have:
- The Narrower the Prediction Interval: More data leads to more precise predictions. As the sample size grows, the Prediction Interval narrows, indicating increased confidence in our predictions.
- The Lower the Risk of Erroneous Predictions: A larger sample size reduces the likelihood that our prediction will be incorrect.
Practical Tips
- Aim for a Larger Sample Size: More historical data improves the reliability of the RWMC simulation and the accuracy of the Prediction Intervals.
- Regularly Update the Model with New Data: As more Epics are completed, continually updating the model with fresh data ensures that predictions remain as accurate as possible.
Conclusion
Estimating the size of Epics in Agile projects is vital for long-term planning, risk quantification, and cost-benefit analysis. The RWMC simulation, when fed with adequate historical data, proves to be a powerful tool for making these estimations with a reasonable level of confidence. As a good practice, always aim to have a larger sample size for improved prediction accuracy and less risk. The key to leveraging this tool effectively lies in gathering a robust sample of historical data and adapting as new data becomes available.
Note: While Prediction Intervals provide a structured way to make future estimates, they are still based on statistical probabilities and past data. It is important to continually revisit and adjust these intervals as new data becomes available and the project progresses.
Advanced Agile Estimation, Planning and Forecasting
Take a deep dive into Story Points, Agile Estimation, Planning and Forecasting on our 1.5 day instructor led online course. Organisations often struggle to accurately estimate work, resulting in damaged stakeholder expectations, late delivery and demotivated staff. In this series of workshops you will learn how to accurately estimate Agile user stories, Features, Epics, projects and even entire programmes of work using the latest techniques from industry experts.
