Agentforce, the GenAI Agent by Salesforce
T-shirt sizing is an agile estimation technique that uses relative sizing to estimate IT project costs. While beneficial in the cost estimation process, it's also important to highlight key characteristics and best practices to ensure an appropriate design and full adoption of this technique.
IT project cost estimation has long presented difficulties and constraints for the IT departments of large corporations. Across industries and organizations, these issues have some common factors such as:
These factors are increasingly present in organizations, and one of the possible reasons is the fact that teams are still working according to the traditional Waterfall model, rather than adopting a more Agile way of working.
Many studies show that the cost estimate could be more accurate and closer to reality if more details about the requirements were available. By adopting an Agile-driven method, several levels and formats of estimations can be performed at different stages throughout the project lifecycle, when different levels of details become available. In general, there are four key stages in the cost estimation process:
Several methods can be used in the estimation of epics, user stories, and tasks. For instance, Functional Point Analysis, expert judgment, historical data analysis, planning poker, and story point analysis are just a few examples.
The most common approach for estimating the costs of projects is using absolute numbers (for example, hours or days).
Generally, the business elaborates an initial list of ideas and potential projects, which is reviewed and filtered by the BA teams. Once a clear list is defined, the Portfolio Project Management team is required to provide an estimation.
The challenge with this method is that, sometimes, project teams find themselves overanalyzing when trying to estimate a certain number of points or days, especially during a project’s initial stage and without many details available.
Lack of detailed information in the early stages of a project is quite common.
However, as we progress further in the project lifecycle process, more inputs are available and things get clearer. This is generally called the “cone of uncertainty”, which assumes that uncertainty should decrease throughout the project lifecycle.
There are several techniques and best practices that help improve the accuracy of cost estimation. One of them is to estimate in relative terms, rather than absolute terms. Estimating in relative terms is more effective as it gives an easy and effective way to identify the first estimate, generally chosen from a predefined list where we can compare different options.
This is exactly what a T-shirt sizing estimation provides.
T-shirt sizing | Sizing description |
---|---|
XS | Extra small |
S | Small |
M | Medium |
L | Large |
XL | Extra large |
Additional categories may be applied such as XXS or XXL, according to the type of business or the needs of each team.
For every defined T-shirt sizing, the associated range of work days or US dollars is quite specific to each organization and needs to be defined and agreed upon across teams. Each organization needs to establish the T-shirt sizing scale relative to the others. For example, for one organization, a medium may be twice as big as a small; for another organization, it may be three times as big. For one organization, for defined XS, the associated range could be 10K – 20K USD, but it could be defined as 50K – 100K USD for another organization. The common understanding of T-shirt sizing scales and associated ranges should be aligned among the projects and/or organizations when the changes in associated ranges are applied. The definitions of T-shirt sizing ranges are often based on past projects and historical information.
Within the T-shirt sizing estimate exercise, the portfolio management team is asked to estimate whether they think a project is extra-small, small, medium, large, extra-large, or even double extra-large. By removing the implied precision of a numerical score, the team is free to think in a more abstract way about the effort involved in a project. By comparing the projects, based on past experience and historical data, the project team can properly break them into defined buckets.
A T-shirt sizing estimate model should be first defined in order to perform T-shirt sizing estimate effectively. The T-shirt sizing estimate model is a combination of two estimation techniques: Parametric estimation and Historical information. There are three steps to define the T-shirt sizing estimate model tool.
Define the T-shirt sizing estimate model – The first step is to categorize the projects into different groups according to their core objective (e.g. new product, bug fix, etc.), with the help of SMEs and experienced people within the organization. For every type of group identified, different components and attributes can be defined which will represent the model input parameters and criteria for future users. This should result in a simple matrix where, for every attribute or criteria, a T-shirt size is assigned.
Validate the model with historical information – In order to ensure that the project categorization has been correctly defined, an additional check could be done using historical data of completed projects. This step is based on the important assumption that historical information and records are available and easily collectible within the organization. During the validation process, the model can be modified or adjusted if needed.
Perform the T-shirt sizing estimate on new projects – Once the T-shirt sizing estimate model is finalized and validated, a pilot can be launched within a small scope of new project demands. The first users should be trained and assisted during this phase, to ensure full adoption of the model. Feedback and remarks should be collected, in order to make further improvements in the future.
As an example, consider a simple project request for a bug fix. The size of such a project could be based on the number of bug fixes to be done:
T-shirt sizing | Bug Fix Volume |
---|---|
XXS | 1-3 bug fix |
XS | 4-6 bug fix |
S | 7-9 bug fix |
M | 10 bug fix and above |
There are immediate benefits following the adoption of the T-shirt sizing estimate approach, and the use of a structured model (designed in Excel or any other suitable software). The main benefits are the following:
Even if the T-shirt sizing estimation model can be a useful tool, there are a few aspects that must be kept in mind, in order to avoid common pitfalls and mistakes. Here are some key points of attention:
In software-related project reality, the portfolio management team becomes more certain of the estimate as they learn more about what they are estimating. T-shirt sizing is an easy way to approach an initiative and is most commonly performed at the initial stages, to address the purpose of strategic planning, which requires very high-level estimates in order to calculate the ROI among different initiatives. T-shirt sizing helps executives and managers get a sense of scale in order to identify the key priorities in the project portfolio. From this perspective, T-shirt sizing is a worthy approach to getting started with relative estimation, but there are practical issues to consider when adopting it and defining the model. The ability to extract the generic and common project attributes for comparison among projects can be complicated since there is no evident mathematical relationship, and it requires engaging expert judgments from more experienced stakeholders.
In addition, with the larger adoption of SAFE, the Lean Agile framework for delivering projects for multiple teams, being able to estimate across teams is paramount.
Sia Partners has developed a strong background and experience on Agile Transformation programs, including initiatives to optimize the project cost estimation process. Through an analytical approach, we have assisted our clients in identifying the key pain points in the cost estimation practice through the end-to-end process. Our specialized teams have helped the core teams in elaborating simple but effective cost estimate models, by involving key users from the beginning, ensuring strong support and adoption in different teams.