Working in small batches is an essential principle in any discipline where feedback loops are important, or you want to learn quickly from your decisions. Working in small batches allows you to rapidly test hypotheses about whether a particular improvement is likely to have the effect you want, and if not, lets you course correct or revisit assumptions. Although this article applies to any type of change that includes organizational transformation and process improvement, it focuses primarily on software delivery.
Working in small batches is part of lean product management. Together with capabilities like visibility of work in the value stream, team experimentation, and visibility into customer feedback, working in small batches predicts software delivery performance and organizational performance.
One reason work is done in large batches is because of the large fixed cost of handing off changes. In traditional phased approaches to software development, handoffs from development to test or from test to IT operations consist of whole releases: months worth of work by teams consisting of tens or hundreds of people. With this traditional approach, collecting feedback on a change can take weeks or months.
In contrast, DevOps practices, which utilize cross-functional teams and lightweight approaches, allow for software to progress from development through test and operations into production in a matter of minutes. However, this rapid progression requires working with code in small batches.
Working in small batches has many benefits:
- It reduces the time it takes to get feedback on changes, making it easier to triage and remediate problems.
- It increases efficiency and motivation.
- It prevents your organization from succumbing to the sunk-cost fallacy.
You can apply the small batches approach at the feature and the product level. As an illustration, a minimum viable product, or MVP, is a prototype of a product with just enough features to enable validated learning about the product and its business model.
Continuous delivery builds upon working in small batches and tries to get every change in version control as early as possible. A goal of continuous delivery is to change the economics of the software delivery process, making it viable to work in small batches. This approach provides fast, comprehensive feedback to teams so that they can improve their work.
How to work in small batches
When you plan new features, try to break them down into work units that can be completed independently and in short timeframes. We recommend that each feature or batch of work follow the agile concept of the INVEST principle:
- Independent. Make batches of work as independent as possible from other batches, so that teams can work on them in any order, and deploy and validate them independent of other batches of work.
- Negotiable. Each batch of work is iterable and can be renegotiated as feedback is received.
- Valuable. Discrete batches of work are usable and provide value to the stakeholders.
- Estimable. Enough information exists about the batches of work that you can easily estimate the scope.
- Small. During a sprint, you should be able to complete batches of work in small increments of time, meaning hours to a couple days.
- Testable. Each batch of work can be tested, monitored, and verified as working in the way users expect.
When features are of an appropriate size, you can split the development of the feature into even smaller batches. This process can be difficult and requires experience to develop. Ideally, your developers should be checking multiple small releasable changes into trunk at least once per day.
The key is to start development at the service or API layer, not at the UI layer. In this way, you can make additions to the API that won't initially be available to users of the app, and check those changes into trunk. You can launch these changes to production without making them visible to users. This approach, called dark launching, allows developers to check in code for small batches that have been completed, but for features that are not yet fully complete. You can then run automated tests against these changes to prove that they behave in the expected way. This way, teams are still working quickly and developing off of master and not long-lived feature branches.
You can also dark launch changes by using a feature toggle, which is a conditional statement based on configuration settings. For example, you can make UI elements visible or invisible, or you can enable or disable service logic. Feature-toggle configuration might be read either at deploy time or runtime. You can use these configuration settings to switch the behavior of new code further down the stack. You can also use similar technique known as branch by abstraction to make larger-scale changes to the system while continuing to develop and release off-trunk without the use of long-lived feature branches.
In this approach, batches of work aren't complete until they're deployed to production and the feedback process has begun to validate the changes. Feedback comes from many sources and in many forms, including users, system monitoring, quality assurance, and automated tests. Your goal is to optimize for speed so that you reduce the cycle time to get changes into the hands of users. This way, you can validate your hypothesis as quickly as possible.
Common pitfalls with working in small batches
When you break down work into small batches, you encounter two pitfalls:
Not breaking up work into small enough pieces. Your first task is to break down the work in a meaningful way. We recommend that you commit code independent of the status of the feature and that individual features take no more than a few days to develop. Any batch of code that takes longer than a week to complete and check is too big. Throughout the development process, it's essential that you analyze how to break down ideas into increments that you can develop iteratively.
Working in small batches but then regrouping the batches before sending them downstream for testing or release. Regrouping work in this way delays the feedback on whether the changes have defects, and whether your users and your organization agree the changes were the right thing to build in the first place.
Ways to reduce the size of work batches
When you slice work into small batches that can be completed in hours, you can typically test and deploy those batches to production in less than an hour (PDF). The key is to decompose the work into small features that allow for rapid development, rather than developing complex features on branches and releasing them infrequently.
To improve small batch development, check your environment and confirm that the following conditions are true:
- Work is decomposed in a way that enables teams to make more frequent production releases.
- Developers are experienced in breaking down work into small changes that can be completed in the space of hours, not days.
To master small batch development, strive to meet each of these conditions in all of your development teams. This practice is a necessary condition for both continuous integration and trunk-based development.
Ways to measure the size of work batches
- Application features are decomposed in a way that supports frequent releases. How often are releases possible? How does this release cadence differ across teams? Are delays in production related to features that are larger?
- Application features are sliced in a way that lets developers complete the work in one week or less. What proportion of features can you complete in one week or less? What features can't you complete in one week or less? Can you commit and release changes before the feature is complete?
- MVPs are defined and set as goals for teams. Is the work decomposed into features that allow for MVPs and rapid development, rather than complex and lengthy processes?
Your measurements depend on the following:
- Knowing your organization's processes.
- Setting goals for reducing waste.
- Looking for ways to reduce complexity in the development process.
- For links to other articles and resources, see the DevOps page.